diff --git "a/IE-en/RE/SciERC/test.json" "b/IE-en/RE/SciERC/test.json"
new file mode 100644--- /dev/null
+++ "b/IE-en/RE/SciERC/test.json"
@@ -0,0 +1,397 @@
+{"text": "Recognition of proper nouns in Japanese text has been studied as a part of the more general problem of morphological analysis in Japanese text processing -LRB- -LSB- 1 -RSB- -LSB- 2 -RSB- -RRB- .", "relation": [{"head": "Recognition of proper nouns", "head_type": "NA", "head_pos": [3, 30], "relation": "part of", "tail": "morphological analysis", "tail_type": "NA", "tail_pos": [112, 134]}, {"head": "proper nouns", "head_type": "NA", "head_pos": [18, 30], "relation": "part of", "tail": "Japanese text", "tail_type": "NA", "tail_pos": [40, 53]}, {"head": "morphological analysis", "head_type": "NA", "head_pos": [106, 128], "relation": "used for", "tail": "Japanese text processing", "tail_type": "NA", "tail_pos": [138, 162]}], "task": "RE"}
+{"text": "It has also been studied in the framework of Japanese information extraction -LRB- -LSB- 3 -RSB- -RRB- in recent years .", "relation": [{"head": "Japanese information extraction", "head_type": "NA", "head_pos": [54, 85], "relation": "used for", "tail": "It", "tail_type": "NA", "tail_pos": [3, 5]}], "task": "RE"}
+{"text": "Our approach to the Multi-lingual Evaluation Task -LRB- MET -RRB- for Japanese text is to consider the given task as a morphological analysis problem in Japanese .", "relation": [{"head": "approach", "head_type": "NA", "head_pos": [7, 15], "relation": "used for", "tail": "Multi-lingual Evaluation Task -LRB- MET -RRB-", "tail_type": "NA", "tail_pos": [29, 74]}, {"head": "Multi-lingual Evaluation Task -LRB- MET -RRB-", "head_type": "NA", "head_pos": [23, 68], "relation": "used for", "tail": "Japanese text", "tail_type": "NA", "tail_pos": [79, 92]}, {"head": "morphological analysis problem", "head_type": "NA", "head_pos": [128, 158], "relation": "used for", "tail": "task", "tail_type": "NA", "tail_pos": [112, 116]}, {"head": "Japanese", "head_type": "NA", "head_pos": [70, 78], "relation": "used for", "tail": "morphological analysis problem", "tail_type": "NA", "tail_pos": [122, 152]}], "task": "RE"}
+{"text": "Our morphological analyzer has done all the necessary work for the recognition and classification of proper names , numerical and temporal expressions , i.e. Named Entity -LRB- NE -RRB- items in the Japanese text .", "relation": [{"head": "morphological analyzer", "head_type": "NA", "head_pos": [7, 29], "relation": "used for", "tail": "recognition and classification of proper names , numerical and temporal expressions", "tail_type": "NA", "tail_pos": [76, 159]}, {"head": "Named Entity -LRB- NE -RRB- items", "head_type": "NA", "head_pos": [167, 200], "relation": "hyponym of", "tail": "proper names , numerical and temporal expressions", "tail_type": "NA", "tail_pos": [104, 153]}, {"head": "Named Entity -LRB- NE -RRB- items", "head_type": "NA", "head_pos": [161, 194], "relation": "part of", "tail": "Japanese text", "tail_type": "NA", "tail_pos": [208, 221]}], "task": "RE"}
+{"text": "Amorph recognizes NE items in two stages : dictionary lookup and rule application .", "relation": [{"head": "Amorph", "head_type": "NA", "head_pos": [3, 9], "relation": "used for", "tail": "NE items", "tail_type": "NA", "tail_pos": [27, 35]}, {"head": "dictionary lookup", "head_type": "NA", "head_pos": [52, 69], "relation": "part of", "tail": "Amorph", "tail_type": "NA", "tail_pos": [3, 9]}, {"head": "dictionary lookup", "head_type": "NA", "head_pos": [46, 63], "relation": "conjunction", "tail": "rule application", "tail_type": "NA", "tail_pos": [74, 90]}, {"head": "rule application", "head_type": "NA", "head_pos": [74, 90], "relation": "part of", "tail": "Amorph", "tail_type": "NA", "tail_pos": [3, 9]}], "task": "RE"}
+{"text": "First , it uses several kinds of dictionaries to segment and tag Japanese character strings .", "relation": [{"head": "dictionaries", "head_type": "NA", "head_pos": [42, 54], "relation": "used for", "tail": "it", "tail_type": "NA", "tail_pos": [11, 13]}, {"head": "dictionaries", "head_type": "NA", "head_pos": [36, 48], "relation": "used for", "tail": "Japanese character strings", "tail_type": "NA", "tail_pos": [74, 100]}], "task": "RE"}
+{"text": "Second , based on the information resulting from the dictionary lookup stage , a set of rules is applied to the segmented strings in order to identify NE items .", "relation": [{"head": "dictionary lookup stage", "head_type": "NA", "head_pos": [56, 79], "relation": "used for", "tail": "rules", "tail_type": "NA", "tail_pos": [97, 102]}, {"head": "rules", "head_type": "NA", "head_pos": [91, 96], "relation": "used for", "tail": "NE items", "tail_type": "NA", "tail_pos": [160, 168]}], "task": "RE"}
+{"text": "We propose to incorporate a priori geometric constraints in a 3 -- D stereo reconstruction scheme to cope with the many cases where image information alone is not sufficient to accurately recover 3 -- D shape .", "relation": [{"head": "priori geometric constraints", "head_type": "NA", "head_pos": [31, 59], "relation": "part of", "tail": "3 -- D stereo reconstruction scheme", "tail_type": "NA", "tail_pos": [71, 106]}, {"head": "image information", "head_type": "NA", "head_pos": [135, 152], "relation": "used for", "tail": "3 -- D shape", "tail_type": "NA", "tail_pos": [205, 217]}], "task": "RE"}
+{"text": "Our approach is based on the iterative deformation of a 3 -- D surface mesh to minimize an objective function .", "relation": [{"head": "iterative deformation of a 3 -- D surface mesh", "head_type": "NA", "head_pos": [38, 84], "relation": "used for", "tail": "approach", "tail_type": "NA", "tail_pos": [7, 15]}, {"head": "iterative deformation of a 3 -- D surface mesh", "head_type": "NA", "head_pos": [32, 78], "relation": "used for", "tail": "objective function", "tail_type": "NA", "tail_pos": [100, 118]}], "task": "RE"}
+{"text": "We show that combining anisotropic meshing with a non-quadratic approach to regularization enables us to obtain satisfactory reconstruction results using triangulations with few vertices .", "relation": [{"head": "anisotropic meshing", "head_type": "NA", "head_pos": [26, 45], "relation": "conjunction", "tail": "non-quadratic approach", "tail_type": "NA", "tail_pos": [59, 81]}, {"head": "anisotropic meshing", "head_type": "NA", "head_pos": [26, 45], "relation": "used for", "tail": "reconstruction", "tail_type": "NA", "tail_pos": [134, 148]}, {"head": "non-quadratic approach", "head_type": "NA", "head_pos": [53, 75], "relation": "used for", "tail": "regularization", "tail_type": "NA", "tail_pos": [85, 99]}, {"head": "non-quadratic approach", "head_type": "NA", "head_pos": [53, 75], "relation": "used for", "tail": "reconstruction", "tail_type": "NA", "tail_pos": [134, 148]}, {"head": "triangulations", "head_type": "NA", "head_pos": [163, 177], "relation": "used for", "tail": "reconstruction", "tail_type": "NA", "tail_pos": [128, 142]}], "task": "RE"}
+{"text": "Structural or numerical constraints can then be added locally to the reconstruction process through a constrained optimization scheme .", "relation": [{"head": "Structural or numerical constraints", "head_type": "NA", "head_pos": [3, 38], "relation": "used for", "tail": "reconstruction process", "tail_type": "NA", "tail_pos": [78, 100]}, {"head": "constrained optimization scheme", "head_type": "NA", "head_pos": [111, 142], "relation": "used for", "tail": "Structural or numerical constraints", "tail_type": "NA", "tail_pos": [3, 38]}], "task": "RE"}
+{"text": "They improve the reconstruction results and enforce their consistency with a priori knowledge about object shape .", "relation": [{"head": "They", "head_type": "NA", "head_pos": [3, 7], "relation": "used for", "tail": "reconstruction", "tail_type": "NA", "tail_pos": [26, 40]}, {"head": "object shape", "head_type": "NA", "head_pos": [109, 121], "relation": "feature of", "tail": "priori knowledge", "tail_type": "NA", "tail_pos": [80, 96]}], "task": "RE"}
+{"text": "The strong description and modeling properties of differential features make them useful tools that can be efficiently used as constraints for 3 -- D reconstruction .", "relation": [{"head": "them", "head_type": "NA", "head_pos": [80, 84], "relation": "used for", "tail": "3 -- D reconstruction", "tail_type": "NA", "tail_pos": [152, 173]}], "task": "RE"}
+{"text": "It is based on a weakly supervised dependency parser that can model speech syntax without relying on any annotated training corpus .", "relation": [{"head": "weakly supervised dependency parser", "head_type": "NA", "head_pos": [20, 55], "relation": "used for", "tail": "speech syntax", "tail_type": "NA", "tail_pos": [77, 90]}], "task": "RE"}
+{"text": "Labeled data is replaced by a few hand-crafted rules that encode basic syntactic knowledge .", "relation": [{"head": "hand-crafted rules", "head_type": "NA", "head_pos": [37, 55], "relation": "used for", "tail": "syntactic knowledge", "tail_type": "NA", "tail_pos": [80, 99]}], "task": "RE"}
+{"text": "Bayesian inference then samples the rules , disambiguating and combining them to create complex tree structures that maximize a discriminative model 's posterior on a target unlabeled corpus .", "relation": [{"head": "Bayesian inference", "head_type": "NA", "head_pos": [3, 21], "relation": "used for", "tail": "rules", "tail_type": "NA", "tail_pos": [45, 50]}, {"head": "them", "head_type": "NA", "head_pos": [76, 80], "relation": "used for", "tail": "complex tree structures", "tail_type": "NA", "tail_pos": [97, 120]}, {"head": "complex tree structures", "head_type": "NA", "head_pos": [91, 114], "relation": "used for", "tail": "discriminative model 's posterior", "tail_type": "NA", "tail_pos": [137, 170]}, {"head": "unlabeled corpus", "head_type": "NA", "head_pos": [183, 199], "relation": "used for", "tail": "discriminative model 's posterior", "tail_type": "NA", "tail_pos": [131, 164]}], "task": "RE"}
+{"text": "This posterior encodes sparse se-lectional preferences between a head word and its dependents .", "relation": [{"head": "posterior", "head_type": "NA", "head_pos": [8, 17], "relation": "used for", "tail": "sparse se-lectional preferences", "tail_type": "NA", "tail_pos": [32, 63]}], "task": "RE"}
+{"text": "The model is evaluated on English and Czech newspaper texts , and is then validated on French broadcast news transcriptions .", "relation": [{"head": "English and Czech newspaper texts", "head_type": "NA", "head_pos": [35, 68], "relation": "evaluate for", "tail": "model", "tail_type": "NA", "tail_pos": [7, 12]}, {"head": "French broadcast news transcriptions", "head_type": "NA", "head_pos": [96, 132], "relation": "evaluate for", "tail": "model", "tail_type": "NA", "tail_pos": [7, 12]}], "task": "RE"}
+{"text": "Listen-Communicate-Show -LRB- LCS -RRB- is a new paradigm for human interaction with data sources .", "relation": [{"head": "Listen-Communicate-Show -LRB- LCS -RRB-", "head_type": "NA", "head_pos": [3, 42], "relation": "used for", "tail": "human interaction with data sources", "tail_type": "NA", "tail_pos": [71, 106]}], "task": "RE"}
+{"text": "We integrate a spoken language understanding system with intelligent mobile agents that mediate between users and information sources .", "relation": [{"head": "intelligent mobile agents", "head_type": "NA", "head_pos": [66, 91], "relation": "part of", "tail": "spoken language understanding system", "tail_type": "NA", "tail_pos": [18, 54]}], "task": "RE"}
+{"text": "We have built and will demonstrate an application of this approach called LCS-Marine .", "relation": [{"head": "approach", "head_type": "NA", "head_pos": [61, 69], "relation": "used for", "tail": "LCS-Marine", "tail_type": "NA", "tail_pos": [83, 93]}], "task": "RE"}
+{"text": "A domain independent model is proposed for the automated interpretation of nominal compounds in English .", "relation": [{"head": "domain independent model", "head_type": "NA", "head_pos": [5, 29], "relation": "used for", "tail": "automated interpretation of nominal compounds", "tail_type": "NA", "tail_pos": [56, 101]}, {"head": "English", "head_type": "NA", "head_pos": [105, 112], "relation": "feature of", "tail": "nominal compounds", "tail_type": "NA", "tail_pos": [78, 95]}], "task": "RE"}
+{"text": "This model is meant to account for productive rules of interpretation which are inferred from the morpho-syntactic and semantic characteristics of the nominal constituents .", "relation": [{"head": "model", "head_type": "NA", "head_pos": [8, 13], "relation": "used for", "tail": "productive rules of interpretation", "tail_type": "NA", "tail_pos": [44, 78]}, {"head": "morpho-syntactic and semantic characteristics", "head_type": "NA", "head_pos": [107, 152], "relation": "used for", "tail": "productive rules of interpretation", "tail_type": "NA", "tail_pos": [38, 72]}, {"head": "morpho-syntactic and semantic characteristics", "head_type": "NA", "head_pos": [101, 146], "relation": "feature of", "tail": "nominal constituents", "tail_type": "NA", "tail_pos": [160, 180]}], "task": "RE"}
+{"text": "In particular , we make extensive use of Pustejovsky 's principles concerning the predicative information associated with nominals .", "relation": [{"head": "nominals", "head_type": "NA", "head_pos": [131, 139], "relation": "feature of", "tail": "predicative information", "tail_type": "NA", "tail_pos": [85, 108]}], "task": "RE"}
+{"text": "We argue that it is necessary to draw a line between generalizable semantic principles and domain-specific semantic information .", "relation": [{"head": "generalizable semantic principles", "head_type": "NA", "head_pos": [56, 89], "relation": "compare", "tail": "domain-specific semantic information", "tail_type": "NA", "tail_pos": [100, 136]}], "task": "RE"}
+{"text": "We explain this distinction and we show how this model may be applied to the interpretation of compounds in real texts , provided that complementary semantic information are retrieved .", "relation": [{"head": "model", "head_type": "NA", "head_pos": [52, 57], "relation": "used for", "tail": "interpretation of compounds", "tail_type": "NA", "tail_pos": [86, 113]}], "task": "RE"}
+{"text": "We present a new method for detecting interest points using histogram information .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [20, 26], "relation": "used for", "tail": "detecting interest points", "tail_type": "NA", "tail_pos": [37, 62]}, {"head": "histogram information", "head_type": "NA", "head_pos": [69, 90], "relation": "used for", "tail": "detecting interest points", "tail_type": "NA", "tail_pos": [31, 56]}], "task": "RE"}
+{"text": "Unlike existing interest point detectors , which measure pixel-wise differences in image intensity , our detectors incorporate histogram-based representations , and thus can find image regions that present a distinct distribution in the neighborhood .", "relation": [{"head": "pixel-wise differences in image intensity", "head_type": "NA", "head_pos": [66, 107], "relation": "evaluate for", "tail": "interest point detectors", "tail_type": "NA", "tail_pos": [19, 43]}, {"head": "histogram-based representations", "head_type": "NA", "head_pos": [136, 167], "relation": "part of", "tail": "detectors", "tail_type": "NA", "tail_pos": [31, 40]}], "task": "RE"}
+{"text": "The proposed detectors are able to capture large-scale structures and distinctive textured patterns , and exhibit strong invariance to rotation , illumination variation , and blur .", "relation": [{"head": "detectors", "head_type": "NA", "head_pos": [16, 25], "relation": "used for", "tail": "large-scale structures", "tail_type": "NA", "tail_pos": [52, 74]}, {"head": "detectors", "head_type": "NA", "head_pos": [16, 25], "relation": "used for", "tail": "distinctive textured patterns", "tail_type": "NA", "tail_pos": [79, 108]}, {"head": "detectors", "head_type": "NA", "head_pos": [16, 25], "relation": "used for", "tail": "rotation", "tail_type": "NA", "tail_pos": [144, 152]}, {"head": "detectors", "head_type": "NA", "head_pos": [16, 25], "relation": "used for", "tail": "illumination variation", "tail_type": "NA", "tail_pos": [155, 177]}, {"head": "detectors", "head_type": "NA", "head_pos": [16, 25], "relation": "used for", "tail": "blur", "tail_type": "NA", "tail_pos": [184, 188]}, {"head": "large-scale structures", "head_type": "NA", "head_pos": [46, 68], "relation": "conjunction", "tail": "distinctive textured patterns", "tail_type": "NA", "tail_pos": [79, 108]}, {"head": "rotation", "head_type": "NA", "head_pos": [138, 146], "relation": "conjunction", "tail": "illumination variation", "tail_type": "NA", "tail_pos": [155, 177]}, {"head": "illumination variation", "head_type": "NA", "head_pos": [149, 171], "relation": "conjunction", "tail": "blur", "tail_type": "NA", "tail_pos": [184, 188]}], "task": "RE"}
+{"text": "The experimental results show that the proposed histogram-based interest point detectors perform particularly well for the tasks of matching textured scenes under blur and illumination changes , in terms of repeatability and distinctiveness .", "relation": [{"head": "histogram-based interest point detectors", "head_type": "NA", "head_pos": [51, 91], "relation": "used for", "tail": "matching textured scenes", "tail_type": "NA", "tail_pos": [141, 165]}, {"head": "repeatability", "head_type": "NA", "head_pos": [216, 229], "relation": "evaluate for", "tail": "histogram-based interest point detectors", "tail_type": "NA", "tail_pos": [51, 91]}, {"head": "repeatability", "head_type": "NA", "head_pos": [210, 223], "relation": "conjunction", "tail": "distinctiveness", "tail_type": "NA", "tail_pos": [234, 249]}, {"head": "distinctiveness", "head_type": "NA", "head_pos": [234, 249], "relation": "evaluate for", "tail": "histogram-based interest point detectors", "tail_type": "NA", "tail_pos": [51, 91]}], "task": "RE"}
+{"text": "An extension of our method to space-time interest point detection for action classification is also presented .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [23, 29], "relation": "used for", "tail": "space-time interest point detection", "tail_type": "NA", "tail_pos": [39, 74]}, {"head": "space-time interest point detection", "head_type": "NA", "head_pos": [33, 68], "relation": "used for", "tail": "action classification", "tail_type": "NA", "tail_pos": [79, 100]}], "task": "RE"}
+{"text": "We have implemented a restricted domain parser called Plume .", "relation": [{"head": "Plume", "head_type": "NA", "head_pos": [63, 68], "relation": "hyponym of", "tail": "restricted domain parser", "tail_type": "NA", "tail_pos": [25, 49]}], "task": "RE"}
+{"text": "Building on previous work at Carnegie-Mellon University e.g. -LSB- 4 , 5 , 8 -RSB- , Plume 's approach to parsing is based on semantic caseframe instantiation .", "relation": [{"head": "Plume 's approach", "head_type": "NA", "head_pos": [88, 105], "relation": "used for", "tail": "parsing", "tail_type": "NA", "tail_pos": [115, 122]}, {"head": "semantic caseframe instantiation", "head_type": "NA", "head_pos": [135, 167], "relation": "used for", "tail": "Plume 's approach", "tail_type": "NA", "tail_pos": [88, 105]}], "task": "RE"}
+{"text": "This has the advantages of efficiency on grammatical input , and robustness in the face of ungrammatical input .", "relation": [{"head": "ungrammatical input", "head_type": "NA", "head_pos": [100, 119], "relation": "feature of", "tail": "robustness", "tail_type": "NA", "tail_pos": [68, 78]}], "task": "RE"}
+{"text": "While Plume is well adapted to simple declarative and imperative utterances , it handles passives , relative clauses and interrogatives in an ad hoc manner leading to patchy syntactic coverage .", "relation": [{"head": "Plume", "head_type": "NA", "head_pos": [9, 14], "relation": "used for", "tail": "declarative and imperative utterances", "tail_type": "NA", "tail_pos": [47, 84]}, {"head": "it", "head_type": "NA", "head_pos": [81, 83], "relation": "used for", "tail": "passives", "tail_type": "NA", "tail_pos": [98, 106]}, {"head": "it", "head_type": "NA", "head_pos": [81, 83], "relation": "used for", "tail": "relative clauses", "tail_type": "NA", "tail_pos": [109, 125]}, {"head": "it", "head_type": "NA", "head_pos": [81, 83], "relation": "used for", "tail": "interrogatives", "tail_type": "NA", "tail_pos": [130, 144]}, {"head": "passives", "head_type": "NA", "head_pos": [92, 100], "relation": "conjunction", "tail": "relative clauses", "tail_type": "NA", "tail_pos": [109, 125]}, {"head": "relative clauses", "head_type": "NA", "head_pos": [103, 119], "relation": "conjunction", "tail": "interrogatives", "tail_type": "NA", "tail_pos": [130, 144]}], "task": "RE"}
+{"text": "This paper outlines Plume as it currently exists and describes our detailed design for extending Plume to handle passives , relative clauses , and interrogatives in a general manner .", "relation": [{"head": "Plume", "head_type": "NA", "head_pos": [20, 25], "relation": "used for", "tail": "passives", "tail_type": "NA", "tail_pos": [122, 130]}, {"head": "Plume", "head_type": "NA", "head_pos": [20, 25], "relation": "used for", "tail": "relative clauses", "tail_type": "NA", "tail_pos": [133, 149]}, {"head": "Plume", "head_type": "NA", "head_pos": [20, 25], "relation": "used for", "tail": "interrogatives", "tail_type": "NA", "tail_pos": [156, 170]}, {"head": "passives", "head_type": "NA", "head_pos": [116, 124], "relation": "conjunction", "tail": "relative clauses", "tail_type": "NA", "tail_pos": [133, 149]}, {"head": "relative clauses", "head_type": "NA", "head_pos": [127, 143], "relation": "conjunction", "tail": "interrogatives", "tail_type": "NA", "tail_pos": [156, 170]}], "task": "RE"}
+{"text": "In this paper , we present an unlexicalized parser for German which employs smoothing and suffix analysis to achieve a labelled bracket F-score of 76.2 , higher than previously reported results on the NEGRA corpus .", "relation": [{"head": "unlexicalized parser", "head_type": "NA", "head_pos": [33, 53], "relation": "used for", "tail": "German", "tail_type": "NA", "tail_pos": [64, 70]}, {"head": "smoothing", "head_type": "NA", "head_pos": [85, 94], "relation": "used for", "tail": "unlexicalized parser", "tail_type": "NA", "tail_pos": [33, 53]}, {"head": "smoothing", "head_type": "NA", "head_pos": [79, 88], "relation": "conjunction", "tail": "suffix analysis", "tail_type": "NA", "tail_pos": [99, 114]}, {"head": "suffix analysis", "head_type": "NA", "head_pos": [99, 114], "relation": "used for", "tail": "unlexicalized parser", "tail_type": "NA", "tail_pos": [33, 53]}, {"head": "labelled bracket F-score", "head_type": "NA", "head_pos": [128, 152], "relation": "evaluate for", "tail": "unlexicalized parser", "tail_type": "NA", "tail_pos": [33, 53]}, {"head": "NEGRA corpus", "head_type": "NA", "head_pos": [210, 222], "relation": "evaluate for", "tail": "unlexicalized parser", "tail_type": "NA", "tail_pos": [33, 53]}], "task": "RE"}
+{"text": "In addition to the high accuracy of the model , the use of smoothing in an unlexicalized parser allows us to better examine the interplay between smoothing and parsing results .", "relation": [{"head": "accuracy", "head_type": "NA", "head_pos": [27, 35], "relation": "evaluate for", "tail": "model", "tail_type": "NA", "tail_pos": [49, 54]}, {"head": "smoothing", "head_type": "NA", "head_pos": [62, 71], "relation": "used for", "tail": "unlexicalized parser", "tail_type": "NA", "tail_pos": [84, 104]}], "task": "RE"}
+{"text": "This paper presents an unsupervised learning approach to disambiguate various relations between named entities by use of various lexical and syntactic features from the contexts .", "relation": [{"head": "unsupervised learning approach", "head_type": "NA", "head_pos": [26, 56], "relation": "used for", "tail": "relations between named entities", "tail_type": "NA", "tail_pos": [87, 119]}, {"head": "lexical and syntactic features", "head_type": "NA", "head_pos": [138, 168], "relation": "used for", "tail": "unsupervised learning approach", "tail_type": "NA", "tail_pos": [26, 56]}], "task": "RE"}
+{"text": "It works by calculating eigenvectors of an adjacency graph 's Laplacian to recover a submanifold of data from a high dimensionality space and then performing cluster number estimation on the eigenvectors .", "relation": [{"head": "It", "head_type": "NA", "head_pos": [3, 5], "relation": "used for", "tail": "submanifold", "tail_type": "NA", "tail_pos": [94, 105]}, {"head": "eigenvectors", "head_type": "NA", "head_pos": [33, 45], "relation": "used for", "tail": "It", "tail_type": "NA", "tail_pos": [3, 5]}, {"head": "adjacency graph 's Laplacian", "head_type": "NA", "head_pos": [52, 80], "relation": "feature of", "tail": "eigenvectors", "tail_type": "NA", "tail_pos": [27, 39]}, {"head": "high dimensionality space", "head_type": "NA", "head_pos": [121, 146], "relation": "used for", "tail": "submanifold", "tail_type": "NA", "tail_pos": [88, 99]}, {"head": "cluster number estimation", "head_type": "NA", "head_pos": [167, 192], "relation": "used for", "tail": "It", "tail_type": "NA", "tail_pos": [3, 5]}, {"head": "cluster number estimation", "head_type": "NA", "head_pos": [161, 186], "relation": "used for", "tail": "eigenvectors", "tail_type": "NA", "tail_pos": [24, 36]}], "task": "RE"}
+{"text": "Experiment results on ACE corpora show that this spectral clustering based approach outperforms the other clustering methods .", "relation": [{"head": "ACE corpora", "head_type": "NA", "head_pos": [25, 36], "relation": "evaluate for", "tail": "spectral clustering based approach", "tail_type": "NA", "tail_pos": [58, 92]}, {"head": "ACE corpora", "head_type": "NA", "head_pos": [25, 36], "relation": "evaluate for", "tail": "clustering methods", "tail_type": "NA", "tail_pos": [115, 133]}, {"head": "spectral clustering based approach", "head_type": "NA", "head_pos": [52, 86], "relation": "compare", "tail": "clustering methods", "tail_type": "NA", "tail_pos": [115, 133]}], "task": "RE"}
+{"text": "This paper proposes a generic mathematical formalism for the combination of various structures : strings , trees , dags , graphs , and products of them .", "relation": [{"head": "strings", "head_type": "NA", "head_pos": [106, 113], "relation": "hyponym of", "tail": "structures", "tail_type": "NA", "tail_pos": [87, 97]}, {"head": "strings", "head_type": "NA", "head_pos": [100, 107], "relation": "conjunction", "tail": "trees", "tail_type": "NA", "tail_pos": [116, 121]}, {"head": "trees", "head_type": "NA", "head_pos": [116, 121], "relation": "hyponym of", "tail": "structures", "tail_type": "NA", "tail_pos": [87, 97]}, {"head": "trees", "head_type": "NA", "head_pos": [110, 115], "relation": "conjunction", "tail": "dags", "tail_type": "NA", "tail_pos": [124, 128]}, {"head": "dags", "head_type": "NA", "head_pos": [124, 128], "relation": "hyponym of", "tail": "structures", "tail_type": "NA", "tail_pos": [87, 97]}, {"head": "dags", "head_type": "NA", "head_pos": [118, 122], "relation": "conjunction", "tail": "graphs", "tail_type": "NA", "tail_pos": [131, 137]}, {"head": "graphs", "head_type": "NA", "head_pos": [131, 137], "relation": "hyponym of", "tail": "structures", "tail_type": "NA", "tail_pos": [87, 97]}], "task": "RE"}
+{"text": "This formalism is both elementary and powerful enough to strongly simulate many grammar formalisms , such as rewriting systems , dependency grammars , TAG , HPSG and LFG .", "relation": [{"head": "formalism", "head_type": "NA", "head_pos": [8, 17], "relation": "used for", "tail": "grammar formalisms", "tail_type": "NA", "tail_pos": [89, 107]}, {"head": "rewriting systems", "head_type": "NA", "head_pos": [118, 135], "relation": "hyponym of", "tail": "grammar formalisms", "tail_type": "NA", "tail_pos": [83, 101]}, {"head": "rewriting systems", "head_type": "NA", "head_pos": [112, 129], "relation": "conjunction", "tail": "dependency grammars", "tail_type": "NA", "tail_pos": [138, 157]}, {"head": "dependency grammars", "head_type": "NA", "head_pos": [138, 157], "relation": "hyponym of", "tail": "grammar formalisms", "tail_type": "NA", "tail_pos": [83, 101]}, {"head": "dependency grammars", "head_type": "NA", "head_pos": [132, 151], "relation": "conjunction", "tail": "TAG", "tail_type": "NA", "tail_pos": [160, 163]}, {"head": "TAG", "head_type": "NA", "head_pos": [160, 163], "relation": "hyponym of", "tail": "grammar formalisms", "tail_type": "NA", "tail_pos": [83, 101]}, {"head": "TAG", "head_type": "NA", "head_pos": [154, 157], "relation": "conjunction", "tail": "HPSG", "tail_type": "NA", "tail_pos": [166, 170]}, {"head": "HPSG", "head_type": "NA", "head_pos": [166, 170], "relation": "hyponym of", "tail": "grammar formalisms", "tail_type": "NA", "tail_pos": [83, 101]}, {"head": "HPSG", "head_type": "NA", "head_pos": [160, 164], "relation": "conjunction", "tail": "LFG", "tail_type": "NA", "tail_pos": [175, 178]}, {"head": "LFG", "head_type": "NA", "head_pos": [175, 178], "relation": "hyponym of", "tail": "grammar formalisms", "tail_type": "NA", "tail_pos": [83, 101]}], "task": "RE"}
+{"text": "A mixed-signal paradigm is presented for high-resolution parallel inner-product computation in very high dimensions , suitable for efficient implementation of kernels in image processing .", "relation": [{"head": "mixed-signal paradigm", "head_type": "NA", "head_pos": [5, 26], "relation": "used for", "tail": "high-resolution parallel inner-product computation", "tail_type": "NA", "tail_pos": [50, 100]}, {"head": "mixed-signal paradigm", "head_type": "NA", "head_pos": [5, 26], "relation": "used for", "tail": "kernels", "tail_type": "NA", "tail_pos": [168, 175]}, {"head": "kernels", "head_type": "NA", "head_pos": [162, 169], "relation": "used for", "tail": "image processing", "tail_type": "NA", "tail_pos": [179, 195]}], "task": "RE"}
+{"text": "At the core of the externally digital architecture is a high-density , low-power analog array performing binary-binary partial matrix-vector multiplication .", "relation": [{"head": "high-density , low-power analog array", "head_type": "NA", "head_pos": [65, 102], "relation": "part of", "tail": "externally digital architecture", "tail_type": "NA", "tail_pos": [22, 53]}, {"head": "binary-binary partial matrix-vector multiplication", "head_type": "NA", "head_pos": [114, 164], "relation": "used for", "tail": "high-density , low-power analog array", "tail_type": "NA", "tail_pos": [59, 96]}], "task": "RE"}
+{"text": "Full digital resolution is maintained even with low-resolution analog-to-digital conversion , owing to random statistics in the analog summation of binary products .", "relation": [{"head": "random statistics", "head_type": "NA", "head_pos": [106, 123], "relation": "part of", "tail": "analog summation of binary products", "tail_type": "NA", "tail_pos": [137, 172]}], "task": "RE"}
+{"text": "A random modulation scheme produces near-Bernoulli statistics even for highly correlated inputs .", "relation": [{"head": "random modulation scheme", "head_type": "NA", "head_pos": [5, 29], "relation": "used for", "tail": "near-Bernoulli statistics", "tail_type": "NA", "tail_pos": [45, 70]}, {"head": "highly correlated inputs", "head_type": "NA", "head_pos": [80, 104], "relation": "used for", "tail": "random modulation scheme", "tail_type": "NA", "tail_pos": [5, 29]}], "task": "RE"}
+{"text": "The approach is validated with real image data , and with experimental results from a CID/DRAM analog array prototype in 0.5 cents m CMOS .", "relation": [{"head": "real image data", "head_type": "NA", "head_pos": [40, 55], "relation": "evaluate for", "tail": "approach", "tail_type": "NA", "tail_pos": [7, 15]}], "task": "RE"}
+{"text": "In this paper we specialize the projective unifocal , bifo-cal , and trifocal tensors to the affine case , and show how the tensors obtained relate to the registered tensors encountered in previous work .", "relation": [{"head": "projective unifocal , bifo-cal , and trifocal tensors", "head_type": "NA", "head_pos": [35, 88], "relation": "used for", "tail": "affine case", "tail_type": "NA", "tail_pos": [102, 113]}], "task": "RE"}
+{"text": "Finally , we show how the estimation of the tensors from point correspondences is achieved through factorization , and discuss the estimation from line correspondences .", "relation": [{"head": "point correspondences", "head_type": "NA", "head_pos": [66, 87], "relation": "used for", "tail": "estimation of the tensors", "tail_type": "NA", "tail_pos": [29, 54]}, {"head": "factorization", "head_type": "NA", "head_pos": [108, 121], "relation": "used for", "tail": "tensors", "tail_type": "NA", "tail_pos": [47, 54]}, {"head": "line correspondences", "head_type": "NA", "head_pos": [156, 176], "relation": "used for", "tail": "estimation", "tail_type": "NA", "tail_pos": [26, 36]}], "task": "RE"}
+{"text": "We propose a corpus-based method -LRB- Biber ,1993 ; Nagao ,1993 ; Smadja ,1993 -RRB- which generates Noun Classifier Associations -LRB- NCA -RRB- to overcome the problems in classifier assignment and semantic construction of noun phrase .", "relation": [{"head": "corpus-based method", "head_type": "NA", "head_pos": [16, 35], "relation": "used for", "tail": "Noun Classifier Associations -LRB- NCA -RRB-", "tail_type": "NA", "tail_pos": [111, 155]}, {"head": "corpus-based method", "head_type": "NA", "head_pos": [16, 35], "relation": "used for", "tail": "classifier assignment", "tail_type": "NA", "tail_pos": [184, 205]}, {"head": "corpus-based method", "head_type": "NA", "head_pos": [16, 35], "relation": "used for", "tail": "semantic construction of noun phrase", "tail_type": "NA", "tail_pos": [210, 246]}, {"head": "Noun Classifier Associations -LRB- NCA -RRB-", "head_type": "NA", "head_pos": [105, 149], "relation": "used for", "tail": "classifier assignment", "tail_type": "NA", "tail_pos": [184, 205]}, {"head": "Noun Classifier Associations -LRB- NCA -RRB-", "head_type": "NA", "head_pos": [105, 149], "relation": "used for", "tail": "semantic construction of noun phrase", "tail_type": "NA", "tail_pos": [210, 246]}, {"head": "classifier assignment", "head_type": "NA", "head_pos": [178, 199], "relation": "conjunction", "tail": "semantic construction of noun phrase", "tail_type": "NA", "tail_pos": [210, 246]}], "task": "RE"}
+{"text": "The NCA is created statistically from a large corpus and recomposed under concept hierarchy constraints and frequency of occurrences .", "relation": [{"head": "concept hierarchy constraints", "head_type": "NA", "head_pos": [83, 112], "relation": "used for", "tail": "NCA", "tail_type": "NA", "tail_pos": [7, 10]}, {"head": "frequency of occurrences", "head_type": "NA", "head_pos": [117, 141], "relation": "used for", "tail": "NCA", "tail_type": "NA", "tail_pos": [7, 10]}], "task": "RE"}
+{"text": "The perception of transparent objects from images is known to be a very hard problem in vision .", "relation": [{"head": "images", "head_type": "NA", "head_pos": [52, 58], "relation": "used for", "tail": "perception of transparent objects", "tail_type": "NA", "tail_pos": [7, 40]}], "task": "RE"}
+{"text": "We show how features that are imaged through a transparent object behave differently from those that are rigidly attached to the scene .", "relation": [{"head": "those", "head_type": "NA", "head_pos": [99, 104], "relation": "compare", "tail": "features", "tail_type": "NA", "tail_pos": [15, 23]}], "task": "RE"}
+{"text": "We present a novel model-based approach to recover the shapes and the poses of transparent objects from known motion .", "relation": [{"head": "model-based approach", "head_type": "NA", "head_pos": [22, 42], "relation": "used for", "tail": "shapes and the poses of transparent objects", "tail_type": "NA", "tail_pos": [64, 107]}, {"head": "known motion", "head_type": "NA", "head_pos": [113, 125], "relation": "used for", "tail": "shapes and the poses of transparent objects", "tail_type": "NA", "tail_pos": [58, 101]}], "task": "RE"}
+{"text": "The objects can be complex in that they may be composed of multiple layers with different refractive indices .", "relation": [{"head": "multiple layers", "head_type": "NA", "head_pos": [68, 83], "relation": "part of", "tail": "they", "tail_type": "NA", "tail_pos": [38, 42]}, {"head": "refractive indices", "head_type": "NA", "head_pos": [99, 117], "relation": "feature of", "tail": "multiple layers", "tail_type": "NA", "tail_pos": [62, 77]}], "task": "RE"}
+{"text": "We have applied it to real scenes that include transparent objects and recovered the shapes of the objects with high accuracy .", "relation": [{"head": "it", "head_type": "NA", "head_pos": [19, 21], "relation": "used for", "tail": "real scenes", "tail_type": "NA", "tail_pos": [31, 42]}, {"head": "it", "head_type": "NA", "head_pos": [19, 21], "relation": "used for", "tail": "shapes of the objects", "tail_type": "NA", "tail_pos": [94, 115]}, {"head": "transparent objects", "head_type": "NA", "head_pos": [56, 75], "relation": "part of", "tail": "real scenes", "tail_type": "NA", "tail_pos": [25, 36]}, {"head": "accuracy", "head_type": "NA", "head_pos": [126, 134], "relation": "evaluate for", "tail": "shapes of the objects", "tail_type": "NA", "tail_pos": [88, 109]}], "task": "RE"}
+{"text": "We propose a novel probabilistic framework for learning visual models of 3D object categories by combining appearance information and geometric constraints .", "relation": [{"head": "probabilistic framework", "head_type": "NA", "head_pos": [22, 45], "relation": "used for", "tail": "visual models of 3D object categories", "tail_type": "NA", "tail_pos": [65, 102]}, {"head": "appearance information", "head_type": "NA", "head_pos": [116, 138], "relation": "used for", "tail": "probabilistic framework", "tail_type": "NA", "tail_pos": [22, 45]}, {"head": "appearance information", "head_type": "NA", "head_pos": [110, 132], "relation": "conjunction", "tail": "geometric constraints", "tail_type": "NA", "tail_pos": [143, 164]}, {"head": "geometric constraints", "head_type": "NA", "head_pos": [143, 164], "relation": "used for", "tail": "probabilistic framework", "tail_type": "NA", "tail_pos": [22, 45]}], "task": "RE"}
+{"text": "A generative framework is used for learning a model that captures the relative position of parts within each of the discretized viewpoints .", "relation": [{"head": "generative framework", "head_type": "NA", "head_pos": [5, 25], "relation": "used for", "tail": "model", "tail_type": "NA", "tail_pos": [55, 60]}], "task": "RE"}
+{"text": "Contrary to most of the existing mixture of viewpoints models , our model establishes explicit correspondences of parts across different viewpoints of the object class .", "relation": [{"head": "model", "head_type": "NA", "head_pos": [58, 63], "relation": "compare", "tail": "mixture of viewpoints models", "tail_type": "NA", "tail_pos": [36, 64]}], "task": "RE"}
+{"text": "Given a new image , detection and classification are achieved by determining the position and viewpoint of the model that maximize recognition scores of the candidate objects .", "relation": [{"head": "image", "head_type": "NA", "head_pos": [15, 20], "relation": "used for", "tail": "detection", "tail_type": "NA", "tail_pos": [29, 38]}, {"head": "image", "head_type": "NA", "head_pos": [15, 20], "relation": "used for", "tail": "classification", "tail_type": "NA", "tail_pos": [43, 57]}, {"head": "detection", "head_type": "NA", "head_pos": [23, 32], "relation": "conjunction", "tail": "classification", "tail_type": "NA", "tail_pos": [43, 57]}, {"head": "position", "head_type": "NA", "head_pos": [90, 98], "relation": "used for", "tail": "detection", "tail_type": "NA", "tail_pos": [23, 32]}, {"head": "position", "head_type": "NA", "head_pos": [90, 98], "relation": "used for", "tail": "classification", "tail_type": "NA", "tail_pos": [37, 51]}, {"head": "position", "head_type": "NA", "head_pos": [84, 92], "relation": "conjunction", "tail": "viewpoint", "tail_type": "NA", "tail_pos": [103, 112]}, {"head": "viewpoint", "head_type": "NA", "head_pos": [103, 112], "relation": "used for", "tail": "detection", "tail_type": "NA", "tail_pos": [23, 32]}, {"head": "viewpoint", "head_type": "NA", "head_pos": [103, 112], "relation": "used for", "tail": "classification", "tail_type": "NA", "tail_pos": [37, 51]}], "task": "RE"}
+{"text": "Our approach is among the first to propose a generative proba-bilistic framework for 3D object categorization .", "relation": [{"head": "generative proba-bilistic framework", "head_type": "NA", "head_pos": [48, 83], "relation": "used for", "tail": "3D object categorization", "tail_type": "NA", "tail_pos": [94, 118]}], "task": "RE"}
+{"text": "We test our algorithm on the detection task and the viewpoint classification task by using '' car '' category from both the Savarese et al. 2007 and PASCAL VOC 2006 datasets .", "relation": [{"head": "algorithm", "head_type": "NA", "head_pos": [15, 24], "relation": "used for", "tail": "detection task", "tail_type": "NA", "tail_pos": [38, 52]}, {"head": "algorithm", "head_type": "NA", "head_pos": [15, 24], "relation": "used for", "tail": "viewpoint classification task", "tail_type": "NA", "tail_pos": [61, 90]}, {"head": "detection task", "head_type": "NA", "head_pos": [32, 46], "relation": "conjunction", "tail": "viewpoint classification task", "tail_type": "NA", "tail_pos": [61, 90]}, {"head": "PASCAL VOC 2006 datasets", "head_type": "NA", "head_pos": [158, 182], "relation": "evaluate for", "tail": "algorithm", "tail_type": "NA", "tail_pos": [15, 24]}], "task": "RE"}
+{"text": "We show promising results in both the detection and viewpoint classification tasks on these two challenging datasets .", "relation": [{"head": "datasets", "head_type": "NA", "head_pos": [117, 125], "relation": "evaluate for", "tail": "detection and viewpoint classification tasks", "tail_type": "NA", "tail_pos": [41, 85]}], "task": "RE"}
+{"text": "We present an application of ambiguity packing and stochastic disambiguation techniques for Lexical-Functional Grammars -LRB- LFG -RRB- to the domain of sentence condensation .", "relation": [{"head": "ambiguity packing and stochastic disambiguation techniques", "head_type": "NA", "head_pos": [32, 90], "relation": "used for", "tail": "Lexical-Functional Grammars -LRB- LFG -RRB-", "tail_type": "NA", "tail_pos": [101, 144]}, {"head": "ambiguity packing and stochastic disambiguation techniques", "head_type": "NA", "head_pos": [32, 90], "relation": "used for", "tail": "sentence condensation", "tail_type": "NA", "tail_pos": [162, 183]}], "task": "RE"}
+{"text": "Our system incorporates a linguistic parser/generator for LFG , a transfer component for parse reduction operating on packed parse forests , and a maximum-entropy model for stochastic output selection .", "relation": [{"head": "linguistic parser/generator", "head_type": "NA", "head_pos": [35, 62], "relation": "part of", "tail": "system", "tail_type": "NA", "tail_pos": [7, 13]}, {"head": "linguistic parser/generator", "head_type": "NA", "head_pos": [29, 56], "relation": "used for", "tail": "LFG", "tail_type": "NA", "tail_pos": [67, 70]}, {"head": "linguistic parser/generator", "head_type": "NA", "head_pos": [29, 56], "relation": "conjunction", "tail": "transfer component", "tail_type": "NA", "tail_pos": [75, 93]}, {"head": "transfer component", "head_type": "NA", "head_pos": [75, 93], "relation": "part of", "tail": "system", "tail_type": "NA", "tail_pos": [7, 13]}, {"head": "transfer component", "head_type": "NA", "head_pos": [69, 87], "relation": "used for", "tail": "parse reduction", "tail_type": "NA", "tail_pos": [98, 113]}, {"head": "transfer component", "head_type": "NA", "head_pos": [69, 87], "relation": "conjunction", "tail": "maximum-entropy model", "tail_type": "NA", "tail_pos": [156, 177]}, {"head": "packed parse forests", "head_type": "NA", "head_pos": [127, 147], "relation": "used for", "tail": "parse reduction", "tail_type": "NA", "tail_pos": [92, 107]}, {"head": "maximum-entropy model", "head_type": "NA", "head_pos": [156, 177], "relation": "part of", "tail": "system", "tail_type": "NA", "tail_pos": [7, 13]}, {"head": "maximum-entropy model", "head_type": "NA", "head_pos": [150, 171], "relation": "used for", "tail": "stochastic output selection", "tail_type": "NA", "tail_pos": [182, 209]}], "task": "RE"}
+{"text": "Furthermore , we propose the use of standard parser evaluation methods for automatically evaluating the summarization quality of sentence condensation systems .", "relation": [{"head": "parser evaluation methods", "head_type": "NA", "head_pos": [48, 73], "relation": "evaluate for", "tail": "summarization quality", "tail_type": "NA", "tail_pos": [113, 134]}, {"head": "summarization quality", "head_type": "NA", "head_pos": [107, 128], "relation": "evaluate for", "tail": "sentence condensation systems", "tail_type": "NA", "tail_pos": [138, 167]}], "task": "RE"}
+{"text": "An experimental evaluation of summarization quality shows a close correlation between the automatic parse-based evaluation and a manual evaluation of generated strings .", "relation": [{"head": "summarization quality", "head_type": "NA", "head_pos": [33, 54], "relation": "evaluate for", "tail": "automatic parse-based evaluation", "tail_type": "NA", "tail_pos": [99, 131]}, {"head": "automatic parse-based evaluation", "head_type": "NA", "head_pos": [93, 125], "relation": "compare", "tail": "manual evaluation", "tail_type": "NA", "tail_pos": [138, 155]}], "task": "RE"}
+{"text": "Overall summarization quality of the proposed system is state-of-the-art , with guaranteed grammaticality of the system output due to the use of a constraint-based parser/generator .", "relation": [{"head": "summarization quality", "head_type": "NA", "head_pos": [11, 32], "relation": "evaluate for", "tail": "system", "tail_type": "NA", "tail_pos": [55, 61]}, {"head": "grammaticality", "head_type": "NA", "head_pos": [100, 114], "relation": "evaluate for", "tail": "system", "tail_type": "NA", "tail_pos": [49, 55]}, {"head": "constraint-based parser/generator", "head_type": "NA", "head_pos": [156, 189], "relation": "used for", "tail": "system", "tail_type": "NA", "tail_pos": [49, 55]}], "task": "RE"}
+{"text": "The robust principal component analysis -LRB- robust PCA -RRB- problem has been considered in many machine learning applications , where the goal is to decompose the data matrix to a low rank part plus a sparse residual .", "relation": [{"head": "robust principal component analysis -LRB- robust PCA -RRB- problem", "head_type": "NA", "head_pos": [7, 73], "relation": "used for", "tail": "machine learning applications", "tail_type": "NA", "tail_pos": [108, 137]}, {"head": "low rank part", "head_type": "NA", "head_pos": [192, 205], "relation": "part of", "tail": "data matrix", "tail_type": "NA", "tail_pos": [169, 180]}, {"head": "low rank part", "head_type": "NA", "head_pos": [186, 199], "relation": "conjunction", "tail": "sparse residual", "tail_type": "NA", "tail_pos": [213, 228]}, {"head": "sparse residual", "head_type": "NA", "head_pos": [213, 228], "relation": "part of", "tail": "data matrix", "tail_type": "NA", "tail_pos": [169, 180]}], "task": "RE"}
+{"text": "While current approaches are developed by only considering the low rank plus sparse structure , in many applications , side information of row and/or column entities may also be given , and it is still unclear to what extent could such information help robust PCA .", "relation": [{"head": "low rank plus sparse structure", "head_type": "NA", "head_pos": [72, 102], "relation": "used for", "tail": "approaches", "tail_type": "NA", "tail_pos": [17, 27]}, {"head": "information", "head_type": "NA", "head_pos": [124, 135], "relation": "used for", "tail": "robust PCA", "tail_type": "NA", "tail_pos": [262, 272]}], "task": "RE"}
+{"text": "Thus , in this paper , we study the problem of robust PCA with side information , where both prior structure and features of entities are exploited for recovery .", "relation": [{"head": "side information", "head_type": "NA", "head_pos": [72, 88], "relation": "used for", "tail": "robust PCA", "tail_type": "NA", "tail_pos": [50, 60]}, {"head": "prior structure", "head_type": "NA", "head_pos": [96, 111], "relation": "conjunction", "tail": "features of entities", "tail_type": "NA", "tail_pos": [122, 142]}, {"head": "prior structure", "head_type": "NA", "head_pos": [96, 111], "relation": "used for", "tail": "recovery", "tail_type": "NA", "tail_pos": [161, 169]}, {"head": "features of entities", "head_type": "NA", "head_pos": [116, 136], "relation": "used for", "tail": "recovery", "tail_type": "NA", "tail_pos": [161, 169]}], "task": "RE"}
+{"text": "We propose a convex problem to incorporate side information in robust PCA and show that the low rank matrix can be exactly recovered via the proposed method under certain conditions .", "relation": [{"head": "convex problem", "head_type": "NA", "head_pos": [16, 30], "relation": "used for", "tail": "side information", "tail_type": "NA", "tail_pos": [52, 68]}, {"head": "side information", "head_type": "NA", "head_pos": [46, 62], "relation": "part of", "tail": "robust PCA", "tail_type": "NA", "tail_pos": [72, 82]}, {"head": "method", "head_type": "NA", "head_pos": [159, 165], "relation": "used for", "tail": "low rank matrix", "tail_type": "NA", "tail_pos": [95, 110]}], "task": "RE"}
+{"text": "In particular , our guarantee suggests that a substantial amount of low rank matrices , which can not be recovered by standard robust PCA , become re-coverable by our proposed method .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [185, 191], "relation": "used for", "tail": "low rank matrices", "tail_type": "NA", "tail_pos": [71, 88]}], "task": "RE"}
+{"text": "The result theoretically justifies the effectiveness of features in robust PCA .", "relation": [{"head": "features", "head_type": "NA", "head_pos": [59, 67], "relation": "feature of", "tail": "robust PCA", "tail_type": "NA", "tail_pos": [77, 87]}], "task": "RE"}
+{"text": "In addition , we conduct synthetic experiments as well as a real application on noisy image classification to show that our method also improves the performance in practice by exploiting side information .", "relation": [{"head": "noisy image classification", "head_type": "NA", "head_pos": [83, 109], "relation": "evaluate for", "tail": "method", "tail_type": "NA", "tail_pos": [133, 139]}, {"head": "side information", "head_type": "NA", "head_pos": [196, 212], "relation": "used for", "tail": "method", "tail_type": "NA", "tail_pos": [127, 133]}], "task": "RE"}
+{"text": "This paper presents necessary and sufficient conditions for the use of demonstrative expressions in English and discusses implications for current discourse processing algorithms .", "relation": [{"head": "demonstrative expressions", "head_type": "NA", "head_pos": [74, 99], "relation": "feature of", "tail": "English", "tail_type": "NA", "tail_pos": [109, 116]}, {"head": "implications", "head_type": "NA", "head_pos": [125, 137], "relation": "used for", "tail": "discourse processing algorithms", "tail_type": "NA", "tail_pos": [156, 187]}], "task": "RE"}
+{"text": "This research is part of a larger study of anaphoric expressions , the results of which will be incorporated into a natural language generation system .", "relation": [{"head": "anaphoric expressions", "head_type": "NA", "head_pos": [46, 67], "relation": "used for", "tail": "natural language generation system", "tail_type": "NA", "tail_pos": [125, 159]}], "task": "RE"}
+{"text": "Using the IEMOCAP database , discrete -LRB- categorical -RRB- and continuous -LRB- attribute -RRB- emotional assessments evaluated by the actors and na ¨ ıve listeners are compared .", "relation": [{"head": "IEMOCAP database", "head_type": "NA", "head_pos": [13, 29], "relation": "used for", "tail": "discrete -LRB- categorical -RRB- and continuous -LRB- attribute -RRB- emotional assessments", "tail_type": "NA", "tail_pos": [38, 129]}], "task": "RE"}
+{"text": "The problem of blind separation of underdetermined instantaneous mixtures of independent signals is addressed through a method relying on nonstationarity of the original signals .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [129, 135], "relation": "used for", "tail": "blind separation of underdetermined instantaneous mixtures of independent signals", "tail_type": "NA", "tail_pos": [18, 99]}, {"head": "nonstationarity", "head_type": "NA", "head_pos": [147, 162], "relation": "used for", "tail": "method", "tail_type": "NA", "tail_pos": [123, 129]}], "task": "RE"}
+{"text": "In comparison with previous works , in this paper it is assumed that the signals are not i.i.d. in each epoch , but obey a first-order autoregressive model .", "relation": [{"head": "first-order autoregressive model", "head_type": "NA", "head_pos": [132, 164], "relation": "used for", "tail": "signals", "tail_type": "NA", "tail_pos": [76, 83]}], "task": "RE"}
+{"text": "This model was shown to be more appropriate for blind separation of natural speech signals .", "relation": [{"head": "model", "head_type": "NA", "head_pos": [8, 13], "relation": "used for", "tail": "blind separation of natural speech signals .", "tail_type": "NA", "tail_pos": [57, 101]}], "task": "RE"}
+{"text": "A separation method is proposed that is nearly statistically efficient -LRB- approaching the corresponding Cramér-Rao lower bound -RRB- , if the separated signals obey the assumed model .", "relation": [{"head": "Cramér-Rao lower bound -RRB-", "head_type": "NA", "head_pos": [116, 144], "relation": "feature of", "tail": "separation method", "tail_type": "NA", "tail_pos": [5, 22]}], "task": "RE"}
+{"text": "In the case of natural speech signals , the method is shown to have separation accuracy better than the state-of-the-art methods .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [53, 59], "relation": "used for", "tail": "natural speech signals", "tail_type": "NA", "tail_pos": [18, 40]}, {"head": "method", "head_type": "NA", "head_pos": [47, 53], "relation": "compare", "tail": "methods", "tail_type": "NA", "tail_pos": [130, 137]}, {"head": "separation accuracy", "head_type": "NA", "head_pos": [77, 96], "relation": "evaluate for", "tail": "method", "tail_type": "NA", "tail_pos": [47, 53]}, {"head": "separation accuracy", "head_type": "NA", "head_pos": [71, 90], "relation": "evaluate for", "tail": "methods", "tail_type": "NA", "tail_pos": [130, 137]}, {"head": "methods", "head_type": "NA", "head_pos": [130, 137], "relation": "used for", "tail": "natural speech signals", "tail_type": "NA", "tail_pos": [18, 40]}], "task": "RE"}
+{"text": "This paper proposes to use a convolution kernel over parse trees to model syntactic structure information for relation extraction .", "relation": [{"head": "convolution kernel over parse trees", "head_type": "NA", "head_pos": [32, 67], "relation": "used for", "tail": "syntactic structure information", "tail_type": "NA", "tail_pos": [83, 114]}, {"head": "syntactic structure information", "head_type": "NA", "head_pos": [77, 108], "relation": "used for", "tail": "relation extraction", "tail_type": "NA", "tail_pos": [119, 138]}], "task": "RE"}
+{"text": "Our study reveals that the syntactic structure features embedded in a parse tree are very effective for relation extraction and these features can be well captured by the convolution tree kernel .", "relation": [{"head": "syntactic structure features", "head_type": "NA", "head_pos": [30, 58], "relation": "feature of", "tail": "parse tree", "tail_type": "NA", "tail_pos": [79, 89]}, {"head": "syntactic structure features", "head_type": "NA", "head_pos": [30, 58], "relation": "used for", "tail": "relation extraction", "tail_type": "NA", "tail_pos": [113, 132]}, {"head": "convolution tree kernel", "head_type": "NA", "head_pos": [180, 203], "relation": "used for", "tail": "features", "tail_type": "NA", "tail_pos": [47, 55]}], "task": "RE"}
+{"text": "Evaluation on the ACE 2003 corpus shows that the convolution kernel over parse trees can achieve comparable performance with the previous best-reported feature-based methods on the 24 ACE relation subtypes .", "relation": [{"head": "ACE 2003 corpus", "head_type": "NA", "head_pos": [21, 36], "relation": "evaluate for", "tail": "convolution kernel over parse trees", "tail_type": "NA", "tail_pos": [58, 93]}, {"head": "feature-based methods", "head_type": "NA", "head_pos": [161, 182], "relation": "compare", "tail": "convolution kernel over parse trees", "tail_type": "NA", "tail_pos": [52, 87]}], "task": "RE"}
+{"text": "It also shows that our method significantly outperforms the previous two dependency tree kernels on the 5 ACE relation major types .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [26, 32], "relation": "compare", "tail": "dependency tree kernels", "tail_type": "NA", "tail_pos": [82, 105]}], "task": "RE"}
+{"text": "This paper presents the results of automatically inducing a Combinatory Categorial Grammar -LRB- CCG -RRB- lexicon from a Turkish dependency treebank .", "relation": [{"head": "Combinatory Categorial Grammar -LRB- CCG -RRB- lexicon", "head_type": "NA", "head_pos": [63, 117], "relation": "part of", "tail": "Turkish dependency treebank", "tail_type": "NA", "tail_pos": [131, 158]}], "task": "RE"}
+{"text": "The fact that Turkish is an agglutinating free word order language presents a challenge for language theories .", "relation": [{"head": "Turkish", "head_type": "NA", "head_pos": [17, 24], "relation": "hyponym of", "tail": "agglutinating free word order language", "tail_type": "NA", "tail_pos": [37, 75]}], "task": "RE"}
+{"text": "We explored possible ways to obtain a compact lexicon , consistent with CCG principles , from a treebank which is an order of magnitude smaller than Penn WSJ .", "relation": [{"head": "compact lexicon", "head_type": "NA", "head_pos": [41, 56], "relation": "part of", "tail": "treebank", "tail_type": "NA", "tail_pos": [105, 113]}, {"head": "treebank", "head_type": "NA", "head_pos": [99, 107], "relation": "compare", "tail": "Penn WSJ", "tail_type": "NA", "tail_pos": [158, 166]}], "task": "RE"}
+{"text": "While sentence extraction as an approach to summarization has been shown to work in documents of certain genres , because of the conversational nature of email communication where utterances are made in relation to one made previously , sentence extraction may not capture the necessary segments of dialogue that would make a summary coherent .", "relation": [{"head": "sentence extraction", "head_type": "NA", "head_pos": [9, 28], "relation": "used for", "tail": "summarization", "tail_type": "NA", "tail_pos": [53, 66]}], "task": "RE"}
+{"text": "In this paper , we present our work on the detection of question-answer pairs in an email conversation for the task of email summarization .", "relation": [{"head": "detection of question-answer pairs", "head_type": "NA", "head_pos": [46, 80], "relation": "used for", "tail": "email summarization", "tail_type": "NA", "tail_pos": [128, 147]}, {"head": "email conversation", "head_type": "NA", "head_pos": [93, 111], "relation": "used for", "tail": "detection of question-answer pairs", "tail_type": "NA", "tail_pos": [46, 80]}], "task": "RE"}
+{"text": "We show that various features based on the structure of email-threads can be used to improve upon lexical similarity of discourse segments for question-answer pairing .", "relation": [{"head": "features", "head_type": "NA", "head_pos": [24, 32], "relation": "used for", "tail": "lexical similarity", "tail_type": "NA", "tail_pos": [107, 125]}, {"head": "features", "head_type": "NA", "head_pos": [24, 32], "relation": "used for", "tail": "question-answer pairing", "tail_type": "NA", "tail_pos": [152, 175]}, {"head": "structure of email-threads", "head_type": "NA", "head_pos": [52, 78], "relation": "used for", "tail": "features", "tail_type": "NA", "tail_pos": [24, 32]}, {"head": "lexical similarity", "head_type": "NA", "head_pos": [101, 119], "relation": "feature of", "tail": "discourse segments", "tail_type": "NA", "tail_pos": [129, 147]}], "task": "RE"}
+{"text": "Specifically , we show how to incorporate a simple prior on the distribution of natural images into support vector machines .", "relation": [{"head": "prior on the distribution of natural images", "head_type": "NA", "head_pos": [54, 97], "relation": "used for", "tail": "support vector machines", "tail_type": "NA", "tail_pos": [109, 132]}], "task": "RE"}
+{"text": "SVMs are known to be robust to overfitting ; however , a few training examples usually do not represent well the structure of the class .", "relation": [{"head": "SVMs", "head_type": "NA", "head_pos": [3, 7], "relation": "used for", "tail": "overfitting", "tail_type": "NA", "tail_pos": [40, 51]}], "task": "RE"}
+{"text": "Our experiments on real data sets show that the resulting detector is more robust to the choice of training examples , and substantially improves both linear and kernel SVM when trained on 10 positive and 10 negative examples .", "relation": [{"head": "real data sets", "head_type": "NA", "head_pos": [22, 36], "relation": "evaluate for", "tail": "detector", "tail_type": "NA", "tail_pos": [67, 75]}, {"head": "detector", "head_type": "NA", "head_pos": [61, 69], "relation": "compare", "tail": "linear and kernel SVM", "tail_type": "NA", "tail_pos": [160, 181]}], "task": "RE"}
+{"text": "Although the study of clustering is centered around an intuitively compelling goal , it has been very difficult to develop a unified framework for reasoning about it at a technical level , and profoundly diverse approaches to clustering abound in the research community .", "relation": [{"head": "unified framework", "head_type": "NA", "head_pos": [128, 145], "relation": "used for", "tail": "reasoning", "tail_type": "NA", "tail_pos": [156, 165]}], "task": "RE"}
+{"text": "Relaxations of these properties expose some of the interesting -LRB- and unavoidable -RRB- trade-offs at work in well-studied clustering techniques such as single-linkage , sum-of-pairs , k-means , and k-median .", "relation": [{"head": "single-linkage", "head_type": "NA", "head_pos": [165, 179], "relation": "hyponym of", "tail": "well-studied clustering techniques", "tail_type": "NA", "tail_pos": [116, 150]}, {"head": "single-linkage", "head_type": "NA", "head_pos": [159, 173], "relation": "conjunction", "tail": "sum-of-pairs", "tail_type": "NA", "tail_pos": [182, 194]}, {"head": "sum-of-pairs", "head_type": "NA", "head_pos": [182, 194], "relation": "hyponym of", "tail": "well-studied clustering techniques", "tail_type": "NA", "tail_pos": [116, 150]}, {"head": "sum-of-pairs", "head_type": "NA", "head_pos": [176, 188], "relation": "conjunction", "tail": "k-means", "tail_type": "NA", "tail_pos": [197, 204]}, {"head": "k-means", "head_type": "NA", "head_pos": [197, 204], "relation": "hyponym of", "tail": "well-studied clustering techniques", "tail_type": "NA", "tail_pos": [116, 150]}, {"head": "k-means", "head_type": "NA", "head_pos": [191, 198], "relation": "conjunction", "tail": "k-median", "tail_type": "NA", "tail_pos": [211, 219]}, {"head": "k-median", "head_type": "NA", "head_pos": [211, 219], "relation": "hyponym of", "tail": "well-studied clustering techniques", "tail_type": "NA", "tail_pos": [116, 150]}], "task": "RE"}
+{"text": "With relevant approach , we identify important contents by PageRank algorithm on the event map constructed from documents .", "relation": [{"head": "PageRank algorithm", "head_type": "NA", "head_pos": [68, 86], "relation": "used for", "tail": "relevant approach", "tail_type": "NA", "tail_pos": [8, 25]}, {"head": "event map", "head_type": "NA", "head_pos": [94, 103], "relation": "used for", "tail": "PageRank algorithm", "tail_type": "NA", "tail_pos": [62, 80]}, {"head": "documents", "head_type": "NA", "head_pos": [121, 130], "relation": "used for", "tail": "event map", "tail_type": "NA", "tail_pos": [88, 97]}], "task": "RE"}
+{"text": "We present a scanning method that recovers dense sub-pixel camera-projector correspondence without requiring any photometric calibration nor preliminary knowledge of their relative geometry .", "relation": [{"head": "scanning method", "head_type": "NA", "head_pos": [16, 31], "relation": "used for", "tail": "dense sub-pixel camera-projector correspondence", "tail_type": "NA", "tail_pos": [52, 99]}], "task": "RE"}
+{"text": "Subpixel accuracy is achieved by considering several zero-crossings defined by the difference between pairs of unstructured patterns .", "relation": [{"head": "zero-crossings", "head_type": "NA", "head_pos": [62, 76], "relation": "used for", "tail": "Subpixel accuracy", "tail_type": "NA", "tail_pos": [3, 20]}], "task": "RE"}
+{"text": "We use gray-level band-pass white noise patterns that increase robustness to indirect lighting and scene discontinuities .", "relation": [{"head": "robustness", "head_type": "NA", "head_pos": [72, 82], "relation": "evaluate for", "tail": "gray-level band-pass white noise patterns", "tail_type": "NA", "tail_pos": [10, 51]}, {"head": "indirect lighting", "head_type": "NA", "head_pos": [86, 103], "relation": "feature of", "tail": "robustness", "tail_type": "NA", "tail_pos": [66, 76]}, {"head": "indirect lighting", "head_type": "NA", "head_pos": [80, 97], "relation": "conjunction", "tail": "scene discontinuities", "tail_type": "NA", "tail_pos": [108, 129]}, {"head": "scene discontinuities", "head_type": "NA", "head_pos": [108, 129], "relation": "feature of", "tail": "robustness", "tail_type": "NA", "tail_pos": [66, 76]}], "task": "RE"}
+{"text": "Simulated and experimental results show that our method recovers scene geometry with high subpixel precision , and that it can handle many challenges of active reconstruction systems .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [52, 58], "relation": "used for", "tail": "scene geometry", "tail_type": "NA", "tail_pos": [74, 88]}, {"head": "subpixel precision", "head_type": "NA", "head_pos": [99, 117], "relation": "feature of", "tail": "scene geometry", "tail_type": "NA", "tail_pos": [68, 82]}, {"head": "it", "head_type": "NA", "head_pos": [81, 83], "relation": "used for", "tail": "active reconstruction systems", "tail_type": "NA", "tail_pos": [162, 191]}], "task": "RE"}
+{"text": "We compare our results to state of the art methods such as mi-cro phase shifting and modulated phase shifting .", "relation": [{"head": "mi-cro phase shifting", "head_type": "NA", "head_pos": [68, 89], "relation": "hyponym of", "tail": "state of the art methods", "tail_type": "NA", "tail_pos": [29, 53]}, {"head": "mi-cro phase shifting", "head_type": "NA", "head_pos": [62, 83], "relation": "conjunction", "tail": "modulated phase shifting", "tail_type": "NA", "tail_pos": [94, 118]}, {"head": "modulated phase shifting", "head_type": "NA", "head_pos": [94, 118], "relation": "hyponym of", "tail": "state of the art methods", "tail_type": "NA", "tail_pos": [29, 53]}], "task": "RE"}
+{"text": "This paper describes a novel system for acquiring adjectival subcategorization frames -LRB- scfs -RRB- and associated frequency information from English corpus data .", "relation": [{"head": "system", "head_type": "NA", "head_pos": [32, 38], "relation": "used for", "tail": "acquiring adjectival subcategorization frames -LRB- scfs -RRB-", "tail_type": "NA", "tail_pos": [49, 111]}], "task": "RE"}
+{"text": "The system incorporates a decision-tree classifier for 30 scf types which tests for the presence of grammatical relations -LRB- grs -RRB- in the output of a robust statistical parser .", "relation": [{"head": "decision-tree classifier", "head_type": "NA", "head_pos": [35, 59], "relation": "part of", "tail": "system", "tail_type": "NA", "tail_pos": [7, 13]}, {"head": "decision-tree classifier", "head_type": "NA", "head_pos": [29, 53], "relation": "used for", "tail": "grammatical relations -LRB- grs -RRB-", "tail_type": "NA", "tail_pos": [109, 146]}], "task": "RE"}
+{"text": "It uses a powerful pattern-matching language to classify grs into frames hierarchically in a way that mirrors inheritance-based lexica .", "relation": [{"head": "pattern-matching language", "head_type": "NA", "head_pos": [28, 53], "relation": "used for", "tail": "It", "tail_type": "NA", "tail_pos": [3, 5]}, {"head": "pattern-matching language", "head_type": "NA", "head_pos": [22, 47], "relation": "used for", "tail": "grs", "tail_type": "NA", "tail_pos": [66, 69]}], "task": "RE"}
+{"text": "The experiments show that the system is able to detect scf types with 70 % precision and 66 % recall rate .", "relation": [{"head": "precision", "head_type": "NA", "head_pos": [84, 93], "relation": "evaluate for", "tail": "system", "tail_type": "NA", "tail_pos": [33, 39]}, {"head": "precision", "head_type": "NA", "head_pos": [78, 87], "relation": "conjunction", "tail": "recall", "tail_type": "NA", "tail_pos": [103, 109]}, {"head": "recall", "head_type": "NA", "head_pos": [103, 109], "relation": "evaluate for", "tail": "system", "tail_type": "NA", "tail_pos": [33, 39]}], "task": "RE"}
+{"text": "A new tool for linguistic annotation of scfs in corpus data is also introduced which can considerably alleviate the process of obtaining training and test data for subcategorization acquisition .", "relation": [{"head": "tool", "head_type": "NA", "head_pos": [9, 13], "relation": "used for", "tail": "linguistic annotation of scfs", "tail_type": "NA", "tail_pos": [24, 53]}, {"head": "training and test data", "head_type": "NA", "head_pos": [140, 162], "relation": "used for", "tail": "subcategorization acquisition", "tail_type": "NA", "tail_pos": [173, 202]}], "task": "RE"}
+{"text": "Machine transliteration/back-transliteration plays an important role in many multilingual speech and language applications .", "relation": [{"head": "Machine transliteration/back-transliteration", "head_type": "NA", "head_pos": [3, 47], "relation": "used for", "tail": "multilingual speech and language applications", "tail_type": "NA", "tail_pos": [86, 131]}], "task": "RE"}
+{"text": "In this paper , a novel framework for machine transliteration/backtransliteration that allows us to carry out direct orthographical mapping -LRB- DOM -RRB- between two different languages is presented .", "relation": [{"head": "framework", "head_type": "NA", "head_pos": [27, 36], "relation": "used for", "tail": "machine transliteration/backtransliteration", "tail_type": "NA", "tail_pos": [47, 90]}, {"head": "machine transliteration/backtransliteration", "head_type": "NA", "head_pos": [41, 84], "relation": "used for", "tail": "direct orthographical mapping -LRB- DOM -RRB-", "tail_type": "NA", "tail_pos": [119, 164]}], "task": "RE"}
+{"text": "Under this framework , a joint source-channel transliteration model , also called n-gram transliteration model -LRB- n-gram TM -RRB- , is further proposed to model the transliteration process .", "relation": [{"head": "framework", "head_type": "NA", "head_pos": [14, 23], "relation": "used for", "tail": "joint source-channel transliteration model", "tail_type": "NA", "tail_pos": [34, 76]}, {"head": "n-gram transliteration model -LRB- n-gram TM -RRB-", "head_type": "NA", "head_pos": [85, 135], "relation": "used for", "tail": "transliteration process", "tail_type": "NA", "tail_pos": [177, 200]}], "task": "RE"}
+{"text": "We evaluate the proposed methods through several transliteration/backtransliteration experiments for English/Chinese and English/Japanese language pairs .", "relation": [{"head": "transliteration/backtransliteration", "head_type": "NA", "head_pos": [58, 93], "relation": "evaluate for", "tail": "methods", "tail_type": "NA", "tail_pos": [28, 35]}, {"head": "transliteration/backtransliteration", "head_type": "NA", "head_pos": [52, 87], "relation": "used for", "tail": "English/Chinese and English/Japanese language pairs", "tail_type": "NA", "tail_pos": [110, 161]}], "task": "RE"}
+{"text": "Our study reveals that the proposed method not only reduces an extensive system development effort but also improves the transliteration accuracy significantly .", "relation": [{"head": "transliteration accuracy", "head_type": "NA", "head_pos": [130, 154], "relation": "evaluate for", "tail": "method", "tail_type": "NA", "tail_pos": [39, 45]}], "task": "RE"}
+{"text": "A bio-inspired model for an analog programmable array processor -LRB- APAP -RRB- , based on studies on the vertebrate retina , has permitted the realization of complex programmable spatio-temporal dynamics in VLSI .", "relation": [{"head": "bio-inspired model", "head_type": "NA", "head_pos": [5, 23], "relation": "used for", "tail": "analog programmable array processor -LRB- APAP -RRB-", "tail_type": "NA", "tail_pos": [37, 89]}, {"head": "bio-inspired model", "head_type": "NA", "head_pos": [5, 23], "relation": "used for", "tail": "complex programmable spatio-temporal dynamics", "tail_type": "NA", "tail_pos": [169, 214]}, {"head": "vertebrate retina", "head_type": "NA", "head_pos": [116, 133], "relation": "used for", "tail": "bio-inspired model", "tail_type": "NA", "tail_pos": [5, 23]}, {"head": "complex programmable spatio-temporal dynamics", "head_type": "NA", "head_pos": [163, 208], "relation": "feature of", "tail": "VLSI", "tail_type": "NA", "tail_pos": [218, 222]}], "task": "RE"}
+{"text": "This model mimics the way in which images are processed in the visual pathway , rendering a feasible alternative for the implementation of early vision applications in standard technologies .", "relation": [{"head": "visual pathway", "head_type": "NA", "head_pos": [72, 86], "relation": "used for", "tail": "images", "tail_type": "NA", "tail_pos": [38, 44]}], "task": "RE"}
+{"text": "Computing power per area and power consumption is amongst the highest reported for a single chip .", "relation": [{"head": "Computing power per area", "head_type": "NA", "head_pos": [3, 27], "relation": "conjunction", "tail": "power consumption", "tail_type": "NA", "tail_pos": [38, 55]}], "task": "RE"}
+{"text": "Another problem with determiners is their inherent ambiguity .", "relation": [{"head": "ambiguity", "head_type": "NA", "head_pos": [60, 69], "relation": "feature of", "tail": "determiners", "tail_type": "NA", "tail_pos": [24, 35]}], "task": "RE"}
+{"text": "In this paper we propose a logical formalism , which , among other things , is suitable for representing determiners without forcing a particular interpretation when their meaning is still not clear .", "relation": [{"head": "logical formalism", "head_type": "NA", "head_pos": [30, 47], "relation": "used for", "tail": "determiners", "tail_type": "NA", "tail_pos": [114, 125]}], "task": "RE"}
+{"text": "We investigate the verbal and nonverbal means for grounding , and propose a design for embodied conversational agents that relies on both kinds of signals to establish common ground in human-computer interaction .", "relation": [{"head": "verbal and nonverbal means", "head_type": "NA", "head_pos": [22, 48], "relation": "used for", "tail": "grounding", "tail_type": "NA", "tail_pos": [59, 68]}, {"head": "design", "head_type": "NA", "head_pos": [79, 85], "relation": "used for", "tail": "embodied conversational agents", "tail_type": "NA", "tail_pos": [96, 126]}, {"head": "common ground", "head_type": "NA", "head_pos": [171, 184], "relation": "used for", "tail": "human-computer interaction", "tail_type": "NA", "tail_pos": [194, 220]}], "task": "RE"}
+{"text": "We analyzed eye gaze , head nods and attentional focus in the context of a direction-giving task .", "relation": [{"head": "eye gaze", "head_type": "NA", "head_pos": [15, 23], "relation": "conjunction", "tail": "head nods", "tail_type": "NA", "tail_pos": [32, 41]}, {"head": "eye gaze", "head_type": "NA", "head_pos": [15, 23], "relation": "part of", "tail": "direction-giving task", "tail_type": "NA", "tail_pos": [84, 105]}, {"head": "head nods", "head_type": "NA", "head_pos": [26, 35], "relation": "conjunction", "tail": "attentional focus", "tail_type": "NA", "tail_pos": [46, 63]}, {"head": "head nods", "head_type": "NA", "head_pos": [26, 35], "relation": "part of", "tail": "direction-giving task", "tail_type": "NA", "tail_pos": [84, 105]}, {"head": "attentional focus", "head_type": "NA", "head_pos": [40, 57], "relation": "part of", "tail": "direction-giving task", "tail_type": "NA", "tail_pos": [84, 105]}], "task": "RE"}
+{"text": "Based on these results , we present an ECA that uses verbal and nonverbal grounding acts to update dialogue state .", "relation": [{"head": "verbal and nonverbal grounding acts", "head_type": "NA", "head_pos": [62, 97], "relation": "used for", "tail": "ECA", "tail_type": "NA", "tail_pos": [42, 45]}, {"head": "verbal and nonverbal grounding acts", "head_type": "NA", "head_pos": [56, 91], "relation": "used for", "tail": "dialogue state", "tail_type": "NA", "tail_pos": [108, 122]}], "task": "RE"}
+{"text": "Sentence boundary detection in speech is important for enriching speech recognition output , making it easier for humans to read and downstream modules to process .", "relation": [{"head": "Sentence boundary detection", "head_type": "NA", "head_pos": [3, 30], "relation": "used for", "tail": "speech recognition output", "tail_type": "NA", "tail_pos": [74, 99]}, {"head": "speech", "head_type": "NA", "head_pos": [40, 46], "relation": "used for", "tail": "Sentence boundary detection", "tail_type": "NA", "tail_pos": [3, 30]}], "task": "RE"}
+{"text": "In previous work , we have developed hidden Markov model -LRB- HMM -RRB- and maximum entropy -LRB- Maxent -RRB- classifiers that integrate textual and prosodic knowledge sources for detecting sentence boundaries .", "relation": [{"head": "hidden Markov model -LRB- HMM -RRB- and maximum entropy -LRB- Maxent -RRB- classifiers", "head_type": "NA", "head_pos": [40, 126], "relation": "used for", "tail": "detecting sentence boundaries", "tail_type": "NA", "tail_pos": [191, 220]}, {"head": "textual and prosodic knowledge sources", "head_type": "NA", "head_pos": [148, 186], "relation": "used for", "tail": "hidden Markov model -LRB- HMM -RRB- and maximum entropy -LRB- Maxent -RRB- classifiers", "tail_type": "NA", "tail_pos": [40, 126]}], "task": "RE"}
+{"text": "In this paper , we evaluate the use of a conditional random field -LRB- CRF -RRB- for this task and relate results with this model to our prior work .", "relation": [{"head": "conditional random field -LRB- CRF -RRB-", "head_type": "NA", "head_pos": [44, 84], "relation": "used for", "tail": "task", "tail_type": "NA", "tail_pos": [100, 104]}], "task": "RE"}
+{"text": "We evaluate across two corpora -LRB- conversational telephone speech and broadcast news speech -RRB- on both human transcriptions and speech recognition output .", "relation": [{"head": "corpora", "head_type": "NA", "head_pos": [26, 33], "relation": "evaluate for", "tail": "human transcriptions", "tail_type": "NA", "tail_pos": [118, 138]}, {"head": "corpora", "head_type": "NA", "head_pos": [26, 33], "relation": "evaluate for", "tail": "speech recognition output", "tail_type": "NA", "tail_pos": [143, 168]}, {"head": "conversational telephone speech", "head_type": "NA", "head_pos": [46, 77], "relation": "hyponym of", "tail": "corpora", "tail_type": "NA", "tail_pos": [26, 33]}, {"head": "conversational telephone speech", "head_type": "NA", "head_pos": [40, 71], "relation": "conjunction", "tail": "broadcast news speech", "tail_type": "NA", "tail_pos": [82, 103]}, {"head": "broadcast news speech", "head_type": "NA", "head_pos": [82, 103], "relation": "hyponym of", "tail": "corpora", "tail_type": "NA", "tail_pos": [26, 33]}, {"head": "human transcriptions", "head_type": "NA", "head_pos": [112, 132], "relation": "conjunction", "tail": "speech recognition output", "tail_type": "NA", "tail_pos": [143, 168]}], "task": "RE"}
+{"text": "In general , our CRF model yields a lower error rate than the HMM and Max-ent models on the NIST sentence boundary detection task in speech , although it is interesting to note that the best results are achieved by three-way voting among the classifiers .", "relation": [{"head": "CRF model", "head_type": "NA", "head_pos": [20, 29], "relation": "compare", "tail": "HMM and Max-ent models", "tail_type": "NA", "tail_pos": [71, 93]}, {"head": "error rate", "head_type": "NA", "head_pos": [51, 61], "relation": "evaluate for", "tail": "CRF model", "tail_type": "NA", "tail_pos": [20, 29]}, {"head": "error rate", "head_type": "NA", "head_pos": [45, 55], "relation": "evaluate for", "tail": "HMM and Max-ent models", "tail_type": "NA", "tail_pos": [71, 93]}, {"head": "NIST sentence boundary detection task", "head_type": "NA", "head_pos": [101, 138], "relation": "evaluate for", "tail": "CRF model", "tail_type": "NA", "tail_pos": [20, 29]}, {"head": "NIST sentence boundary detection task", "head_type": "NA", "head_pos": [101, 138], "relation": "evaluate for", "tail": "HMM and Max-ent models", "tail_type": "NA", "tail_pos": [65, 87]}, {"head": "speech", "head_type": "NA", "head_pos": [142, 148], "relation": "feature of", "tail": "NIST sentence boundary detection task", "tail_type": "NA", "tail_pos": [95, 132]}, {"head": "classifiers", "head_type": "NA", "head_pos": [251, 262], "relation": "used for", "tail": "three-way voting", "tail_type": "NA", "tail_pos": [218, 234]}], "task": "RE"}
+{"text": "This probably occurs because each model has different strengths and weaknesses for modeling the knowledge sources .", "relation": [{"head": "model", "head_type": "NA", "head_pos": [37, 42], "relation": "used for", "tail": "knowledge sources", "tail_type": "NA", "tail_pos": [105, 122]}], "task": "RE"}
+{"text": "We propose a novel approach to associate objects across multiple PTZ cameras that can be used to perform camera handoff in wide-area surveillance scenarios .", "relation": [{"head": "approach", "head_type": "NA", "head_pos": [22, 30], "relation": "used for", "tail": "camera handoff in wide-area surveillance scenarios", "tail_type": "NA", "tail_pos": [114, 164]}], "task": "RE"}
+{"text": "While previous approaches relied on geometric , appearance , or correlation-based information for establishing correspondences between static cameras , they each have well-known limitations and are not extendable to wide-area settings with PTZ cameras .", "relation": [{"head": "geometric , appearance , or correlation-based information", "head_type": "NA", "head_pos": [45, 102], "relation": "used for", "tail": "approaches", "tail_type": "NA", "tail_pos": [18, 28]}], "task": "RE"}
+{"text": "Towards this goal , we also propose a novel Multiple Instance Learning -LRB- MIL -RRB- formulation for the problem based on the logistic softmax function of covariance-based region features within a MAP estimation framework .", "relation": [{"head": "logistic softmax function of covariance-based region features", "head_type": "NA", "head_pos": [137, 198], "relation": "used for", "tail": "Multiple Instance Learning -LRB- MIL -RRB- formulation", "tail_type": "NA", "tail_pos": [47, 101]}, {"head": "MAP estimation framework", "head_type": "NA", "head_pos": [208, 232], "relation": "used for", "tail": "Multiple Instance Learning -LRB- MIL -RRB- formulation", "tail_type": "NA", "tail_pos": [47, 101]}], "task": "RE"}
+{"text": "We demonstrate our approach with multiple PTZ camera sequences in typical outdoor surveillance settings and show a comparison with state-of-the-art approaches .", "relation": [{"head": "approach", "head_type": "NA", "head_pos": [22, 30], "relation": "used for", "tail": "outdoor surveillance settings", "tail_type": "NA", "tail_pos": [83, 112]}, {"head": "approach", "head_type": "NA", "head_pos": [22, 30], "relation": "compare", "tail": "state-of-the-art approaches", "tail_type": "NA", "tail_pos": [140, 167]}, {"head": "multiple PTZ camera sequences", "head_type": "NA", "head_pos": [42, 71], "relation": "used for", "tail": "approach", "tail_type": "NA", "tail_pos": [22, 30]}], "task": "RE"}
+{"text": "This paper solves a specialized regression problem to obtain sampling probabilities for records in databases .", "relation": [{"head": "specialized regression problem", "head_type": "NA", "head_pos": [23, 53], "relation": "used for", "tail": "sampling probabilities", "tail_type": "NA", "tail_pos": [70, 92]}, {"head": "sampling probabilities", "head_type": "NA", "head_pos": [64, 86], "relation": "used for", "tail": "records", "tail_type": "NA", "tail_pos": [97, 104]}, {"head": "records", "head_type": "NA", "head_pos": [91, 98], "relation": "part of", "tail": "databases", "tail_type": "NA", "tail_pos": [108, 117]}], "task": "RE"}
+{"text": "The goal is to sample a small set of records over which evaluating aggregate queries can be done both efficiently and accurately .", "relation": [{"head": "aggregate queries", "head_type": "NA", "head_pos": [76, 93], "relation": "evaluate for", "tail": "records", "tail_type": "NA", "tail_pos": [40, 47]}], "task": "RE"}
+{"text": "We provide a principled and provable solution for this problem ; it is parameterless and requires no data insights .", "relation": [{"head": "principled and provable solution", "head_type": "NA", "head_pos": [16, 48], "relation": "used for", "tail": "problem", "tail_type": "NA", "tail_pos": [64, 71]}], "task": "RE"}
+{"text": "Moreover , a cost zero solution always exists and can only be excluded by hard budget constraints .", "relation": [{"head": "hard budget constraints", "head_type": "NA", "head_pos": [83, 106], "relation": "used for", "tail": "cost zero solution", "tail_type": "NA", "tail_pos": [16, 34]}], "task": "RE"}
+{"text": "Our extensive experimental results significantly improve over both uniform sampling and standard stratified sampling which are de-facto the industry standards .", "relation": [{"head": "uniform sampling", "head_type": "NA", "head_pos": [70, 86], "relation": "conjunction", "tail": "stratified sampling", "tail_type": "NA", "tail_pos": [106, 125]}], "task": "RE"}
+{"text": "We consider the problem of computing the Kullback-Leibler distance , also called the relative entropy , between a probabilistic context-free grammar and a probabilistic finite automaton .", "relation": [{"head": "probabilistic context-free grammar", "head_type": "NA", "head_pos": [117, 151], "relation": "compare", "tail": "probabilistic finite automaton", "tail_type": "NA", "tail_pos": [164, 194]}], "task": "RE"}
+{"text": "We show that there is a closed-form -LRB- analytical -RRB- solution for one part of the Kullback-Leibler distance , viz the cross-entropy .", "relation": [{"head": "closed-form -LRB- analytical -RRB- solution", "head_type": "NA", "head_pos": [27, 70], "relation": "used for", "tail": "Kullback-Leibler distance", "tail_type": "NA", "tail_pos": [97, 122]}, {"head": "closed-form -LRB- analytical -RRB- solution", "head_type": "NA", "head_pos": [27, 70], "relation": "used for", "tail": "cross-entropy", "tail_type": "NA", "tail_pos": [133, 146]}, {"head": "cross-entropy", "head_type": "NA", "head_pos": [133, 146], "relation": "part of", "tail": "Kullback-Leibler distance", "tail_type": "NA", "tail_pos": [91, 116]}], "task": "RE"}
+{"text": "We discuss several applications of the result to the problem of distributional approximation of probabilistic context-free grammars by means of probabilistic finite automata .", "relation": [{"head": "distributional approximation", "head_type": "NA", "head_pos": [67, 95], "relation": "feature of", "tail": "probabilistic context-free grammars", "tail_type": "NA", "tail_pos": [105, 140]}, {"head": "probabilistic finite automata", "head_type": "NA", "head_pos": [153, 182], "relation": "used for", "tail": "distributional approximation", "tail_type": "NA", "tail_pos": [67, 95]}], "task": "RE"}
+{"text": "In spite of over two decades of intense research , illumination and pose invariance remain prohibitively challenging aspects of face recognition for most practical applications .", "relation": [{"head": "illumination", "head_type": "NA", "head_pos": [54, 66], "relation": "conjunction", "tail": "pose invariance", "tail_type": "NA", "tail_pos": [77, 92]}, {"head": "illumination", "head_type": "NA", "head_pos": [54, 66], "relation": "part of", "tail": "face recognition", "tail_type": "NA", "tail_pos": [137, 153]}, {"head": "pose invariance", "head_type": "NA", "head_pos": [71, 86], "relation": "part of", "tail": "face recognition", "tail_type": "NA", "tail_pos": [137, 153]}], "task": "RE"}
+{"text": "The objective of this work is to recognize faces using video sequences both for training and recognition input , in a realistic , unconstrained setup in which lighting , pose and user motion pattern have a wide variability and face images are of low resolution .", "relation": [{"head": "lighting", "head_type": "NA", "head_pos": [162, 170], "relation": "conjunction", "tail": "pose", "tail_type": "NA", "tail_pos": [179, 183]}, {"head": "pose", "head_type": "NA", "head_pos": [173, 177], "relation": "conjunction", "tail": "user motion pattern", "tail_type": "NA", "tail_pos": [188, 207]}, {"head": "resolution", "head_type": "NA", "head_pos": [259, 269], "relation": "feature of", "tail": "face images", "tail_type": "NA", "tail_pos": [230, 241]}], "task": "RE"}
+{"text": "In particular there are three areas of novelty : -LRB- i -RRB- we show how a photometric model of image formation can be combined with a statistical model of generic face appearance variation , learnt offline , to generalize in the presence of extreme illumination changes ; -LRB- ii -RRB- we use the smoothness of geodesically local appearance manifold structure and a robust same-identity likelihood to achieve invariance to unseen head poses ; and -LRB- iii -RRB- we introduce an accurate video sequence '' reillumination '' algorithm to achieve robustness to face motion patterns in video .", "relation": [{"head": "photometric model", "head_type": "NA", "head_pos": [80, 97], "relation": "used for", "tail": "image formation", "tail_type": "NA", "tail_pos": [107, 122]}, {"head": "photometric model", "head_type": "NA", "head_pos": [80, 97], "relation": "conjunction", "tail": "statistical model", "tail_type": "NA", "tail_pos": [146, 163]}, {"head": "statistical model", "head_type": "NA", "head_pos": [140, 157], "relation": "used for", "tail": "generic face appearance variation", "tail_type": "NA", "tail_pos": [167, 200]}, {"head": "statistical model", "head_type": "NA", "head_pos": [140, 157], "relation": "used for", "tail": "extreme illumination changes", "tail_type": "NA", "tail_pos": [253, 281]}, {"head": "smoothness", "head_type": "NA", "head_pos": [304, 314], "relation": "feature of", "tail": "geodesically local appearance manifold structure", "tail_type": "NA", "tail_pos": [324, 372]}, {"head": "geodesically local appearance manifold structure", "head_type": "NA", "head_pos": [318, 366], "relation": "conjunction", "tail": "robust same-identity likelihood", "tail_type": "NA", "tail_pos": [379, 410]}, {"head": "robustness", "head_type": "NA", "head_pos": [558, 568], "relation": "evaluate for", "tail": "video sequence '' reillumination '' algorithm", "tail_type": "NA", "tail_pos": [495, 540]}, {"head": "face motion patterns", "head_type": "NA", "head_pos": [572, 592], "relation": "feature of", "tail": "robustness", "tail_type": "NA", "tail_pos": [552, 562]}, {"head": "face motion patterns", "head_type": "NA", "head_pos": [566, 586], "relation": "part of", "tail": "video", "tail_type": "NA", "tail_pos": [492, 497]}], "task": "RE"}
+{"text": "We describe a fully automatic recognition system based on the proposed method and an extensive evaluation on 171 individuals and over 1300 video sequences with extreme illumination , pose and head motion variation .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [80, 86], "relation": "used for", "tail": "fully automatic recognition system", "tail_type": "NA", "tail_pos": [17, 51]}, {"head": "video sequences", "head_type": "NA", "head_pos": [148, 163], "relation": "evaluate for", "tail": "fully automatic recognition system", "tail_type": "NA", "tail_pos": [17, 51]}, {"head": "illumination", "head_type": "NA", "head_pos": [177, 189], "relation": "feature of", "tail": "video sequences", "tail_type": "NA", "tail_pos": [142, 157]}, {"head": "illumination", "head_type": "NA", "head_pos": [171, 183], "relation": "conjunction", "tail": "pose", "tail_type": "NA", "tail_pos": [65, 69]}, {"head": "pose", "head_type": "NA", "head_pos": [65, 69], "relation": "feature of", "tail": "video sequences", "tail_type": "NA", "tail_pos": [142, 157]}, {"head": "pose", "head_type": "NA", "head_pos": [65, 69], "relation": "conjunction", "tail": "head motion variation", "tail_type": "NA", "tail_pos": [201, 222]}, {"head": "head motion variation", "head_type": "NA", "head_pos": [201, 222], "relation": "feature of", "tail": "video sequences", "tail_type": "NA", "tail_pos": [142, 157]}], "task": "RE"}
+{"text": "On this challenging data set our system consistently demonstrated a nearly perfect recognition rate -LRB- over 99.7 % on all three databases -RRB- , significantly out-performing state-of-the-art commercial software and methods from the literature .", "relation": [{"head": "data set", "head_type": "NA", "head_pos": [23, 31], "relation": "evaluate for", "tail": "system", "tail_type": "NA", "tail_pos": [42, 48]}, {"head": "system", "head_type": "NA", "head_pos": [36, 42], "relation": "compare", "tail": "commercial software", "tail_type": "NA", "tail_pos": [204, 223]}, {"head": "system", "head_type": "NA", "head_pos": [36, 42], "relation": "compare", "tail": "methods", "tail_type": "NA", "tail_pos": [228, 235]}, {"head": "recognition rate", "head_type": "NA", "head_pos": [92, 108], "relation": "evaluate for", "tail": "system", "tail_type": "NA", "tail_pos": [36, 42]}, {"head": "commercial software", "head_type": "NA", "head_pos": [198, 217], "relation": "conjunction", "tail": "methods", "tail_type": "NA", "tail_pos": [228, 235]}], "task": "RE"}
+{"text": "We present Minimum Bayes-Risk -LRB- MBR -RRB- decoding for statistical machine translation .", "relation": [{"head": "Minimum Bayes-Risk -LRB- MBR -RRB- decoding", "head_type": "NA", "head_pos": [14, 57], "relation": "used for", "tail": "statistical machine translation", "tail_type": "NA", "tail_pos": [68, 99]}], "task": "RE"}
+{"text": "This statistical approach aims to minimize expected loss of translation errors under loss functions that measure translation performance .", "relation": [{"head": "loss functions", "head_type": "NA", "head_pos": [88, 102], "relation": "evaluate for", "tail": "translation", "tail_type": "NA", "tail_pos": [60, 71]}], "task": "RE"}
+{"text": "We describe a hierarchy of loss functions that incorporate different levels of linguistic information from word strings , word-to-word alignments from an MT system , and syntactic structure from parse-trees of source and target language sentences .", "relation": [{"head": "linguistic information", "head_type": "NA", "head_pos": [88, 110], "relation": "used for", "tail": "loss functions", "tail_type": "NA", "tail_pos": [30, 44]}, {"head": "word-to-word alignments", "head_type": "NA", "head_pos": [131, 154], "relation": "used for", "tail": "loss functions", "tail_type": "NA", "tail_pos": [30, 44]}, {"head": "word-to-word alignments", "head_type": "NA", "head_pos": [125, 148], "relation": "part of", "tail": "MT system", "tail_type": "NA", "tail_pos": [163, 172]}, {"head": "syntactic structure", "head_type": "NA", "head_pos": [179, 198], "relation": "used for", "tail": "loss functions", "tail_type": "NA", "tail_pos": [30, 44]}, {"head": "parse-trees", "head_type": "NA", "head_pos": [204, 215], "relation": "part of", "tail": "syntactic structure", "tail_type": "NA", "tail_pos": [173, 192]}], "task": "RE"}
+{"text": "We report the performance of the MBR decoders on a Chinese-to-English translation task .", "relation": [{"head": "MBR decoders", "head_type": "NA", "head_pos": [36, 48], "relation": "used for", "tail": "Chinese-to-English translation task", "tail_type": "NA", "tail_pos": [60, 95]}], "task": "RE"}
+{"text": "Our results show that MBR decoding can be used to tune statistical MT performance for specific loss functions .", "relation": [{"head": "MBR decoding", "head_type": "NA", "head_pos": [25, 37], "relation": "used for", "tail": "statistical MT", "tail_type": "NA", "tail_pos": [64, 78]}, {"head": "MBR decoding", "head_type": "NA", "head_pos": [25, 37], "relation": "used for", "tail": "loss functions", "tail_type": "NA", "tail_pos": [104, 118]}], "task": "RE"}
+{"text": "This paper presents a critical discussion of the various approaches that have been used in the evaluation of Natural Language systems .", "relation": [{"head": "approaches", "head_type": "NA", "head_pos": [60, 70], "relation": "used for", "tail": "evaluation of Natural Language systems", "tail_type": "NA", "tail_pos": [104, 142]}], "task": "RE"}
+{"text": "We conclude that previous approaches have neglected to evaluate systems in the context of their use , e.g. solving a task requiring data retrieval .", "relation": [{"head": "approaches", "head_type": "NA", "head_pos": [29, 39], "relation": "evaluate for", "tail": "systems", "tail_type": "NA", "tail_pos": [73, 80]}, {"head": "systems", "head_type": "NA", "head_pos": [67, 74], "relation": "used for", "tail": "task", "tail_type": "NA", "tail_pos": [126, 130]}, {"head": "data retrieval", "head_type": "NA", "head_pos": [141, 155], "relation": "part of", "tail": "task", "tail_type": "NA", "tail_pos": [120, 124]}], "task": "RE"}
+{"text": "In the second half of the paper , we report a laboratory study using the Wizard of Oz technique to identify NL requirements for carrying out this task .", "relation": [{"head": "Wizard of Oz technique", "head_type": "NA", "head_pos": [76, 98], "relation": "used for", "tail": "NL requirements", "tail_type": "NA", "tail_pos": [117, 132]}, {"head": "Wizard of Oz technique", "head_type": "NA", "head_pos": [76, 98], "relation": "used for", "tail": "task", "tail_type": "NA", "tail_pos": [155, 159]}], "task": "RE"}
+{"text": "We evaluate the demands that task dialogues collected using this technique , place upon a prototype Natural Language system .", "relation": [{"head": "task dialogues", "head_type": "NA", "head_pos": [32, 46], "relation": "used for", "tail": "prototype Natural Language system", "tail_type": "NA", "tail_pos": [99, 132]}, {"head": "technique", "head_type": "NA", "head_pos": [74, 83], "relation": "used for", "tail": "task dialogues", "tail_type": "NA", "tail_pos": [32, 46]}], "task": "RE"}
+{"text": "We present results on addressee identification in four-participants face-to-face meetings using Bayesian Network and Naive Bayes classifiers .", "relation": [{"head": "Bayesian Network", "head_type": "NA", "head_pos": [105, 121], "relation": "used for", "tail": "addressee identification in four-participants face-to-face meetings", "tail_type": "NA", "tail_pos": [25, 92]}, {"head": "Naive Bayes classifiers", "head_type": "NA", "head_pos": [126, 149], "relation": "used for", "tail": "addressee identification in four-participants face-to-face meetings", "tail_type": "NA", "tail_pos": [25, 92]}, {"head": "Naive Bayes classifiers", "head_type": "NA", "head_pos": [126, 149], "relation": "conjunction", "tail": "Bayesian Network", "tail_type": "NA", "tail_pos": [99, 115]}], "task": "RE"}
+{"text": "First , we investigate how well the addressee of a dialogue act can be predicted based on gaze , utterance and conversational context features .", "relation": [{"head": "gaze", "head_type": "NA", "head_pos": [99, 103], "relation": "used for", "tail": "addressee of a dialogue act", "tail_type": "NA", "tail_pos": [39, 66]}, {"head": "gaze", "head_type": "NA", "head_pos": [93, 97], "relation": "conjunction", "tail": "utterance", "tail_type": "NA", "tail_pos": [106, 115]}, {"head": "utterance", "head_type": "NA", "head_pos": [106, 115], "relation": "used for", "tail": "addressee of a dialogue act", "tail_type": "NA", "tail_pos": [39, 66]}, {"head": "utterance", "head_type": "NA", "head_pos": [100, 109], "relation": "conjunction", "tail": "conversational context features", "tail_type": "NA", "tail_pos": [120, 151]}, {"head": "conversational context features", "head_type": "NA", "head_pos": [120, 151], "relation": "used for", "tail": "addressee of a dialogue act", "tail_type": "NA", "tail_pos": [39, 66]}], "task": "RE"}
+{"text": "Both classifiers perform the best when conversational context and utterance features are combined with speaker 's gaze information .", "relation": [{"head": "conversational context", "head_type": "NA", "head_pos": [48, 70], "relation": "used for", "tail": "classifiers", "tail_type": "NA", "tail_pos": [8, 19]}, {"head": "conversational context", "head_type": "NA", "head_pos": [42, 64], "relation": "conjunction", "tail": "utterance features", "tail_type": "NA", "tail_pos": [75, 93]}, {"head": "utterance features", "head_type": "NA", "head_pos": [75, 93], "relation": "used for", "tail": "classifiers", "tail_type": "NA", "tail_pos": [8, 19]}, {"head": "speaker 's gaze information", "head_type": "NA", "head_pos": [112, 139], "relation": "used for", "tail": "classifiers", "tail_type": "NA", "tail_pos": [8, 19]}, {"head": "speaker 's gaze information", "head_type": "NA", "head_pos": [112, 139], "relation": "conjunction", "tail": "utterance features", "tail_type": "NA", "tail_pos": [69, 87]}], "task": "RE"}
+{"text": "Towards deep analysis of compositional classes of paraphrases , we have examined a class-oriented framework for collecting paraphrase examples , in which sentential paraphrases are collected for each paraphrase class separately by means of automatic candidate generation and manual judgement .", "relation": [{"head": "class-oriented framework", "head_type": "NA", "head_pos": [92, 116], "relation": "used for", "tail": "compositional classes of paraphrases", "tail_type": "NA", "tail_pos": [28, 64]}, {"head": "class-oriented framework", "head_type": "NA", "head_pos": [86, 110], "relation": "used for", "tail": "paraphrase examples", "tail_type": "NA", "tail_pos": [132, 151]}, {"head": "automatic candidate generation", "head_type": "NA", "head_pos": [249, 279], "relation": "used for", "tail": "sentential paraphrases", "tail_type": "NA", "tail_pos": [157, 179]}, {"head": "automatic candidate generation", "head_type": "NA", "head_pos": [243, 273], "relation": "conjunction", "tail": "manual judgement", "tail_type": "NA", "tail_pos": [284, 300]}, {"head": "manual judgement", "head_type": "NA", "head_pos": [284, 300], "relation": "used for", "tail": "sentential paraphrases", "tail_type": "NA", "tail_pos": [157, 179]}], "task": "RE"}
+{"text": "The purpose of this research is to test the efficacy of applying automated evaluation techniques , originally devised for the evaluation of human language learners , to the output of machine translation -LRB- MT -RRB- systems .", "relation": [{"head": "automated evaluation techniques", "head_type": "NA", "head_pos": [68, 99], "relation": "used for", "tail": "evaluation of human language learners", "tail_type": "NA", "tail_pos": [135, 172]}], "task": "RE"}
+{"text": "We believe that these evaluation techniques will provide information about both the human language learning process , the translation process and the development of machine translation systems .", "relation": [{"head": "evaluation techniques", "head_type": "NA", "head_pos": [25, 46], "relation": "used for", "tail": "human language learning process", "tail_type": "NA", "tail_pos": [93, 124]}, {"head": "evaluation techniques", "head_type": "NA", "head_pos": [25, 46], "relation": "used for", "tail": "translation process", "tail_type": "NA", "tail_pos": [131, 150]}, {"head": "evaluation techniques", "head_type": "NA", "head_pos": [25, 46], "relation": "used for", "tail": "machine translation systems", "tail_type": "NA", "tail_pos": [174, 201]}, {"head": "human language learning process", "head_type": "NA", "head_pos": [87, 118], "relation": "conjunction", "tail": "translation process", "tail_type": "NA", "tail_pos": [131, 150]}, {"head": "translation process", "head_type": "NA", "head_pos": [125, 144], "relation": "conjunction", "tail": "machine translation systems", "tail_type": "NA", "tail_pos": [174, 201]}], "task": "RE"}
+{"text": "A language learning experiment showed that assessors can differentiate native from non-native language essays in less than 100 words .", "relation": [{"head": "language learning", "head_type": "NA", "head_pos": [5, 22], "relation": "evaluate for", "tail": "assessors", "tail_type": "NA", "tail_pos": [52, 61]}], "task": "RE"}
+{"text": "Some of the extracts were expert human translations , others were machine translation outputs .", "relation": [{"head": "machine translation outputs", "head_type": "NA", "head_pos": [75, 102], "relation": "conjunction", "tail": "expert human translations", "tail_type": "NA", "tail_pos": [29, 54]}], "task": "RE"}
+{"text": "The subjects were given three minutes per extract to determine whether they believed the sample output to be an expert human translation or a machine translation .", "relation": [{"head": "expert human translation", "head_type": "NA", "head_pos": [115, 139], "relation": "compare", "tail": "machine translation", "tail_type": "NA", "tail_pos": [151, 170]}], "task": "RE"}
+{"text": "This paper presents a machine learning approach to bare slice disambiguation in dialogue .", "relation": [{"head": "machine learning approach", "head_type": "NA", "head_pos": [25, 50], "relation": "used for", "tail": "bare slice disambiguation", "tail_type": "NA", "tail_pos": [60, 85]}, {"head": "dialogue", "head_type": "NA", "head_pos": [89, 97], "relation": "used for", "tail": "bare slice disambiguation", "tail_type": "NA", "tail_pos": [54, 79]}], "task": "RE"}
+{"text": "We extract a set of heuristic principles from a corpus-based sample and formulate them as probabilistic Horn clauses .", "relation": [{"head": "corpus-based sample", "head_type": "NA", "head_pos": [57, 76], "relation": "used for", "tail": "heuristic principles", "tail_type": "NA", "tail_pos": [23, 43]}, {"head": "probabilistic Horn clauses", "head_type": "NA", "head_pos": [99, 125], "relation": "feature of", "tail": "heuristic principles", "tail_type": "NA", "tail_pos": [23, 43]}], "task": "RE"}
+{"text": "We then use the predicates of such clauses to create a set of domain independent features to annotate an input dataset , and run two different machine learning algorithms : SLIPPER , a rule-based learning algorithm , and TiMBL , a memory-based system .", "relation": [{"head": "SLIPPER", "head_type": "NA", "head_pos": [176, 183], "relation": "hyponym of", "tail": "rule-based learning algorithm", "tail_type": "NA", "tail_pos": [194, 223]}, {"head": "rule-based learning algorithm", "head_type": "NA", "head_pos": [194, 223], "relation": "part of", "tail": "machine learning algorithms", "tail_type": "NA", "tail_pos": [146, 173]}, {"head": "rule-based learning algorithm", "head_type": "NA", "head_pos": [188, 217], "relation": "compare", "tail": "memory-based system", "tail_type": "NA", "tail_pos": [240, 259]}, {"head": "TiMBL", "head_type": "NA", "head_pos": [224, 229], "relation": "hyponym of", "tail": "memory-based system", "tail_type": "NA", "tail_pos": [240, 259]}, {"head": "memory-based system", "head_type": "NA", "head_pos": [240, 259], "relation": "part of", "tail": "machine learning algorithms", "tail_type": "NA", "tail_pos": [146, 173]}], "task": "RE"}
+{"text": "The results show that the features in terms of which we formulate our heuristic principles have significant predictive power , and that rules that closely resemble our Horn clauses can be learnt automatically from these features .", "relation": [{"head": "features", "head_type": "NA", "head_pos": [29, 37], "relation": "feature of", "tail": "heuristic principles", "tail_type": "NA", "tail_pos": [79, 99]}], "task": "RE"}
+{"text": "We suggest a new goal and evaluation criterion for word similarity measures .", "relation": [{"head": "evaluation criterion", "head_type": "NA", "head_pos": [29, 49], "relation": "used for", "tail": "word similarity measures", "tail_type": "NA", "tail_pos": [60, 84]}], "task": "RE"}
+{"text": "The new criterion -- meaning-entailing substitutability -- fits the needs of semantic-oriented NLP applications and can be evaluated directly -LRB- independent of an application -RRB- at a good level of human agreement .", "relation": [{"head": "meaning-entailing substitutability", "head_type": "NA", "head_pos": [24, 58], "relation": "used for", "tail": "semantic-oriented NLP applications", "tail_type": "NA", "tail_pos": [86, 120]}, {"head": "human agreement", "head_type": "NA", "head_pos": [212, 227], "relation": "evaluate for", "tail": "meaning-entailing substitutability", "tail_type": "NA", "tail_pos": [24, 58]}], "task": "RE"}
+{"text": "Motivated by this semantic criterion we analyze the empirical quality of distributional word feature vectors and its impact on word similarity results , proposing an objective measure for evaluating feature vector quality .", "relation": [{"head": "semantic criterion", "head_type": "NA", "head_pos": [21, 39], "relation": "evaluate for", "tail": "distributional word feature vectors", "tail_type": "NA", "tail_pos": [82, 117]}, {"head": "distributional word feature vectors", "head_type": "NA", "head_pos": [76, 111], "relation": "used for", "tail": "word similarity", "tail_type": "NA", "tail_pos": [136, 151]}, {"head": "measure", "head_type": "NA", "head_pos": [179, 186], "relation": "evaluate for", "tail": "feature vector quality", "tail_type": "NA", "tail_pos": [208, 230]}], "task": "RE"}
+{"text": "Finally , a novel feature weighting and selection function is presented , which yields superior feature vectors and better word similarity performance .", "relation": [{"head": "feature weighting and selection function", "head_type": "NA", "head_pos": [21, 61], "relation": "used for", "tail": "feature vectors", "tail_type": "NA", "tail_pos": [105, 120]}, {"head": "feature weighting and selection function", "head_type": "NA", "head_pos": [21, 61], "relation": "used for", "tail": "word similarity", "tail_type": "NA", "tail_pos": [132, 147]}, {"head": "feature vectors", "head_type": "NA", "head_pos": [99, 114], "relation": "conjunction", "tail": "word similarity", "tail_type": "NA", "tail_pos": [132, 147]}], "task": "RE"}
+{"text": "This phenomenon causes many image processing techniques to fail as they assume the presence of only one layer at each examined site e.g. motion estimation and object recognition .", "relation": [{"head": "motion estimation", "head_type": "NA", "head_pos": [140, 157], "relation": "conjunction", "tail": "object recognition", "tail_type": "NA", "tail_pos": [168, 186]}], "task": "RE"}
+{"text": "This work presents an automated technique for detecting reflections in image sequences by analyzing motion trajectories of feature points .", "relation": [{"head": "technique", "head_type": "NA", "head_pos": [35, 44], "relation": "used for", "tail": "detecting reflections in image sequences", "tail_type": "NA", "tail_pos": [55, 95]}, {"head": "motion trajectories", "head_type": "NA", "head_pos": [109, 128], "relation": "used for", "tail": "technique", "tail_type": "NA", "tail_pos": [35, 44]}, {"head": "feature points", "head_type": "NA", "head_pos": [132, 146], "relation": "feature of", "tail": "motion trajectories", "tail_type": "NA", "tail_pos": [103, 122]}], "task": "RE"}
+{"text": "It models reflection as regions containing two different layers moving over each other .", "relation": [{"head": "It", "head_type": "NA", "head_pos": [3, 5], "relation": "used for", "tail": "reflection", "tail_type": "NA", "tail_pos": [19, 29]}], "task": "RE"}
+{"text": "We present a strong detector based on combining a set of weak detectors .", "relation": [{"head": "detectors", "head_type": "NA", "head_pos": [71, 80], "relation": "used for", "tail": "detector", "tail_type": "NA", "tail_pos": [23, 31]}], "task": "RE"}
+{"text": "We use novel priors , generate sparse and dense detection maps and our results show high detection rate with rejection to pathological motion and occlusion .", "relation": [{"head": "priors", "head_type": "NA", "head_pos": [16, 22], "relation": "used for", "tail": "sparse and dense detection maps", "tail_type": "NA", "tail_pos": [40, 71]}, {"head": "pathological motion", "head_type": "NA", "head_pos": [125, 144], "relation": "conjunction", "tail": "occlusion", "tail_type": "NA", "tail_pos": [155, 164]}], "task": "RE"}
+{"text": "This paper considers the problem of reconstructing the motion of a 3D articulated tree from 2D point correspondences subject to some temporal prior .", "relation": [{"head": "2D point correspondences", "head_type": "NA", "head_pos": [101, 125], "relation": "used for", "tail": "reconstructing the motion of a 3D articulated tree", "tail_type": "NA", "tail_pos": [39, 89]}], "task": "RE"}
+{"text": "Hitherto , smooth motion has been encouraged using a trajectory basis , yielding a hard combinatorial problem with time complexity growing exponentially in the number of frames .", "relation": [{"head": "trajectory basis", "head_type": "NA", "head_pos": [62, 78], "relation": "used for", "tail": "smooth motion", "tail_type": "NA", "tail_pos": [14, 27]}, {"head": "time complexity", "head_type": "NA", "head_pos": [124, 139], "relation": "evaluate for", "tail": "hard combinatorial problem", "tail_type": "NA", "tail_pos": [86, 112]}], "task": "RE"}
+{"text": "Branch and bound strategies have previously attempted to curb this complexity whilst maintaining global optimality .", "relation": [{"head": "Branch and bound strategies", "head_type": "NA", "head_pos": [3, 30], "relation": "used for", "tail": "complexity", "tail_type": "NA", "tail_pos": [76, 86]}, {"head": "global optimality", "head_type": "NA", "head_pos": [106, 123], "relation": "feature of", "tail": "Branch and bound strategies", "tail_type": "NA", "tail_pos": [3, 30]}], "task": "RE"}
+{"text": "However , they provide no guarantee of being more efficient than exhaustive search .", "relation": [{"head": "they", "head_type": "NA", "head_pos": [13, 17], "relation": "compare", "tail": "exhaustive search", "tail_type": "NA", "tail_pos": [74, 91]}], "task": "RE"}
+{"text": "Extension to affine projection enables reconstruction without estimating cameras .", "relation": [{"head": "affine projection", "head_type": "NA", "head_pos": [16, 33], "relation": "used for", "tail": "reconstruction", "tail_type": "NA", "tail_pos": [48, 62]}], "task": "RE"}
+{"text": "Topical blog post retrieval is the task of ranking blog posts with respect to their relevance for a given topic .", "relation": [{"head": "Topical blog post retrieval", "head_type": "NA", "head_pos": [3, 30], "relation": "hyponym of", "tail": "ranking blog posts", "tail_type": "NA", "tail_pos": [52, 70]}, {"head": "relevance", "head_type": "NA", "head_pos": [93, 102], "relation": "feature of", "tail": "blog posts", "tail_type": "NA", "tail_pos": [54, 64]}], "task": "RE"}
+{"text": "To improve topical blog post retrieval we incorporate textual credibility indicators in the retrieval process .", "relation": [{"head": "textual credibility indicators", "head_type": "NA", "head_pos": [63, 93], "relation": "used for", "tail": "topical blog post retrieval", "tail_type": "NA", "tail_pos": [14, 41]}, {"head": "textual credibility indicators", "head_type": "NA", "head_pos": [57, 87], "relation": "part of", "tail": "retrieval process", "tail_type": "NA", "tail_pos": [101, 118]}], "task": "RE"}
+{"text": "We describe how to estimate these indicators and how to integrate them into a retrieval approach based on language models .", "relation": [{"head": "them", "head_type": "NA", "head_pos": [69, 73], "relation": "part of", "tail": "retrieval approach", "tail_type": "NA", "tail_pos": [87, 105]}, {"head": "language models", "head_type": "NA", "head_pos": [115, 130], "relation": "used for", "tail": "them", "tail_type": "NA", "tail_pos": [69, 73]}], "task": "RE"}
+{"text": "Experiments on the TREC Blog track test set show that both groups of credibility indicators significantly improve retrieval effectiveness ; the best performance is achieved when combining them .", "relation": [{"head": "TREC Blog track test set", "head_type": "NA", "head_pos": [22, 46], "relation": "evaluate for", "tail": "credibility indicators", "tail_type": "NA", "tail_pos": [78, 100]}, {"head": "retrieval effectiveness", "head_type": "NA", "head_pos": [123, 146], "relation": "evaluate for", "tail": "credibility indicators", "tail_type": "NA", "tail_pos": [72, 94]}], "task": "RE"}
+{"text": "We investigate the problem of learning to predict moves in the board game of Go from game records of expert players .", "relation": [{"head": "game records of expert players", "head_type": "NA", "head_pos": [94, 124], "relation": "used for", "tail": "board game of Go", "tail_type": "NA", "tail_pos": [66, 82]}], "task": "RE"}
+{"text": "This distribution has numerous applications in computer Go , including serving as an efficient stand-alone Go player .", "relation": [{"head": "distribution", "head_type": "NA", "head_pos": [8, 20], "relation": "used for", "tail": "computer Go", "tail_type": "NA", "tail_pos": [56, 67]}], "task": "RE"}
+{"text": "It would also be effective as a move selector and move sorter for game tree search and as a training tool for Go players .", "relation": [{"head": "It", "head_type": "NA", "head_pos": [3, 5], "relation": "used for", "tail": "move selector", "tail_type": "NA", "tail_pos": [41, 54]}, {"head": "It", "head_type": "NA", "head_pos": [3, 5], "relation": "used for", "tail": "move sorter", "tail_type": "NA", "tail_pos": [59, 70]}, {"head": "It", "head_type": "NA", "head_pos": [3, 5], "relation": "used for", "tail": "training tool", "tail_type": "NA", "tail_pos": [101, 114]}, {"head": "move selector", "head_type": "NA", "head_pos": [35, 48], "relation": "conjunction", "tail": "move sorter", "tail_type": "NA", "tail_pos": [59, 70]}, {"head": "move selector", "head_type": "NA", "head_pos": [35, 48], "relation": "used for", "tail": "game tree search", "tail_type": "NA", "tail_pos": [75, 91]}, {"head": "move sorter", "head_type": "NA", "head_pos": [53, 64], "relation": "used for", "tail": "game tree search", "tail_type": "NA", "tail_pos": [75, 91]}, {"head": "training tool", "head_type": "NA", "head_pos": [95, 108], "relation": "used for", "tail": "Go players", "tail_type": "NA", "tail_pos": [119, 129]}], "task": "RE"}
+{"text": "Our method has two major components : a -RRB- a pattern extraction scheme for efficiently harvesting patterns of given size and shape from expert game records and b -RRB- a Bayesian learning algorithm -LRB- in two variants -RRB- that learns a distribution over the values of a move given a board position based on the local pattern context .", "relation": [{"head": "pattern extraction scheme", "head_type": "NA", "head_pos": [57, 82], "relation": "part of", "tail": "method", "tail_type": "NA", "tail_pos": [7, 13]}, {"head": "pattern extraction scheme", "head_type": "NA", "head_pos": [51, 76], "relation": "conjunction", "tail": "Bayesian learning algorithm", "tail_type": "NA", "tail_pos": [182, 209]}, {"head": "Bayesian learning algorithm", "head_type": "NA", "head_pos": [182, 209], "relation": "part of", "tail": "method", "tail_type": "NA", "tail_pos": [7, 13]}], "task": "RE"}
+{"text": "The system is trained on 181,000 expert games and shows excellent prediction performance as indicated by its ability to perfectly predict the moves made by professional Go players in 34 % of test positions .", "relation": [{"head": "expert games", "head_type": "NA", "head_pos": [42, 54], "relation": "used for", "tail": "system", "tail_type": "NA", "tail_pos": [7, 13]}], "task": "RE"}
+{"text": "We present a novel approach for automatically acquiring English topic signatures .", "relation": [{"head": "approach", "head_type": "NA", "head_pos": [22, 30], "relation": "used for", "tail": "automatically acquiring English topic signatures", "tail_type": "NA", "tail_pos": [41, 89]}], "task": "RE"}
+{"text": "Topic signatures can be useful in a number of Natural Language Processing -LRB- NLP -RRB- applications , such as Word Sense Disambiguation -LRB- WSD -RRB- and Text Summarisation .", "relation": [{"head": "Topic signatures", "head_type": "NA", "head_pos": [3, 19], "relation": "used for", "tail": "Natural Language Processing -LRB- NLP -RRB- applications", "tail_type": "NA", "tail_pos": [55, 111]}, {"head": "Topic signatures", "head_type": "NA", "head_pos": [3, 19], "relation": "used for", "tail": "Word Sense Disambiguation -LRB- WSD -RRB-", "tail_type": "NA", "tail_pos": [122, 163]}, {"head": "Topic signatures", "head_type": "NA", "head_pos": [3, 19], "relation": "used for", "tail": "Text Summarisation", "tail_type": "NA", "tail_pos": [168, 186]}, {"head": "Word Sense Disambiguation -LRB- WSD -RRB-", "head_type": "NA", "head_pos": [122, 163], "relation": "hyponym of", "tail": "Natural Language Processing -LRB- NLP -RRB- applications", "tail_type": "NA", "tail_pos": [49, 105]}, {"head": "Word Sense Disambiguation -LRB- WSD -RRB-", "head_type": "NA", "head_pos": [116, 157], "relation": "conjunction", "tail": "Text Summarisation", "tail_type": "NA", "tail_pos": [168, 186]}, {"head": "Text Summarisation", "head_type": "NA", "head_pos": [168, 186], "relation": "hyponym of", "tail": "Natural Language Processing -LRB- NLP -RRB- applications", "tail_type": "NA", "tail_pos": [49, 105]}], "task": "RE"}
+{"text": "Our method takes advantage of the different way in which word senses are lexicalised in English and Chinese , and also exploits the large amount of Chinese text available in corpora and on the Web .", "relation": [{"head": "Chinese text", "head_type": "NA", "head_pos": [151, 163], "relation": "part of", "tail": "corpora", "tail_type": "NA", "tail_pos": [183, 190]}, {"head": "Chinese text", "head_type": "NA", "head_pos": [151, 163], "relation": "part of", "tail": "Web", "tail_type": "NA", "tail_pos": [202, 205]}, {"head": "corpora", "head_type": "NA", "head_pos": [177, 184], "relation": "conjunction", "tail": "Web", "tail_type": "NA", "tail_pos": [202, 205]}], "task": "RE"}
+{"text": "We evaluated the topic signatures on a WSD task , where we trained a second-order vector cooccurrence algorithm on standard WSD datasets , with promising results .", "relation": [{"head": "WSD task", "head_type": "NA", "head_pos": [48, 56], "relation": "evaluate for", "tail": "topic signatures", "tail_type": "NA", "tail_pos": [20, 36]}, {"head": "WSD datasets", "head_type": "NA", "head_pos": [133, 145], "relation": "used for", "tail": "second-order vector cooccurrence algorithm", "tail_type": "NA", "tail_pos": [72, 114]}], "task": "RE"}
+{"text": "Joint matrix triangularization is often used for estimating the joint eigenstructure of a set M of matrices , with applications in signal processing and machine learning .", "relation": [{"head": "Joint matrix triangularization", "head_type": "NA", "head_pos": [3, 33], "relation": "used for", "tail": "joint eigenstructure", "tail_type": "NA", "tail_pos": [73, 93]}, {"head": "joint eigenstructure", "head_type": "NA", "head_pos": [67, 87], "relation": "used for", "tail": "signal processing", "tail_type": "NA", "tail_pos": [140, 157]}, {"head": "joint eigenstructure", "head_type": "NA", "head_pos": [67, 87], "relation": "used for", "tail": "machine learning", "tail_type": "NA", "tail_pos": [162, 178]}, {"head": "signal processing", "head_type": "NA", "head_pos": [134, 151], "relation": "conjunction", "tail": "machine learning", "tail_type": "NA", "tail_pos": [162, 178]}], "task": "RE"}
+{"text": "Our main result is a first-order upper bound on the distance between any approximate joint triangularizer of the matrices in M ' and any exact joint triangularizer of the matrices in M .", "relation": [{"head": "approximate joint triangularizer", "head_type": "NA", "head_pos": [76, 108], "relation": "conjunction", "tail": "exact joint triangularizer", "tail_type": "NA", "tail_pos": [146, 172]}], "task": "RE"}
+{"text": "To our knowledge , this is the first a posteriori bound for joint matrix decomposition .", "relation": [{"head": "posteriori bound", "head_type": "NA", "head_pos": [42, 58], "relation": "used for", "tail": "joint matrix decomposition", "tail_type": "NA", "tail_pos": [69, 95]}], "task": "RE"}
+{"text": "The psycholinguistic literature provides evidence for syntactic priming , i.e. , the tendency to repeat structures .", "relation": [{"head": "psycholinguistic literature", "head_type": "NA", "head_pos": [7, 34], "relation": "used for", "tail": "syntactic priming", "tail_type": "NA", "tail_pos": [63, 80]}], "task": "RE"}
+{"text": "This paper describes a method for incorporating priming into an incremental probabilistic parser .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [26, 32], "relation": "used for", "tail": "priming", "tail_type": "NA", "tail_pos": [57, 64]}, {"head": "priming", "head_type": "NA", "head_pos": [51, 58], "relation": "used for", "tail": "incremental probabilistic parser", "tail_type": "NA", "tail_pos": [73, 105]}], "task": "RE"}
+{"text": "These models simulate the reading time advantage for parallel structures found in human data , and also yield a small increase in overall parsing accuracy .", "relation": [{"head": "parallel structures", "head_type": "NA", "head_pos": [56, 75], "relation": "part of", "tail": "human data", "tail_type": "NA", "tail_pos": [91, 101]}], "task": "RE"}
+{"text": "Learned confidence measures gain increasing importance for outlier removal and quality improvement in stereo vision .", "relation": [{"head": "Learned confidence measures", "head_type": "NA", "head_pos": [3, 30], "relation": "used for", "tail": "outlier removal", "tail_type": "NA", "tail_pos": [68, 83]}, {"head": "Learned confidence measures", "head_type": "NA", "head_pos": [3, 30], "relation": "used for", "tail": "quality improvement", "tail_type": "NA", "tail_pos": [88, 107]}, {"head": "outlier removal", "head_type": "NA", "head_pos": [62, 77], "relation": "conjunction", "tail": "quality improvement", "tail_type": "NA", "tail_pos": [88, 107]}, {"head": "outlier removal", "head_type": "NA", "head_pos": [62, 77], "relation": "part of", "tail": "stereo vision", "tail_type": "NA", "tail_pos": [111, 124]}, {"head": "quality improvement", "head_type": "NA", "head_pos": [82, 101], "relation": "part of", "tail": "stereo vision", "tail_type": "NA", "tail_pos": [111, 124]}], "task": "RE"}
+{"text": "However , acquiring the necessary training data is typically a tedious and time consuming task that involves manual interaction , active sensing devices and/or synthetic scenes .", "relation": [{"head": "manual interaction", "head_type": "NA", "head_pos": [118, 136], "relation": "used for", "tail": "task", "tail_type": "NA", "tail_pos": [93, 97]}, {"head": "manual interaction", "head_type": "NA", "head_pos": [112, 130], "relation": "conjunction", "tail": "active sensing devices", "tail_type": "NA", "tail_pos": [139, 161]}, {"head": "active sensing devices", "head_type": "NA", "head_pos": [139, 161], "relation": "used for", "tail": "task", "tail_type": "NA", "tail_pos": [93, 97]}, {"head": "active sensing devices", "head_type": "NA", "head_pos": [133, 155], "relation": "conjunction", "tail": "synthetic scenes", "tail_type": "NA", "tail_pos": [169, 185]}, {"head": "synthetic scenes", "head_type": "NA", "head_pos": [169, 185], "relation": "used for", "tail": "task", "tail_type": "NA", "tail_pos": [93, 97]}], "task": "RE"}
+{"text": "The key idea of our approach is to use different view points for reasoning about contradictions and consistencies between multiple depth maps generated with the same stereo algorithm .", "relation": [{"head": "view points", "head_type": "NA", "head_pos": [58, 69], "relation": "used for", "tail": "approach", "tail_type": "NA", "tail_pos": [23, 31]}], "task": "RE"}
+{"text": "Among other experiments , we demonstrate the potential of our approach by boosting the performance of three learned confidence measures on the KITTI2012 dataset by simply training them on a vast amount of automatically generated training data rather than a limited amount of laser ground truth data .", "relation": [{"head": "approach", "head_type": "NA", "head_pos": [65, 73], "relation": "used for", "tail": "learned confidence measures", "tail_type": "NA", "tail_pos": [117, 144]}, {"head": "KITTI2012 dataset", "head_type": "NA", "head_pos": [152, 169], "relation": "evaluate for", "tail": "learned confidence measures", "tail_type": "NA", "tail_pos": [111, 138]}, {"head": "automatically generated training data", "head_type": "NA", "head_pos": [214, 251], "relation": "used for", "tail": "them", "tail_type": "NA", "tail_pos": [183, 187]}, {"head": "laser ground truth data", "head_type": "NA", "head_pos": [284, 307], "relation": "compare", "tail": "automatically generated training data", "tail_type": "NA", "tail_pos": [208, 245]}], "task": "RE"}
+{"text": "An important area of learning in autonomous agents is the ability to learn domain-speciic models of actions to be used by planning systems .", "relation": [{"head": "learning in autonomous agents", "head_type": "NA", "head_pos": [24, 53], "relation": "used for", "tail": "domain-speciic models of actions", "tail_type": "NA", "tail_pos": [84, 116]}, {"head": "planning systems", "head_type": "NA", "head_pos": [131, 147], "relation": "used for", "tail": "domain-speciic models of actions", "tail_type": "NA", "tail_pos": [78, 110]}], "task": "RE"}
+{"text": "These methods diier from previous work in the area in two ways : the use of an action model formalism which is better suited to the needs of a re-active agent , and successful implementation of noise-handling mechanisms .", "relation": [{"head": "action model formalism", "head_type": "NA", "head_pos": [88, 110], "relation": "used for", "tail": "methods", "tail_type": "NA", "tail_pos": [9, 16]}, {"head": "action model formalism", "head_type": "NA", "head_pos": [82, 104], "relation": "used for", "tail": "re-active agent", "tail_type": "NA", "tail_pos": [152, 167]}, {"head": "noise-handling mechanisms", "head_type": "NA", "head_pos": [203, 228], "relation": "used for", "tail": "methods", "tail_type": "NA", "tail_pos": [9, 16]}], "task": "RE"}
+{"text": "Training instances are generated from experience and observation , and a variant of GOLEM is used to learn action models from these instances .", "relation": [{"head": "GOLEM", "head_type": "NA", "head_pos": [87, 92], "relation": "used for", "tail": "action models", "tail_type": "NA", "tail_pos": [116, 129]}], "task": "RE"}
+{"text": "The integrated learning system has been experimentally validated in simulated construction and ooce domains .", "relation": [{"head": "simulated construction", "head_type": "NA", "head_pos": [77, 99], "relation": "evaluate for", "tail": "integrated learning system", "tail_type": "NA", "tail_pos": [7, 33]}, {"head": "simulated construction", "head_type": "NA", "head_pos": [71, 93], "relation": "conjunction", "tail": "ooce domains", "tail_type": "NA", "tail_pos": [104, 116]}, {"head": "ooce domains", "head_type": "NA", "head_pos": [104, 116], "relation": "evaluate for", "tail": "integrated learning system", "tail_type": "NA", "tail_pos": [7, 33]}], "task": "RE"}
+{"text": "This paper describes FERRET , an interactive question-answering -LRB- Q/A -RRB- system designed to address the challenges of integrating automatic Q/A applications into real-world environments .", "relation": [{"head": "FERRET", "head_type": "NA", "head_pos": [24, 30], "relation": "hyponym of", "tail": "interactive question-answering -LRB- Q/A -RRB- system", "tail_type": "NA", "tail_pos": [42, 95]}, {"head": "FERRET", "head_type": "NA", "head_pos": [24, 30], "relation": "used for", "tail": "integrating automatic Q/A applications into real-world environments", "tail_type": "NA", "tail_pos": [134, 201]}], "task": "RE"}
+{"text": "FERRET utilizes a novel approach to Q/A known as predictive questioning which attempts to identify the questions -LRB- and answers -RRB- that users need by analyzing how a user interacts with a system while gathering information related to a particular scenario .", "relation": [{"head": "approach", "head_type": "NA", "head_pos": [33, 41], "relation": "used for", "tail": "FERRET", "tail_type": "NA", "tail_pos": [3, 9]}, {"head": "approach", "head_type": "NA", "head_pos": [27, 35], "relation": "used for", "tail": "Q/A", "tail_type": "NA", "tail_pos": [45, 48]}], "task": "RE"}
+{"text": "In order to build robust automatic abstracting systems , there is a need for better training resources than are currently available .", "relation": [{"head": "training resources", "head_type": "NA", "head_pos": [93, 111], "relation": "used for", "tail": "automatic abstracting systems", "tail_type": "NA", "tail_pos": [28, 57]}], "task": "RE"}
+{"text": "In this paper , we introduce an annotation scheme for scientific articles which can be used to build such a resource in a consistent way .", "relation": [{"head": "annotation scheme", "head_type": "NA", "head_pos": [35, 52], "relation": "used for", "tail": "scientific articles", "tail_type": "NA", "tail_pos": [63, 82]}, {"head": "annotation scheme", "head_type": "NA", "head_pos": [35, 52], "relation": "used for", "tail": "resource", "tail_type": "NA", "tail_pos": [117, 125]}], "task": "RE"}
+{"text": "The seven categories of the scheme are based on rhetorical moves of argumentation .", "relation": [{"head": "rhetorical moves of argumentation", "head_type": "NA", "head_pos": [57, 90], "relation": "used for", "tail": "scheme", "tail_type": "NA", "tail_pos": [31, 37]}], "task": "RE"}
+{"text": "The automated segmentation of images into semantically meaningful parts requires shape information since low-level feature analysis alone often fails to reach this goal .", "relation": [{"head": "images", "head_type": "NA", "head_pos": [39, 45], "relation": "used for", "tail": "automated segmentation", "tail_type": "NA", "tail_pos": [7, 29]}], "task": "RE"}
+{"text": "We introduce a novel method of shape constrained image segmentation which is based on mixtures of feature distributions for color and texture as well as probabilistic shape knowledge .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [24, 30], "relation": "used for", "tail": "shape constrained image segmentation", "tail_type": "NA", "tail_pos": [40, 76]}, {"head": "mixtures of feature distributions", "head_type": "NA", "head_pos": [95, 128], "relation": "used for", "tail": "method", "tail_type": "NA", "tail_pos": [24, 30]}, {"head": "mixtures of feature distributions", "head_type": "NA", "head_pos": [89, 122], "relation": "used for", "tail": "color", "tail_type": "NA", "tail_pos": [133, 138]}, {"head": "mixtures of feature distributions", "head_type": "NA", "head_pos": [89, 122], "relation": "used for", "tail": "texture", "tail_type": "NA", "tail_pos": [143, 150]}, {"head": "mixtures of feature distributions", "head_type": "NA", "head_pos": [89, 122], "relation": "used for", "tail": "probabilistic shape knowledge", "tail_type": "NA", "tail_pos": [162, 191]}, {"head": "color", "head_type": "NA", "head_pos": [127, 132], "relation": "conjunction", "tail": "texture", "tail_type": "NA", "tail_pos": [143, 150]}, {"head": "texture", "head_type": "NA", "head_pos": [137, 144], "relation": "conjunction", "tail": "probabilistic shape knowledge", "tail_type": "NA", "tail_pos": [162, 191]}], "task": "RE"}
+{"text": "The combined approach is formulated in the framework of Bayesian statistics to account for the robust-ness requirement in image understanding .", "relation": [{"head": "approach", "head_type": "NA", "head_pos": [16, 24], "relation": "used for", "tail": "robust-ness requirement in image understanding", "tail_type": "NA", "tail_pos": [104, 150]}, {"head": "Bayesian statistics", "head_type": "NA", "head_pos": [65, 84], "relation": "used for", "tail": "approach", "tail_type": "NA", "tail_pos": [16, 24]}], "task": "RE"}
+{"text": "The goal of this work is the enrichment of human-machine interactions in a natural language environment .", "relation": [{"head": "natural language environment", "head_type": "NA", "head_pos": [84, 112], "relation": "feature of", "tail": "human-machine interactions", "tail_type": "NA", "tail_pos": [46, 72]}], "task": "RE"}
+{"text": "This paper highlights a particular class of miscommunication -- reference problems -- by describing a case study and techniques for avoiding failures of reference .", "relation": [{"head": "reference problems", "head_type": "NA", "head_pos": [73, 91], "relation": "hyponym of", "tail": "miscommunication", "tail_type": "NA", "tail_pos": [47, 63]}, {"head": "techniques", "head_type": "NA", "head_pos": [120, 130], "relation": "used for", "tail": "failures of reference", "tail_type": "NA", "tail_pos": [150, 171]}], "task": "RE"}
+{"text": "This paper examines the benefits of system combination for unsupervised WSD .", "relation": [{"head": "system combination", "head_type": "NA", "head_pos": [39, 57], "relation": "used for", "tail": "unsupervised WSD", "tail_type": "NA", "tail_pos": [68, 84]}], "task": "RE"}
+{"text": "We investigate several voting - and arbiter-based combination strategies over a diverse pool of unsupervised WSD systems .", "relation": [{"head": "voting - and arbiter-based combination strategies", "head_type": "NA", "head_pos": [26, 75], "relation": "used for", "tail": "unsupervised WSD systems", "tail_type": "NA", "tail_pos": [105, 129]}], "task": "RE"}
+{"text": "Our combination methods rely on predominant senses which are derived automatically from raw text .", "relation": [{"head": "predominant senses", "head_type": "NA", "head_pos": [41, 59], "relation": "used for", "tail": "combination methods", "tail_type": "NA", "tail_pos": [7, 26]}, {"head": "raw text", "head_type": "NA", "head_pos": [97, 105], "relation": "used for", "tail": "predominant senses", "tail_type": "NA", "tail_pos": [35, 53]}], "task": "RE"}
+{"text": "Experiments using the SemCor and Senseval-3 data sets demonstrate that our ensembles yield significantly better results when compared with state-of-the-art .", "relation": [{"head": "SemCor and Senseval-3 data sets", "head_type": "NA", "head_pos": [25, 56], "relation": "evaluate for", "tail": "ensembles", "tail_type": "NA", "tail_pos": [84, 93]}, {"head": "SemCor and Senseval-3 data sets", "head_type": "NA", "head_pos": [25, 56], "relation": "evaluate for", "tail": "state-of-the-art", "tail_type": "NA", "tail_pos": [148, 164]}], "task": "RE"}
+{"text": "The applicability of many current information extraction techniques is severely limited by the need for supervised training data .", "relation": [{"head": "supervised training data", "head_type": "NA", "head_pos": [113, 137], "relation": "used for", "tail": "information extraction techniques", "tail_type": "NA", "tail_pos": [37, 70]}], "task": "RE"}
+{"text": "We demonstrate that for certain field structured extraction tasks , such as classified advertisements and bibliographic citations , small amounts of prior knowledge can be used to learn effective models in a primarily unsupervised fashion .", "relation": [{"head": "classified advertisements", "head_type": "NA", "head_pos": [85, 110], "relation": "hyponym of", "tail": "field structured extraction tasks", "tail_type": "NA", "tail_pos": [35, 68]}, {"head": "classified advertisements", "head_type": "NA", "head_pos": [79, 104], "relation": "conjunction", "tail": "bibliographic citations", "tail_type": "NA", "tail_pos": [115, 138]}, {"head": "bibliographic citations", "head_type": "NA", "head_pos": [115, 138], "relation": "hyponym of", "tail": "field structured extraction tasks", "tail_type": "NA", "tail_pos": [35, 68]}, {"head": "prior knowledge", "head_type": "NA", "head_pos": [158, 173], "relation": "used for", "tail": "field structured extraction tasks", "tail_type": "NA", "tail_pos": [35, 68]}], "task": "RE"}
+{"text": "Although hidden Markov models -LRB- HMMs -RRB- provide a suitable generative model for field structured text , general unsupervised HMM learning fails to learn useful structure in either of our domains .", "relation": [{"head": "hidden Markov models -LRB- HMMs -RRB-", "head_type": "NA", "head_pos": [12, 49], "relation": "used for", "tail": "generative model", "tail_type": "NA", "tail_pos": [75, 91]}, {"head": "generative model", "head_type": "NA", "head_pos": [69, 85], "relation": "used for", "tail": "field structured text", "tail_type": "NA", "tail_pos": [96, 117]}], "task": "RE"}
+{"text": "In both domains , we found that unsupervised methods can attain accuracies with 400 unlabeled examples comparable to those attained by supervised methods on 50 labeled examples , and that semi-supervised methods can make good use of small amounts of labeled data .", "relation": [{"head": "unsupervised methods", "head_type": "NA", "head_pos": [35, 55], "relation": "compare", "tail": "supervised methods", "tail_type": "NA", "tail_pos": [37, 55]}, {"head": "accuracies", "head_type": "NA", "head_pos": [73, 83], "relation": "evaluate for", "tail": "unsupervised methods", "tail_type": "NA", "tail_pos": [35, 55]}, {"head": "accuracies", "head_type": "NA", "head_pos": [67, 77], "relation": "evaluate for", "tail": "supervised methods", "tail_type": "NA", "tail_pos": [34, 52]}, {"head": "unlabeled examples", "head_type": "NA", "head_pos": [93, 111], "relation": "used for", "tail": "unsupervised methods", "tail_type": "NA", "tail_pos": [35, 55]}, {"head": "labeled examples", "head_type": "NA", "head_pos": [86, 102], "relation": "used for", "tail": "supervised methods", "tail_type": "NA", "tail_pos": [34, 52]}, {"head": "labeled data", "head_type": "NA", "head_pos": [259, 271], "relation": "used for", "tail": "semi-supervised methods", "tail_type": "NA", "tail_pos": [191, 214]}], "task": "RE"}
+{"text": "This paper gives an overall account of a prototype natural language question answering system , called Chat-80 .", "relation": [{"head": "Chat-80", "head_type": "NA", "head_pos": [112, 119], "relation": "hyponym of", "tail": "natural language question answering system", "tail_type": "NA", "tail_pos": [54, 96]}], "task": "RE"}
+{"text": "The system is implemented entirely in Prolog , a programming language based on logic .", "relation": [{"head": "Prolog", "head_type": "NA", "head_pos": [47, 53], "relation": "used for", "tail": "system", "tail_type": "NA", "tail_pos": [7, 13]}, {"head": "Prolog", "head_type": "NA", "head_pos": [41, 47], "relation": "hyponym of", "tail": "programming language", "tail_type": "NA", "tail_pos": [58, 78]}, {"head": "logic", "head_type": "NA", "head_pos": [88, 93], "relation": "used for", "tail": "programming language", "tail_type": "NA", "tail_pos": [52, 72]}], "task": "RE"}
+{"text": "With the aid of a logic-based grammar formalism called extraposition grammars , Chat-80 translates English questions into the Prolog subset of logic .", "relation": [{"head": "extraposition grammars", "head_type": "NA", "head_pos": [64, 86], "relation": "hyponym of", "tail": "logic-based grammar formalism", "tail_type": "NA", "tail_pos": [21, 50]}, {"head": "extraposition grammars", "head_type": "NA", "head_pos": [58, 80], "relation": "used for", "tail": "Chat-80", "tail_type": "NA", "tail_pos": [89, 96]}], "task": "RE"}
+{"text": "The resulting logical expression is then transformed by a planning algorithm into efficient Prolog , cf. query optimisation in a relational database .", "relation": [{"head": "planning algorithm", "head_type": "NA", "head_pos": [67, 85], "relation": "used for", "tail": "logical expression", "tail_type": "NA", "tail_pos": [17, 35]}, {"head": "relational database", "head_type": "NA", "head_pos": [138, 157], "relation": "used for", "tail": "query optimisation", "tail_type": "NA", "tail_pos": [108, 126]}], "task": "RE"}
+{"text": "Human action recognition from well-segmented 3D skeleton data has been intensively studied and attracting an increasing attention .", "relation": [{"head": "well-segmented 3D skeleton data", "head_type": "NA", "head_pos": [39, 70], "relation": "used for", "tail": "Human action recognition", "tail_type": "NA", "tail_pos": [3, 27]}], "task": "RE"}
+{"text": "Online action detection goes one step further and is more challenging , which identifies the action type and localizes the action positions on the fly from the untrimmed stream .", "relation": [{"head": "Online action detection", "head_type": "NA", "head_pos": [3, 26], "relation": "used for", "tail": "action type", "tail_type": "NA", "tail_pos": [102, 113]}, {"head": "Online action detection", "head_type": "NA", "head_pos": [3, 26], "relation": "used for", "tail": "action positions", "tail_type": "NA", "tail_pos": [132, 148]}, {"head": "action type", "head_type": "NA", "head_pos": [96, 107], "relation": "conjunction", "tail": "action positions", "tail_type": "NA", "tail_pos": [132, 148]}, {"head": "untrimmed stream", "head_type": "NA", "head_pos": [169, 185], "relation": "used for", "tail": "Online action detection", "tail_type": "NA", "tail_pos": [3, 26]}], "task": "RE"}
+{"text": "In this paper , we study the problem of online action detection from the streaming skeleton data .", "relation": [{"head": "streaming skeleton data", "head_type": "NA", "head_pos": [82, 105], "relation": "used for", "tail": "online action detection", "tail_type": "NA", "tail_pos": [43, 66]}], "task": "RE"}
+{"text": "We propose a multi-task end-to-end Joint Classification-Regression Recurrent Neural Network to better explore the action type and temporal localiza-tion information .", "relation": [{"head": "multi-task end-to-end Joint Classification-Regression Recurrent Neural Network", "head_type": "NA", "head_pos": [16, 94], "relation": "used for", "tail": "action type", "tail_type": "NA", "tail_pos": [123, 134]}, {"head": "multi-task end-to-end Joint Classification-Regression Recurrent Neural Network", "head_type": "NA", "head_pos": [16, 94], "relation": "used for", "tail": "temporal localiza-tion information", "tail_type": "NA", "tail_pos": [139, 173]}, {"head": "action type", "head_type": "NA", "head_pos": [117, 128], "relation": "conjunction", "tail": "temporal localiza-tion information", "tail_type": "NA", "tail_pos": [139, 173]}], "task": "RE"}
+{"text": "By employing a joint classification and regression optimization objective , this network is capable of automatically localizing the start and end points of actions more accurately .", "relation": [{"head": "joint classification and regression optimization objective", "head_type": "NA", "head_pos": [18, 76], "relation": "used for", "tail": "network", "tail_type": "NA", "tail_pos": [90, 97]}], "task": "RE"}
+{"text": "Specifically , by leveraging the merits of the deep Long Short-Term Memory -LRB- LSTM -RRB- subnetwork , the proposed model automatically captures the complex long-range temporal dynamics , which naturally avoids the typical sliding window design and thus ensures high computational efficiency .", "relation": [{"head": "deep Long Short-Term Memory -LRB- LSTM -RRB- subnetwork", "head_type": "NA", "head_pos": [50, 105], "relation": "used for", "tail": "model", "tail_type": "NA", "tail_pos": [127, 132]}, {"head": "long-range temporal dynamics", "head_type": "NA", "head_pos": [168, 196], "relation": "feature of", "tail": "model", "tail_type": "NA", "tail_pos": [121, 126]}], "task": "RE"}
+{"text": "To evaluate our proposed model , we build a large streaming video dataset with annotations .", "relation": [{"head": "streaming video dataset", "head_type": "NA", "head_pos": [59, 82], "relation": "evaluate for", "tail": "model", "tail_type": "NA", "tail_pos": [28, 33]}], "task": "RE"}
+{"text": "Experimental results on our dataset and the public G3D dataset both demonstrate very promising performance of our scheme .", "relation": [{"head": "dataset", "head_type": "NA", "head_pos": [31, 38], "relation": "conjunction", "tail": "G3D dataset", "tail_type": "NA", "tail_pos": [60, 71]}], "task": "RE"}
+{"text": "The task of machine translation -LRB- MT -RRB- evaluation is closely related to the task of sentence-level semantic equivalence classification .", "relation": [{"head": "machine translation -LRB- MT -RRB- evaluation", "head_type": "NA", "head_pos": [15, 60], "relation": "conjunction", "tail": "sentence-level semantic equivalence classification", "tail_type": "NA", "tail_pos": [101, 151]}], "task": "RE"}
+{"text": "This paper investigates the utility of applying standard MT evaluation methods -LRB- BLEU , NIST , WER and PER -RRB- to building classifiers to predict semantic equivalence and entailment .", "relation": [{"head": "MT evaluation methods", "head_type": "NA", "head_pos": [60, 81], "relation": "used for", "tail": "classifiers", "tail_type": "NA", "tail_pos": [138, 149]}, {"head": "BLEU", "head_type": "NA", "head_pos": [94, 98], "relation": "hyponym of", "tail": "MT evaluation methods", "tail_type": "NA", "tail_pos": [60, 81]}, {"head": "BLEU", "head_type": "NA", "head_pos": [88, 92], "relation": "conjunction", "tail": "NIST", "tail_type": "NA", "tail_pos": [101, 105]}, {"head": "NIST", "head_type": "NA", "head_pos": [101, 105], "relation": "hyponym of", "tail": "MT evaluation methods", "tail_type": "NA", "tail_pos": [60, 81]}, {"head": "NIST", "head_type": "NA", "head_pos": [95, 99], "relation": "conjunction", "tail": "WER", "tail_type": "NA", "tail_pos": [108, 111]}, {"head": "WER", "head_type": "NA", "head_pos": [108, 111], "relation": "hyponym of", "tail": "MT evaluation methods", "tail_type": "NA", "tail_pos": [60, 81]}, {"head": "WER", "head_type": "NA", "head_pos": [102, 105], "relation": "conjunction", "tail": "PER", "tail_type": "NA", "tail_pos": [116, 119]}, {"head": "PER", "head_type": "NA", "head_pos": [116, 119], "relation": "hyponym of", "tail": "MT evaluation methods", "tail_type": "NA", "tail_pos": [60, 81]}, {"head": "classifiers", "head_type": "NA", "head_pos": [132, 143], "relation": "used for", "tail": "semantic equivalence", "tail_type": "NA", "tail_pos": [161, 181]}, {"head": "classifiers", "head_type": "NA", "head_pos": [132, 143], "relation": "used for", "tail": "entailment", "tail_type": "NA", "tail_pos": [186, 196]}, {"head": "semantic equivalence", "head_type": "NA", "head_pos": [155, 175], "relation": "conjunction", "tail": "entailment", "tail_type": "NA", "tail_pos": [186, 196]}], "task": "RE"}
+{"text": "We also introduce a novel classification method based on PER which leverages part of speech information of the words contributing to the word matches and non-matches in the sentence .", "relation": [{"head": "PER", "head_type": "NA", "head_pos": [66, 69], "relation": "used for", "tail": "classification method", "tail_type": "NA", "tail_pos": [29, 50]}, {"head": "PER", "head_type": "NA", "head_pos": [60, 63], "relation": "used for", "tail": "part of speech information", "tail_type": "NA", "tail_pos": [86, 112]}, {"head": "part of speech information", "head_type": "NA", "head_pos": [80, 106], "relation": "used for", "tail": "word matches and non-matches", "tail_type": "NA", "tail_pos": [146, 174]}], "task": "RE"}
+{"text": "Our results show that MT evaluation techniques are able to produce useful features for paraphrase classification and to a lesser extent entailment .", "relation": [{"head": "MT evaluation techniques", "head_type": "NA", "head_pos": [25, 49], "relation": "used for", "tail": "features", "tail_type": "NA", "tail_pos": [83, 91]}, {"head": "MT evaluation techniques", "head_type": "NA", "head_pos": [25, 49], "relation": "used for", "tail": "paraphrase classification", "tail_type": "NA", "tail_pos": [96, 121]}, {"head": "MT evaluation techniques", "head_type": "NA", "head_pos": [25, 49], "relation": "used for", "tail": "entailment", "tail_type": "NA", "tail_pos": [145, 155]}, {"head": "paraphrase classification", "head_type": "NA", "head_pos": [90, 115], "relation": "conjunction", "tail": "entailment", "tail_type": "NA", "tail_pos": [145, 155]}], "task": "RE"}
+{"text": "Our technique gives a substantial improvement in paraphrase classification accuracy over all of the other models used in the experiments .", "relation": [{"head": "technique", "head_type": "NA", "head_pos": [7, 16], "relation": "compare", "tail": "models", "tail_type": "NA", "tail_pos": [115, 121]}, {"head": "paraphrase classification accuracy", "head_type": "NA", "head_pos": [58, 92], "relation": "evaluate for", "tail": "technique", "tail_type": "NA", "tail_pos": [7, 16]}, {"head": "paraphrase classification accuracy", "head_type": "NA", "head_pos": [52, 86], "relation": "evaluate for", "tail": "models", "tail_type": "NA", "tail_pos": [115, 121]}], "task": "RE"}
+{"text": "Given an object model and a black-box measure of similarity between the model and candidate targets , we consider visual object tracking as a numerical optimization problem .", "relation": [{"head": "numerical optimization problem", "head_type": "NA", "head_pos": [151, 181], "relation": "used for", "tail": "visual object tracking", "tail_type": "NA", "tail_pos": [117, 139]}], "task": "RE"}
+{"text": "During normal tracking conditions when the object is visible from frame to frame , local optimization is used to track the local mode of the similarity measure in a parameter space of translation , rotation and scale .", "relation": [{"head": "local optimization", "head_type": "NA", "head_pos": [86, 104], "relation": "used for", "tail": "local mode of the similarity measure", "tail_type": "NA", "tail_pos": [132, 168]}, {"head": "parameter space of translation , rotation and scale", "head_type": "NA", "head_pos": [174, 225], "relation": "used for", "tail": "local mode of the similarity measure", "tail_type": "NA", "tail_pos": [126, 162]}], "task": "RE"}
+{"text": "However , when the object becomes partially or totally occluded , such local tracking is prone to failure , especially when common prediction techniques like the Kalman filter do not provide a good estimate of object parameters in future frames .", "relation": [{"head": "Kalman filter", "head_type": "NA", "head_pos": [171, 184], "relation": "part of", "tail": "prediction techniques", "tail_type": "NA", "tail_pos": [134, 155]}], "task": "RE"}
+{"text": "To recover from these inevitable tracking failures , we consider object detection as a global optimization problem and solve it via Adaptive Simulated Annealing -LRB- ASA -RRB- , a method that avoids becoming trapped at local modes and is much faster than exhaustive search .", "relation": [{"head": "global optimization problem", "head_type": "NA", "head_pos": [96, 123], "relation": "used for", "tail": "object detection", "tail_type": "NA", "tail_pos": [68, 84]}, {"head": "Adaptive Simulated Annealing -LRB- ASA -RRB-", "head_type": "NA", "head_pos": [141, 185], "relation": "used for", "tail": "it", "tail_type": "NA", "tail_pos": [26, 28]}, {"head": "method", "head_type": "NA", "head_pos": [184, 190], "relation": "compare", "tail": "exhaustive search", "tail_type": "NA", "tail_pos": [265, 282]}], "task": "RE"}
+{"text": "As a Monte Carlo approach , ASA stochastically samples the parameter space , in contrast to local deterministic search .", "relation": [{"head": "ASA", "head_type": "NA", "head_pos": [37, 40], "relation": "hyponym of", "tail": "Monte Carlo approach", "tail_type": "NA", "tail_pos": [8, 28]}, {"head": "ASA", "head_type": "NA", "head_pos": [31, 34], "relation": "compare", "tail": "local deterministic search", "tail_type": "NA", "tail_pos": [101, 127]}], "task": "RE"}
+{"text": "We apply cluster analysis on the sampled parameter space to redetect the object and renew the local tracker .", "relation": [{"head": "cluster analysis", "head_type": "NA", "head_pos": [12, 28], "relation": "used for", "tail": "sampled parameter space", "tail_type": "NA", "tail_pos": [42, 65]}, {"head": "cluster analysis", "head_type": "NA", "head_pos": [12, 28], "relation": "used for", "tail": "local tracker", "tail_type": "NA", "tail_pos": [103, 116]}], "task": "RE"}
+{"text": "Our numerical hybrid local and global mode-seeking tracker is validated on challenging airborne videos with heavy occlusion and large camera motions .", "relation": [{"head": "airborne videos", "head_type": "NA", "head_pos": [96, 111], "relation": "evaluate for", "tail": "numerical hybrid local and global mode-seeking tracker", "tail_type": "NA", "tail_pos": [7, 61]}, {"head": "heavy occlusion", "head_type": "NA", "head_pos": [117, 132], "relation": "feature of", "tail": "airborne videos", "tail_type": "NA", "tail_pos": [90, 105]}, {"head": "heavy occlusion", "head_type": "NA", "head_pos": [111, 126], "relation": "conjunction", "tail": "camera motions", "tail_type": "NA", "tail_pos": [143, 157]}, {"head": "camera motions", "head_type": "NA", "head_pos": [143, 157], "relation": "feature of", "tail": "airborne videos", "tail_type": "NA", "tail_pos": [90, 105]}], "task": "RE"}
+{"text": "Our approach outperforms state-of-the-art trackers on the VIVID benchmark datasets .", "relation": [{"head": "state-of-the-art trackers", "head_type": "NA", "head_pos": [34, 59], "relation": "compare", "tail": "approach", "tail_type": "NA", "tail_pos": [7, 15]}, {"head": "VIVID benchmark datasets", "head_type": "NA", "head_pos": [67, 91], "relation": "evaluate for", "tail": "approach", "tail_type": "NA", "tail_pos": [7, 15]}, {"head": "VIVID benchmark datasets", "head_type": "NA", "head_pos": [67, 91], "relation": "evaluate for", "tail": "state-of-the-art trackers", "tail_type": "NA", "tail_pos": [28, 53]}], "task": "RE"}
+{"text": "Techniques for automatically training modules of a natural language generator have recently been proposed , but a fundamental concern is whether the quality of utterances produced with trainable components can compete with hand-crafted template-based or rule-based approaches .", "relation": [{"head": "Techniques", "head_type": "NA", "head_pos": [3, 13], "relation": "used for", "tail": "automatically training modules", "tail_type": "NA", "tail_pos": [24, 54]}, {"head": "automatically training modules", "head_type": "NA", "head_pos": [18, 48], "relation": "part of", "tail": "natural language generator", "tail_type": "NA", "tail_pos": [60, 86]}, {"head": "utterances", "head_type": "NA", "head_pos": [163, 173], "relation": "evaluate for", "tail": "trainable components", "tail_type": "NA", "tail_pos": [194, 214]}, {"head": "utterances", "head_type": "NA", "head_pos": [163, 173], "relation": "evaluate for", "tail": "hand-crafted template-based or rule-based approaches", "tail_type": "NA", "tail_pos": [232, 284]}, {"head": "trainable components", "head_type": "NA", "head_pos": [188, 208], "relation": "compare", "tail": "hand-crafted template-based or rule-based approaches", "tail_type": "NA", "tail_pos": [232, 284]}], "task": "RE"}
+{"text": "In this paper We experimentally evaluate a trainable sentence planner for a spoken dialogue system by eliciting subjective human judgments .", "relation": [{"head": "trainable sentence planner", "head_type": "NA", "head_pos": [46, 72], "relation": "used for", "tail": "spoken dialogue system", "tail_type": "NA", "tail_pos": [85, 107]}, {"head": "subjective human judgments", "head_type": "NA", "head_pos": [121, 147], "relation": "evaluate for", "tail": "trainable sentence planner", "tail_type": "NA", "tail_pos": [46, 72]}], "task": "RE"}
+{"text": "In order to perform an exhaustive comparison , we also evaluate a hand-crafted template-based generation component , two rule-based sentence planners , and two baseline sentence planners .", "relation": [{"head": "hand-crafted template-based generation component", "head_type": "NA", "head_pos": [69, 117], "relation": "conjunction", "tail": "rule-based sentence planners", "tail_type": "NA", "tail_pos": [130, 158]}, {"head": "rule-based sentence planners", "head_type": "NA", "head_pos": [124, 152], "relation": "conjunction", "tail": "baseline sentence planners", "tail_type": "NA", "tail_pos": [169, 195]}], "task": "RE"}
+{"text": "We show that the trainable sentence planner performs better than the rule-based systems and the baselines , and as well as the hand-crafted system .", "relation": [{"head": "trainable sentence planner", "head_type": "NA", "head_pos": [20, 46], "relation": "compare", "tail": "rule-based systems", "tail_type": "NA", "tail_pos": [78, 96]}, {"head": "trainable sentence planner", "head_type": "NA", "head_pos": [20, 46], "relation": "compare", "tail": "baselines", "tail_type": "NA", "tail_pos": [105, 114]}, {"head": "trainable sentence planner", "head_type": "NA", "head_pos": [20, 46], "relation": "compare", "tail": "hand-crafted system", "tail_type": "NA", "tail_pos": [136, 155]}, {"head": "rule-based systems", "head_type": "NA", "head_pos": [72, 90], "relation": "conjunction", "tail": "baselines", "tail_type": "NA", "tail_pos": [105, 114]}, {"head": "baselines", "head_type": "NA", "head_pos": [99, 108], "relation": "conjunction", "tail": "hand-crafted system", "tail_type": "NA", "tail_pos": [136, 155]}], "task": "RE"}
+{"text": "A new algorithm is proposed for novel view generation in one-to-one teleconferencing applications .", "relation": [{"head": "algorithm", "head_type": "NA", "head_pos": [9, 18], "relation": "used for", "tail": "novel view generation", "tail_type": "NA", "tail_pos": [41, 62]}, {"head": "novel view generation", "head_type": "NA", "head_pos": [35, 56], "relation": "used for", "tail": "one-to-one teleconferencing applications", "tail_type": "NA", "tail_pos": [66, 106]}], "task": "RE"}
+{"text": "Given the video streams acquired by two cameras placed on either side of a computer monitor , the proposed algorithm synthesises images from a virtual camera in arbitrary position -LRB- typically located within the monitor -RRB- to facilitate eye contact .", "relation": [{"head": "cameras", "head_type": "NA", "head_pos": [49, 56], "relation": "used for", "tail": "video streams", "tail_type": "NA", "tail_pos": [13, 26]}, {"head": "algorithm", "head_type": "NA", "head_pos": [110, 119], "relation": "used for", "tail": "eye contact", "tail_type": "NA", "tail_pos": [252, 263]}, {"head": "virtual camera", "head_type": "NA", "head_pos": [152, 166], "relation": "used for", "tail": "images", "tail_type": "NA", "tail_pos": [132, 138]}, {"head": "arbitrary position", "head_type": "NA", "head_pos": [170, 188], "relation": "feature of", "tail": "virtual camera", "tail_type": "NA", "tail_pos": [146, 160]}], "task": "RE"}
+{"text": "Our technique is based on an improved , dynamic-programming , stereo algorithm for efficient novel-view generation .", "relation": [{"head": "technique", "head_type": "NA", "head_pos": [7, 16], "relation": "used for", "tail": "novel-view generation", "tail_type": "NA", "tail_pos": [102, 123]}, {"head": "dynamic-programming , stereo algorithm", "head_type": "NA", "head_pos": [49, 87], "relation": "used for", "tail": "technique", "tail_type": "NA", "tail_pos": [7, 16]}], "task": "RE"}
+{"text": "The two main contributions of this paper are : i -RRB- a new type of three-plane graph for dense-stereo dynamic-programming , that encourages correct occlusion labeling ; ii -RRB- a compact geometric derivation for novel-view synthesis by direct projection of the minimum-cost surface .", "relation": [{"head": "three-plane graph", "head_type": "NA", "head_pos": [72, 89], "relation": "used for", "tail": "dense-stereo dynamic-programming", "tail_type": "NA", "tail_pos": [100, 132]}, {"head": "dense-stereo dynamic-programming", "head_type": "NA", "head_pos": [94, 126], "relation": "used for", "tail": "occlusion labeling", "tail_type": "NA", "tail_pos": [159, 177]}, {"head": "compact geometric derivation", "head_type": "NA", "head_pos": [185, 213], "relation": "used for", "tail": "novel-view synthesis", "tail_type": "NA", "tail_pos": [224, 244]}, {"head": "direct projection of the minimum-cost surface", "head_type": "NA", "head_pos": [248, 293], "relation": "used for", "tail": "compact geometric derivation", "tail_type": "NA", "tail_pos": [185, 213]}], "task": "RE"}
+{"text": "Furthermore , this paper presents a novel algorithm for the temporal maintenance of a background model to enhance the rendering of occlusions and reduce temporal artefacts -LRB- flicker -RRB- ; and a cost aggregation algorithm that acts directly on our three-dimensional matching cost space .", "relation": [{"head": "algorithm", "head_type": "NA", "head_pos": [45, 54], "relation": "used for", "tail": "temporal maintenance of a background model", "tail_type": "NA", "tail_pos": [69, 111]}, {"head": "algorithm", "head_type": "NA", "head_pos": [45, 54], "relation": "used for", "tail": "rendering of occlusions", "tail_type": "NA", "tail_pos": [127, 150]}, {"head": "algorithm", "head_type": "NA", "head_pos": [45, 54], "relation": "used for", "tail": "temporal artefacts -LRB- flicker -RRB-", "tail_type": "NA", "tail_pos": [162, 200]}, {"head": "cost aggregation algorithm", "head_type": "NA", "head_pos": [209, 235], "relation": "conjunction", "tail": "algorithm", "tail_type": "NA", "tail_pos": [45, 54]}, {"head": "cost aggregation algorithm", "head_type": "NA", "head_pos": [203, 229], "relation": "used for", "tail": "three-dimensional matching cost space", "tail_type": "NA", "tail_pos": [262, 299]}], "task": "RE"}
+{"text": "Examples are given that demonstrate the robustness of the new algorithm to spatial and temporal artefacts for long stereo video streams .", "relation": [{"head": "robustness", "head_type": "NA", "head_pos": [43, 53], "relation": "evaluate for", "tail": "algorithm", "tail_type": "NA", "tail_pos": [71, 80]}, {"head": "algorithm", "head_type": "NA", "head_pos": [65, 74], "relation": "used for", "tail": "spatial and temporal artefacts", "tail_type": "NA", "tail_pos": [84, 114]}, {"head": "spatial and temporal artefacts", "head_type": "NA", "head_pos": [78, 108], "relation": "used for", "tail": "long stereo video streams", "tail_type": "NA", "tail_pos": [119, 144]}], "task": "RE"}
+{"text": "We further demonstrate synthesis from a freely translating virtual camera .", "relation": [{"head": "translating virtual camera", "head_type": "NA", "head_pos": [56, 82], "relation": "used for", "tail": "synthesis", "tail_type": "NA", "tail_pos": [26, 35]}], "task": "RE"}
+{"text": "To a large extent , these statistics reflect semantic constraints and thus are used to disambiguate anaphora references and syntactic ambiguities .", "relation": [{"head": "semantic constraints", "head_type": "NA", "head_pos": [48, 68], "relation": "used for", "tail": "anaphora references", "tail_type": "NA", "tail_pos": [109, 128]}, {"head": "semantic constraints", "head_type": "NA", "head_pos": [48, 68], "relation": "used for", "tail": "syntactic ambiguities", "tail_type": "NA", "tail_pos": [133, 154]}, {"head": "anaphora references", "head_type": "NA", "head_pos": [103, 122], "relation": "conjunction", "tail": "syntactic ambiguities", "tail_type": "NA", "tail_pos": [133, 154]}], "task": "RE"}
+{"text": "The results of the experiment show that in most of the cases the cooccurrence statistics indeed reflect the semantic constraints and thus provide a basis for a useful disambiguation tool .", "relation": [{"head": "cooccurrence statistics", "head_type": "NA", "head_pos": [68, 91], "relation": "used for", "tail": "disambiguation tool", "tail_type": "NA", "tail_pos": [176, 195]}], "task": "RE"}
+{"text": "We present a novel method for discovering parallel sentences in comparable , non-parallel corpora .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [22, 28], "relation": "used for", "tail": "discovering parallel sentences", "tail_type": "NA", "tail_pos": [39, 69]}, {"head": "comparable , non-parallel corpora", "head_type": "NA", "head_pos": [73, 106], "relation": "used for", "tail": "discovering parallel sentences", "tail_type": "NA", "tail_pos": [33, 63]}], "task": "RE"}
+{"text": "Using this approach , we extract parallel data from large Chinese , Arabic , and English non-parallel newspaper corpora .", "relation": [{"head": "approach", "head_type": "NA", "head_pos": [14, 22], "relation": "used for", "tail": "parallel data", "tail_type": "NA", "tail_pos": [42, 55]}, {"head": "parallel data", "head_type": "NA", "head_pos": [36, 49], "relation": "part of", "tail": "Chinese , Arabic , and English non-parallel newspaper corpora", "tail_type": "NA", "tail_pos": [67, 128]}], "task": "RE"}
+{"text": "We evaluate the quality of the extracted data by showing that it improves the performance of a state-of-the-art statistical machine translation system .", "relation": [{"head": "it", "head_type": "NA", "head_pos": [20, 22], "relation": "used for", "tail": "statistical machine translation system", "tail_type": "NA", "tail_pos": [121, 159]}], "task": "RE"}
+{"text": "We also show that a good-quality MT system can be built from scratch by starting with a very small parallel corpus -LRB- 100,000 words -RRB- and exploiting a large non-parallel corpus .", "relation": [{"head": "parallel corpus", "head_type": "NA", "head_pos": [108, 123], "relation": "used for", "tail": "MT system", "tail_type": "NA", "tail_pos": [36, 45]}, {"head": "parallel corpus", "head_type": "NA", "head_pos": [102, 117], "relation": "conjunction", "tail": "non-parallel corpus", "tail_type": "NA", "tail_pos": [173, 192]}, {"head": "non-parallel corpus", "head_type": "NA", "head_pos": [173, 192], "relation": "used for", "tail": "MT system", "tail_type": "NA", "tail_pos": [36, 45]}], "task": "RE"}
+{"text": "Thus , our method can be applied with great benefit to language pairs for which only scarce resources are available .", "relation": [{"head": "scarce resources", "head_type": "NA", "head_pos": [94, 110], "relation": "used for", "tail": "method", "tail_type": "NA", "tail_pos": [14, 20]}], "task": "RE"}
+{"text": "In this paper , we describe a search procedure for statistical machine translation -LRB- MT -RRB- based on dynamic programming -LRB- DP -RRB- .", "relation": [{"head": "search procedure", "head_type": "NA", "head_pos": [33, 49], "relation": "used for", "tail": "statistical machine translation -LRB- MT -RRB-", "tail_type": "NA", "tail_pos": [60, 106]}, {"head": "dynamic programming -LRB- DP -RRB-", "head_type": "NA", "head_pos": [116, 150], "relation": "used for", "tail": "statistical machine translation -LRB- MT -RRB-", "tail_type": "NA", "tail_pos": [54, 100]}], "task": "RE"}
+{"text": "Starting from a DP-based solution to the traveling salesman problem , we present a novel technique to restrict the possible word reordering between source and target language in order to achieve an efficient search algorithm .", "relation": [{"head": "technique", "head_type": "NA", "head_pos": [92, 101], "relation": "used for", "tail": "search algorithm", "tail_type": "NA", "tail_pos": [217, 233]}], "task": "RE"}
+{"text": "The experimental tests are carried out on the Verbmobil task -LRB- German-English , 8000-word vocabulary -RRB- , which is a limited-domain spoken-language task .", "relation": [{"head": "Verbmobil task", "head_type": "NA", "head_pos": [49, 63], "relation": "hyponym of", "tail": "limited-domain spoken-language task", "tail_type": "NA", "tail_pos": [133, 168]}], "task": "RE"}
+{"text": "A purely functional implementation of LR-parsers is given , together with a simple correctness proof .", "relation": [{"head": "correctness proof", "head_type": "NA", "head_pos": [92, 109], "relation": "conjunction", "tail": "LR-parsers", "tail_type": "NA", "tail_pos": [41, 51]}], "task": "RE"}
+{"text": "It is presented as a generalization of the recursive descent parser .", "relation": [{"head": "recursive descent parser", "head_type": "NA", "head_pos": [52, 76], "relation": "used for", "tail": "It", "tail_type": "NA", "tail_pos": [3, 5]}], "task": "RE"}
+{"text": "For non-LR grammars the time-complexity of our parser is cubic if the functions that constitute the parser are implemented as memo-functions , i.e. functions that memorize the results of previous invocations .", "relation": [{"head": "time-complexity", "head_type": "NA", "head_pos": [27, 42], "relation": "evaluate for", "tail": "parser", "tail_type": "NA", "tail_pos": [56, 62]}, {"head": "parser", "head_type": "NA", "head_pos": [56, 62], "relation": "used for", "tail": "non-LR grammars", "tail_type": "NA", "tail_pos": [7, 22]}, {"head": "memo-functions", "head_type": "NA", "head_pos": [135, 149], "relation": "used for", "tail": "parser", "tail_type": "NA", "tail_pos": [47, 53]}], "task": "RE"}
+{"text": "Memo-functions also facilitate a simple way to construct a very compact representation of the parse forest .", "relation": [{"head": "Memo-functions", "head_type": "NA", "head_pos": [3, 17], "relation": "used for", "tail": "parse forest", "tail_type": "NA", "tail_pos": [103, 115]}], "task": "RE"}
+{"text": "For LR -LRB- 0 -RRB- grammars , our algorithm is closely related to the recursive ascent parsers recently discovered by Kruse-man Aretz -LSB- 1 -RSB- and Roberts -LSB- 2 -RSB- .", "relation": [{"head": "algorithm", "head_type": "NA", "head_pos": [45, 54], "relation": "used for", "tail": "LR -LRB- 0 -RRB- grammars", "tail_type": "NA", "tail_pos": [7, 32]}, {"head": "algorithm", "head_type": "NA", "head_pos": [39, 48], "relation": "conjunction", "tail": "recursive ascent parsers", "tail_type": "NA", "tail_pos": [81, 105]}], "task": "RE"}
+{"text": "Extended CF grammars -LRB- grammars with regular expressions at the right hand side -RRB- can be parsed with a simple modification of the LR-parser for normal CF grammars .", "relation": [{"head": "regular expressions", "head_type": "NA", "head_pos": [50, 69], "relation": "feature of", "tail": "grammars", "tail_type": "NA", "tail_pos": [12, 20]}, {"head": "LR-parser", "head_type": "NA", "head_pos": [147, 156], "relation": "used for", "tail": "Extended CF grammars", "tail_type": "NA", "tail_pos": [3, 23]}, {"head": "LR-parser", "head_type": "NA", "head_pos": [141, 150], "relation": "used for", "tail": "CF grammars", "tail_type": "NA", "tail_pos": [9, 20]}], "task": "RE"}
+{"text": "In this theory , discourse structure is composed of three separate but interrelated components : the structure of the sequence of utterances -LRB- called the linguistic structure -RRB- , a structure of purposes -LRB- called the intentional structure -RRB- , and the state of focus of attention -LRB- called the attentional state -RRB- .", "relation": [{"head": "components", "head_type": "NA", "head_pos": [93, 103], "relation": "part of", "tail": "discourse structure", "tail_type": "NA", "tail_pos": [20, 39]}, {"head": "linguistic structure", "head_type": "NA", "head_pos": [167, 187], "relation": "part of", "tail": "components", "tail_type": "NA", "tail_pos": [87, 97]}, {"head": "linguistic structure", "head_type": "NA", "head_pos": [161, 181], "relation": "conjunction", "tail": "intentional structure", "tail_type": "NA", "tail_pos": [237, 258]}, {"head": "intentional structure", "head_type": "NA", "head_pos": [237, 258], "relation": "part of", "tail": "components", "tail_type": "NA", "tail_pos": [87, 97]}, {"head": "intentional structure", "head_type": "NA", "head_pos": [231, 252], "relation": "conjunction", "tail": "attentional state", "tail_type": "NA", "tail_pos": [320, 337]}, {"head": "attentional state", "head_type": "NA", "head_pos": [320, 337], "relation": "part of", "tail": "components", "tail_type": "NA", "tail_pos": [87, 97]}], "task": "RE"}
+{"text": "The intentional structure captures the discourse-relevant purposes , expressed in each of the linguistic segments as well as relationships among them .", "relation": [{"head": "intentional structure", "head_type": "NA", "head_pos": [7, 28], "relation": "used for", "tail": "discourse-relevant purposes", "tail_type": "NA", "tail_pos": [48, 75]}], "task": "RE"}
+{"text": "The distinction among these components is essential to provide an adequate explanation of such discourse phenomena as cue phrases , referring expressions , and interruptions .", "relation": [{"head": "cue phrases", "head_type": "NA", "head_pos": [127, 138], "relation": "hyponym of", "tail": "discourse phenomena", "tail_type": "NA", "tail_pos": [98, 117]}, {"head": "cue phrases", "head_type": "NA", "head_pos": [121, 132], "relation": "conjunction", "tail": "referring expressions", "tail_type": "NA", "tail_pos": [141, 162]}, {"head": "referring expressions", "head_type": "NA", "head_pos": [141, 162], "relation": "hyponym of", "tail": "discourse phenomena", "tail_type": "NA", "tail_pos": [98, 117]}, {"head": "referring expressions", "head_type": "NA", "head_pos": [135, 156], "relation": "conjunction", "tail": "interruptions", "tail_type": "NA", "tail_pos": [169, 182]}, {"head": "interruptions", "head_type": "NA", "head_pos": [169, 182], "relation": "hyponym of", "tail": "discourse phenomena", "tail_type": "NA", "tail_pos": [98, 117]}], "task": "RE"}
+{"text": "We examine the relationship between the two grammatical formalisms : Tree Adjoining Grammars and Head Grammars .", "relation": [{"head": "Tree Adjoining Grammars", "head_type": "NA", "head_pos": [78, 101], "relation": "hyponym of", "tail": "grammatical formalisms", "tail_type": "NA", "tail_pos": [47, 69]}, {"head": "Tree Adjoining Grammars", "head_type": "NA", "head_pos": [72, 95], "relation": "compare", "tail": "Head Grammars", "tail_type": "NA", "tail_pos": [106, 119]}, {"head": "Head Grammars", "head_type": "NA", "head_pos": [106, 119], "relation": "hyponym of", "tail": "grammatical formalisms", "tail_type": "NA", "tail_pos": [47, 69]}], "task": "RE"}
+{"text": "We then turn to a discussion comparing the linguistic expressiveness of the two formalisms .", "relation": [{"head": "linguistic expressiveness", "head_type": "NA", "head_pos": [46, 71], "relation": "feature of", "tail": "formalisms", "tail_type": "NA", "tail_pos": [89, 99]}], "task": "RE"}
+{"text": "We provide a unified account of sentence-level and text-level anaphora within the framework of a dependency-based grammar model .", "relation": [{"head": "dependency-based grammar model", "head_type": "NA", "head_pos": [106, 136], "relation": "used for", "tail": "sentence-level and text-level anaphora", "tail_type": "NA", "tail_pos": [35, 73]}], "task": "RE"}
+{"text": "Criteria for anaphora resolution within sentence boundaries rephrase major concepts from GB 's binding theory , while those for text-level anaphora incorporate an adapted version of a Grosz-Sidner-style focus model .", "relation": [{"head": "Criteria", "head_type": "NA", "head_pos": [3, 11], "relation": "used for", "tail": "anaphora resolution within sentence boundaries", "tail_type": "NA", "tail_pos": [22, 68]}, {"head": "GB 's binding theory", "head_type": "NA", "head_pos": [98, 118], "relation": "used for", "tail": "Criteria", "tail_type": "NA", "tail_pos": [3, 11]}, {"head": "those", "head_type": "NA", "head_pos": [121, 126], "relation": "used for", "tail": "text-level anaphora", "tail_type": "NA", "tail_pos": [137, 156]}, {"head": "Grosz-Sidner-style focus model", "head_type": "NA", "head_pos": [193, 223], "relation": "part of", "tail": "those", "tail_type": "NA", "tail_pos": [121, 126]}], "task": "RE"}
+{"text": "Coedition of a natural language text and its representation in some interlingual form seems the best and simplest way to share text revision across languages .", "relation": [{"head": "Coedition", "head_type": "NA", "head_pos": [3, 12], "relation": "used for", "tail": "text revision", "tail_type": "NA", "tail_pos": [136, 149]}, {"head": "natural language text", "head_type": "NA", "head_pos": [24, 45], "relation": "used for", "tail": "Coedition", "tail_type": "NA", "tail_pos": [3, 12]}], "task": "RE"}
+{"text": "The modified graph is then sent to the UNL-L0 deconverter and the result shown .", "relation": [{"head": "graph", "head_type": "NA", "head_pos": [16, 21], "relation": "used for", "tail": "UNL-L0 deconverter", "tail_type": "NA", "tail_pos": [48, 66]}], "task": "RE"}
+{"text": "On the internal side , liaisons are established between elements of the text and the graph by using broadly available resources such as a LO-English or better a L0-UNL dictionary , a morphosyntactic parser of L0 , and a canonical graph2tree transformation .", "relation": [{"head": "resources", "head_type": "NA", "head_pos": [127, 136], "relation": "used for", "tail": "liaisons", "tail_type": "NA", "tail_pos": [26, 34]}, {"head": "LO-English or better a L0-UNL dictionary", "head_type": "NA", "head_pos": [147, 187], "relation": "hyponym of", "tail": "resources", "tail_type": "NA", "tail_pos": [121, 130]}, {"head": "LO-English or better a L0-UNL dictionary", "head_type": "NA", "head_pos": [141, 181], "relation": "conjunction", "tail": "morphosyntactic parser of L0", "tail_type": "NA", "tail_pos": [192, 220]}, {"head": "morphosyntactic parser of L0", "head_type": "NA", "head_pos": [192, 220], "relation": "hyponym of", "tail": "resources", "tail_type": "NA", "tail_pos": [121, 130]}, {"head": "morphosyntactic parser of L0", "head_type": "NA", "head_pos": [186, 214], "relation": "conjunction", "tail": "canonical graph2tree transformation", "tail_type": "NA", "tail_pos": [229, 264]}, {"head": "canonical graph2tree transformation", "head_type": "NA", "head_pos": [229, 264], "relation": "hyponym of", "tail": "resources", "tail_type": "NA", "tail_pos": [121, 130]}], "task": "RE"}
+{"text": "Establishing a `` best '' correspondence between the '' UNL-tree + L0 '' and the '' MS-L0 structure '' , a lattice , may be done using the dictionary and trying to align the tree and the selected trajectory with as few crossing liaisons as possible .", "relation": [{"head": "UNL-tree + L0", "head_type": "NA", "head_pos": [59, 72], "relation": "conjunction", "tail": "MS-L0 structure", "tail_type": "NA", "tail_pos": [93, 108]}, {"head": "dictionary", "head_type": "NA", "head_pos": [148, 158], "relation": "used for", "tail": "lattice", "tail_type": "NA", "tail_pos": [110, 117]}], "task": "RE"}
+{"text": "A central goal of this research is to merge approaches from pivot MT , interactive MT , and multilingual text authoring .", "relation": [{"head": "pivot MT", "head_type": "NA", "head_pos": [63, 71], "relation": "conjunction", "tail": "interactive MT", "tail_type": "NA", "tail_pos": [80, 94]}, {"head": "interactive MT", "head_type": "NA", "head_pos": [74, 88], "relation": "conjunction", "tail": "multilingual text authoring", "tail_type": "NA", "tail_pos": [101, 128]}], "task": "RE"}
+{"text": "We report experiments conducted on a multilingual corpus to estimate the number of analogies among the sentences that it contains .", "relation": [{"head": "multilingual corpus", "head_type": "NA", "head_pos": [40, 59], "relation": "evaluate for", "tail": "analogies", "tail_type": "NA", "tail_pos": [92, 101]}], "task": "RE"}
+{"text": "Our goal is to learn a Mahalanobis distance by minimizing a loss defined on the weighted sum of the precision at different ranks .", "relation": [{"head": "loss", "head_type": "NA", "head_pos": [69, 73], "relation": "used for", "tail": "Mahalanobis distance", "tail_type": "NA", "tail_pos": [26, 46]}, {"head": "weighted sum", "head_type": "NA", "head_pos": [83, 95], "relation": "feature of", "tail": "precision", "tail_type": "NA", "tail_pos": [109, 118]}], "task": "RE"}
+{"text": "Our core motivation is that minimizing a weighted rank loss is a natural criterion for many problems in computer vision such as person re-identification .", "relation": [{"head": "weighted rank loss", "head_type": "NA", "head_pos": [44, 62], "relation": "used for", "tail": "computer vision", "tail_type": "NA", "tail_pos": [113, 128]}, {"head": "weighted rank loss", "head_type": "NA", "head_pos": [44, 62], "relation": "used for", "tail": "person re-identification", "tail_type": "NA", "tail_pos": [137, 161]}, {"head": "person re-identification", "head_type": "NA", "head_pos": [137, 161], "relation": "hyponym of", "tail": "computer vision", "tail_type": "NA", "tail_pos": [107, 122]}], "task": "RE"}
+{"text": "We propose a novel metric learning formulation called Weighted Approximate Rank Component Analysis -LRB- WARCA -RRB- .", "relation": [{"head": "Weighted Approximate Rank Component Analysis -LRB- WARCA -RRB-", "head_type": "NA", "head_pos": [63, 125], "relation": "hyponym of", "tail": "metric learning formulation", "tail_type": "NA", "tail_pos": [22, 49]}], "task": "RE"}
+{"text": "We then derive a scalable stochastic gradient descent algorithm for the resulting learning problem .", "relation": [{"head": "stochastic gradient descent algorithm", "head_type": "NA", "head_pos": [29, 66], "relation": "used for", "tail": "learning problem", "tail_type": "NA", "tail_pos": [91, 107]}], "task": "RE"}
+{"text": "We also derive an efficient non-linear extension of WARCA by using the kernel trick .", "relation": [{"head": "kernel trick", "head_type": "NA", "head_pos": [80, 92], "relation": "used for", "tail": "non-linear extension of WARCA", "tail_type": "NA", "tail_pos": [31, 60]}], "task": "RE"}
+{"text": "Kernel space embedding decouples the training and prediction costs from the data dimension and enables us to plug inarbitrary distance measures which are more natural for the features .", "relation": [{"head": "Kernel space embedding", "head_type": "NA", "head_pos": [3, 25], "relation": "used for", "tail": "inarbitrary distance measures", "tail_type": "NA", "tail_pos": [123, 152]}], "task": "RE"}
+{"text": "We also address a more general problem of matrix rank degeneration & non-isolated minima in the low-rank matrix optimization by using new type of regularizer which approximately enforces the or-thonormality of the learned matrix very efficiently .", "relation": [{"head": "matrix rank degeneration", "head_type": "NA", "head_pos": [45, 69], "relation": "conjunction", "tail": "non-isolated minima", "tail_type": "NA", "tail_pos": [78, 97]}, {"head": "matrix rank degeneration", "head_type": "NA", "head_pos": [45, 69], "relation": "feature of", "tail": "low-rank matrix optimization", "tail_type": "NA", "tail_pos": [105, 133]}, {"head": "non-isolated minima", "head_type": "NA", "head_pos": [72, 91], "relation": "feature of", "tail": "low-rank matrix optimization", "tail_type": "NA", "tail_pos": [105, 133]}, {"head": "regularizer", "head_type": "NA", "head_pos": [155, 166], "relation": "used for", "tail": "low-rank matrix optimization", "tail_type": "NA", "tail_pos": [99, 127]}, {"head": "regularizer", "head_type": "NA", "head_pos": [149, 160], "relation": "used for", "tail": "or-thonormality", "tail_type": "NA", "tail_pos": [200, 215]}, {"head": "or-thonormality", "head_type": "NA", "head_pos": [194, 209], "relation": "feature of", "tail": "learned matrix", "tail_type": "NA", "tail_pos": [223, 237]}], "task": "RE"}
+{"text": "We validate this new method on nine standard person re-identification datasets including two large scale Market-1501 and CUHK03 datasets and show that we improve upon the current state-of-the-art methods on all of them .", "relation": [{"head": "person re-identification datasets", "head_type": "NA", "head_pos": [54, 87], "relation": "evaluate for", "tail": "method", "tail_type": "NA", "tail_pos": [24, 30]}, {"head": "scale Market-1501", "head_type": "NA", "head_pos": [108, 125], "relation": "hyponym of", "tail": "person re-identification datasets", "tail_type": "NA", "tail_pos": [48, 81]}, {"head": "CUHK03 datasets", "head_type": "NA", "head_pos": [130, 145], "relation": "hyponym of", "tail": "person re-identification datasets", "tail_type": "NA", "tail_pos": [48, 81]}, {"head": "CUHK03 datasets", "head_type": "NA", "head_pos": [130, 145], "relation": "conjunction", "tail": "scale Market-1501", "tail_type": "NA", "tail_pos": [102, 119]}], "task": "RE"}
+{"text": "In this paper , we discuss language model adaptation methods given a word list and a raw corpus .", "relation": [{"head": "word list", "head_type": "NA", "head_pos": [78, 87], "relation": "used for", "tail": "language model adaptation methods", "tail_type": "NA", "tail_pos": [30, 63]}, {"head": "word list", "head_type": "NA", "head_pos": [72, 81], "relation": "conjunction", "tail": "raw corpus", "tail_type": "NA", "tail_pos": [94, 104]}, {"head": "raw corpus", "head_type": "NA", "head_pos": [94, 104], "relation": "used for", "tail": "language model adaptation methods", "tail_type": "NA", "tail_pos": [30, 63]}], "task": "RE"}
+{"text": "In this situation , the general method is to segment the raw corpus automatically using a word list , correct the output sentences by hand , and build a model from the segmented corpus .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [35, 41], "relation": "used for", "tail": "raw corpus", "tail_type": "NA", "tail_pos": [66, 76]}, {"head": "word list", "head_type": "NA", "head_pos": [99, 108], "relation": "used for", "tail": "method", "tail_type": "NA", "tail_pos": [35, 41]}, {"head": "segmented corpus", "head_type": "NA", "head_pos": [177, 193], "relation": "used for", "tail": "model", "tail_type": "NA", "tail_pos": [156, 161]}], "task": "RE"}
+{"text": "In the experiments , we used a variety of methods for preparing a segmented corpus and compared the language models by their speech recognition accuracies .", "relation": [{"head": "methods", "head_type": "NA", "head_pos": [45, 52], "relation": "used for", "tail": "preparing a segmented corpus", "tail_type": "NA", "tail_pos": [63, 91]}, {"head": "speech recognition accuracies", "head_type": "NA", "head_pos": [134, 163], "relation": "evaluate for", "tail": "language models", "tail_type": "NA", "tail_pos": [103, 118]}], "task": "RE"}
+{"text": "Many practical modeling problems involve discrete data that are best represented as draws from multinomial or categorical distributions .", "relation": [{"head": "discrete data", "head_type": "NA", "head_pos": [50, 63], "relation": "used for", "tail": "modeling problems", "tail_type": "NA", "tail_pos": [18, 35]}, {"head": "multinomial or categorical distributions", "head_type": "NA", "head_pos": [104, 144], "relation": "used for", "tail": "modeling problems", "tail_type": "NA", "tail_pos": [18, 35]}], "task": "RE"}
+{"text": "For example , nucleotides in a DNA sequence , children 's names in a given state and year , and text documents are all commonly modeled with multinomial distributions .", "relation": [{"head": "multinomial distributions", "head_type": "NA", "head_pos": [150, 175], "relation": "used for", "tail": "nucleotides in a DNA sequence", "tail_type": "NA", "tail_pos": [17, 46]}, {"head": "multinomial distributions", "head_type": "NA", "head_pos": [150, 175], "relation": "used for", "tail": "text documents", "tail_type": "NA", "tail_pos": [99, 113]}], "task": "RE"}
+{"text": "Here , we leverage a logistic stick-breaking representation and recent innovations in Pólya-gamma augmentation to reformu-late the multinomial distribution in terms of latent variables with jointly Gaussian likelihoods , enabling us to take advantage of a host of Bayesian inference techniques for Gaussian models with minimal overhead .", "relation": [{"head": "logistic stick-breaking representation", "head_type": "NA", "head_pos": [24, 62], "relation": "used for", "tail": "multinomial distribution", "tail_type": "NA", "tail_pos": [140, 164]}, {"head": "Pólya-gamma augmentation", "head_type": "NA", "head_pos": [89, 113], "relation": "used for", "tail": "multinomial distribution", "tail_type": "NA", "tail_pos": [140, 164]}, {"head": "latent variables", "head_type": "NA", "head_pos": [177, 193], "relation": "part of", "tail": "multinomial distribution", "tail_type": "NA", "tail_pos": [134, 158]}, {"head": "jointly Gaussian likelihoods", "head_type": "NA", "head_pos": [199, 227], "relation": "feature of", "tail": "latent variables", "tail_type": "NA", "tail_pos": [171, 187]}, {"head": "Bayesian inference techniques", "head_type": "NA", "head_pos": [267, 296], "relation": "used for", "tail": "Gaussian models", "tail_type": "NA", "tail_pos": [307, 322]}, {"head": "minimal overhead", "head_type": "NA", "head_pos": [328, 344], "relation": "feature of", "tail": "Gaussian models", "tail_type": "NA", "tail_pos": [301, 316]}], "task": "RE"}
+{"text": "MINPRAN , a new robust operator , nds good ts in data sets where more than 50 % of the points are outliers .", "relation": [{"head": "MINPRAN", "head_type": "NA", "head_pos": [3, 10], "relation": "hyponym of", "tail": "robust operator", "tail_type": "NA", "tail_pos": [25, 40]}], "task": "RE"}
+{"text": "Unlike other techniques that handle large outlier percentages , MINPRAN does not rely on a known error bound for the good data .", "relation": [{"head": "techniques", "head_type": "NA", "head_pos": [16, 26], "relation": "used for", "tail": "large outlier percentages", "tail_type": "NA", "tail_pos": [45, 70]}, {"head": "techniques", "head_type": "NA", "head_pos": [16, 26], "relation": "compare", "tail": "MINPRAN", "tail_type": "NA", "tail_pos": [73, 80]}], "task": "RE"}
+{"text": "Based on this , MINPRAN uses random sampling to search for the t and the number of inliers to the t that are least likely to have occurred randomly .", "relation": [{"head": "random sampling", "head_type": "NA", "head_pos": [38, 53], "relation": "used for", "tail": "MINPRAN", "tail_type": "NA", "tail_pos": [19, 26]}], "task": "RE"}
+{"text": "MINPRAN 's properties are connrmed experimentally on synthetic data and compare favorably to least median of squares .", "relation": [{"head": "synthetic data", "head_type": "NA", "head_pos": [62, 76], "relation": "evaluate for", "tail": "MINPRAN", "tail_type": "NA", "tail_pos": [3, 10]}, {"head": "least median of squares", "head_type": "NA", "head_pos": [102, 125], "relation": "compare", "tail": "MINPRAN", "tail_type": "NA", "tail_pos": [3, 10]}], "task": "RE"}
+{"text": "Related work applies MINPRAN to complex range and intensity data 23 -RSB- .", "relation": [{"head": "MINPRAN", "head_type": "NA", "head_pos": [24, 31], "relation": "used for", "tail": "complex range", "tail_type": "NA", "tail_pos": [41, 54]}, {"head": "MINPRAN", "head_type": "NA", "head_pos": [24, 31], "relation": "used for", "tail": "intensity data", "tail_type": "NA", "tail_pos": [59, 73]}], "task": "RE"}
+{"text": "Metagrammatical formalisms that combine context-free phrase structure rules and metarules -LRB- MPS grammars -RRB- allow concise statement of generalizations about the syntax of natural languages .", "relation": [{"head": "context-free phrase structure rules", "head_type": "NA", "head_pos": [49, 84], "relation": "part of", "tail": "Metagrammatical formalisms", "tail_type": "NA", "tail_pos": [3, 29]}, {"head": "context-free phrase structure rules", "head_type": "NA", "head_pos": [43, 78], "relation": "conjunction", "tail": "metarules -LRB- MPS grammars -RRB-", "tail_type": "NA", "tail_pos": [89, 123]}, {"head": "metarules -LRB- MPS grammars -RRB-", "head_type": "NA", "head_pos": [89, 123], "relation": "part of", "tail": "Metagrammatical formalisms", "tail_type": "NA", "tail_pos": [3, 29]}], "task": "RE"}
+{"text": "We evaluate several proposals for constraining them , basing our assessment on computational tractability and explanatory adequacy .", "relation": [{"head": "computational tractability and explanatory adequacy", "head_type": "NA", "head_pos": [88, 139], "relation": "evaluate for", "tail": "them", "tail_type": "NA", "tail_pos": [50, 54]}], "task": "RE"}
+{"text": "The unique properties of tree-adjoining grammars -LRB- TAG -RRB- present a challenge for the application of TAGs beyond the limited confines of syntax , for instance , to the task of semantic interpretation or automatic translation of natural language .", "relation": [{"head": "TAGs", "head_type": "NA", "head_pos": [111, 115], "relation": "used for", "tail": "semantic interpretation", "tail_type": "NA", "tail_pos": [192, 215]}, {"head": "TAGs", "head_type": "NA", "head_pos": [111, 115], "relation": "used for", "tail": "automatic translation of natural language", "tail_type": "NA", "tail_pos": [219, 260]}, {"head": "semantic interpretation", "head_type": "NA", "head_pos": [186, 209], "relation": "conjunction", "tail": "automatic translation of natural language", "tail_type": "NA", "tail_pos": [219, 260]}], "task": "RE"}
+{"text": "The formalism 's intended usage is to relate expressions of natural languages to their associated semantics represented in a logical form language , or to their translates in another natural language ; in summary , we intend it to allow TAGs to be used beyond their role in syntax proper .", "relation": [{"head": "logical form language", "head_type": "NA", "head_pos": [134, 155], "relation": "used for", "tail": "semantics", "tail_type": "NA", "tail_pos": [101, 110]}, {"head": "TAGs", "head_type": "NA", "head_pos": [240, 244], "relation": "used for", "tail": "syntax proper", "tail_type": "NA", "tail_pos": [283, 296]}], "task": "RE"}
+{"text": "A model-based approach to on-line cursive handwriting analysis and recognition is presented and evaluated .", "relation": [{"head": "model-based approach", "head_type": "NA", "head_pos": [5, 25], "relation": "used for", "tail": "on-line cursive handwriting analysis and recognition", "tail_type": "NA", "tail_pos": [35, 87]}], "task": "RE"}
+{"text": "In this model , on-line handwriting is considered as a modulation of a simple cycloidal pen motion , described by two coupled oscillations with a constant linear drift along the line of the writing .", "relation": [{"head": "model", "head_type": "NA", "head_pos": [11, 16], "relation": "used for", "tail": "on-line handwriting", "tail_type": "NA", "tail_pos": [25, 44]}, {"head": "on-line handwriting", "head_type": "NA", "head_pos": [19, 38], "relation": "part of", "tail": "cycloidal pen motion", "tail_type": "NA", "tail_pos": [87, 107]}], "task": "RE"}
+{"text": "A general procedure for the estimation and quantization of these cycloidal motion parameters for arbitrary handwriting is presented .", "relation": [{"head": "cycloidal motion parameters", "head_type": "NA", "head_pos": [68, 95], "relation": "used for", "tail": "arbitrary handwriting", "tail_type": "NA", "tail_pos": [106, 127]}], "task": "RE"}
+{"text": "The result is a discrete motor control representation of the continuous pen motion , via the quantized levels of the model parameters .", "relation": [{"head": "discrete motor control representation", "head_type": "NA", "head_pos": [19, 56], "relation": "used for", "tail": "continuous pen motion", "tail_type": "NA", "tail_pos": [70, 91]}], "task": "RE"}
+{"text": "This motor control representation enables successful word spotting and matching of cursive scripts .", "relation": [{"head": "motor control representation", "head_type": "NA", "head_pos": [8, 36], "relation": "used for", "tail": "word spotting", "tail_type": "NA", "tail_pos": [62, 75]}, {"head": "motor control representation", "head_type": "NA", "head_pos": [8, 36], "relation": "used for", "tail": "matching of cursive scripts", "tail_type": "NA", "tail_pos": [80, 107]}, {"head": "word spotting", "head_type": "NA", "head_pos": [56, 69], "relation": "conjunction", "tail": "matching of cursive scripts", "tail_type": "NA", "tail_pos": [80, 107]}], "task": "RE"}
+{"text": "Our experiments clearly indicate the potential of this dynamic representation for complete cursive handwriting recognition .", "relation": [{"head": "dynamic representation", "head_type": "NA", "head_pos": [58, 80], "relation": "used for", "tail": "cursive handwriting recognition", "tail_type": "NA", "tail_pos": [100, 131]}], "task": "RE"}
+{"text": "In the Object Recognition task , there exists a di-chotomy between the categorization of objects and estimating object pose , where the former necessitates a view-invariant representation , while the latter requires a representation capable of capturing pose information over different categories of objects .", "relation": [{"head": "categorization of objects", "head_type": "NA", "head_pos": [80, 105], "relation": "part of", "tail": "Object Recognition task", "tail_type": "NA", "tail_pos": [10, 33]}, {"head": "categorization of objects", "head_type": "NA", "head_pos": [74, 99], "relation": "conjunction", "tail": "estimating object pose", "tail_type": "NA", "tail_pos": [110, 132]}, {"head": "estimating object pose", "head_type": "NA", "head_pos": [110, 132], "relation": "part of", "tail": "Object Recognition task", "tail_type": "NA", "tail_pos": [10, 33]}, {"head": "view-invariant representation", "head_type": "NA", "head_pos": [167, 196], "relation": "used for", "tail": "former", "tail_type": "NA", "tail_pos": [139, 145]}, {"head": "representation", "head_type": "NA", "head_pos": [173, 187], "relation": "used for", "tail": "latter", "tail_type": "NA", "tail_pos": [203, 209]}, {"head": "representation", "head_type": "NA", "head_pos": [173, 187], "relation": "used for", "tail": "pose information", "tail_type": "NA", "tail_pos": [263, 279]}], "task": "RE"}
+{"text": "With the rise of deep archi-tectures , the prime focus has been on object category recognition .", "relation": [{"head": "deep archi-tectures", "head_type": "NA", "head_pos": [20, 39], "relation": "used for", "tail": "object category recognition", "tail_type": "NA", "tail_pos": [76, 103]}], "task": "RE"}
+{"text": "In contrast , object pose estimation using these approaches has received relatively less attention .", "relation": [{"head": "approaches", "head_type": "NA", "head_pos": [58, 68], "relation": "used for", "tail": "object pose estimation", "tail_type": "NA", "tail_pos": [17, 39]}], "task": "RE"}
+{"text": "In this work , we study how Convolutional Neural Networks -LRB- CNN -RRB- architectures can be adapted to the task of simultaneous object recognition and pose estimation .", "relation": [{"head": "Convolutional Neural Networks -LRB- CNN -RRB- architectures", "head_type": "NA", "head_pos": [31, 90], "relation": "used for", "tail": "object recognition", "tail_type": "NA", "tail_pos": [140, 158]}, {"head": "Convolutional Neural Networks -LRB- CNN -RRB- architectures", "head_type": "NA", "head_pos": [31, 90], "relation": "used for", "tail": "pose estimation", "tail_type": "NA", "tail_pos": [163, 178]}, {"head": "object recognition", "head_type": "NA", "head_pos": [134, 152], "relation": "conjunction", "tail": "pose estimation", "tail_type": "NA", "tail_pos": [163, 178]}], "task": "RE"}
+{"text": "We investigate and analyze the layers of various CNN models and extensively compare between them with the goal of discovering how the layers of distributed representations within CNNs represent object pose information and how this contradicts with object category representations .", "relation": [{"head": "layers", "head_type": "NA", "head_pos": [34, 40], "relation": "part of", "tail": "CNN models", "tail_type": "NA", "tail_pos": [58, 68]}, {"head": "layers of distributed representations", "head_type": "NA", "head_pos": [137, 174], "relation": "part of", "tail": "CNNs", "tail_type": "NA", "tail_pos": [188, 192]}, {"head": "layers of distributed representations", "head_type": "NA", "head_pos": [137, 174], "relation": "used for", "tail": "object pose information", "tail_type": "NA", "tail_pos": [203, 226]}, {"head": "this", "head_type": "NA", "head_pos": [229, 233], "relation": "compare", "tail": "object category representations", "tail_type": "NA", "tail_pos": [257, 288]}], "task": "RE"}
+{"text": "It is particularly valuable to empirical MT research .", "relation": [{"head": "It", "head_type": "NA", "head_pos": [3, 5], "relation": "used for", "tail": "empirical MT research", "tail_type": "NA", "tail_pos": [40, 61]}], "task": "RE"}
+{"text": "In this paper , we explore geometric structures of 3D lines in ray space for improving light field triangulation and stereo matching .", "relation": [{"head": "geometric structures of 3D lines", "head_type": "NA", "head_pos": [30, 62], "relation": "used for", "tail": "light field triangulation", "tail_type": "NA", "tail_pos": [96, 121]}, {"head": "geometric structures of 3D lines", "head_type": "NA", "head_pos": [30, 62], "relation": "used for", "tail": "stereo matching", "tail_type": "NA", "tail_pos": [126, 141]}, {"head": "ray space", "head_type": "NA", "head_pos": [72, 81], "relation": "feature of", "tail": "geometric structures of 3D lines", "tail_type": "NA", "tail_pos": [30, 62]}, {"head": "light field triangulation", "head_type": "NA", "head_pos": [90, 115], "relation": "conjunction", "tail": "stereo matching", "tail_type": "NA", "tail_pos": [126, 141]}], "task": "RE"}
+{"text": "Such a triangulation provides a piecewise-linear interpolant useful for light field super-resolution .", "relation": [{"head": "triangulation", "head_type": "NA", "head_pos": [10, 23], "relation": "used for", "tail": "piecewise-linear interpolant", "tail_type": "NA", "tail_pos": [41, 69]}, {"head": "piecewise-linear interpolant", "head_type": "NA", "head_pos": [35, 63], "relation": "used for", "tail": "light field super-resolution", "tail_type": "NA", "tail_pos": [81, 109]}], "task": "RE"}
+{"text": "Experiments on synthetic and real data show that both our triangulation and LAGC algorithms outperform state-of-the-art solutions in accuracy and visual quality .", "relation": [{"head": "synthetic and real data", "head_type": "NA", "head_pos": [18, 41], "relation": "evaluate for", "tail": "triangulation and LAGC algorithms", "tail_type": "NA", "tail_pos": [67, 100]}, {"head": "synthetic and real data", "head_type": "NA", "head_pos": [18, 41], "relation": "evaluate for", "tail": "state-of-the-art solutions", "tail_type": "NA", "tail_pos": [112, 138]}, {"head": "triangulation and LAGC algorithms", "head_type": "NA", "head_pos": [61, 94], "relation": "compare", "tail": "state-of-the-art solutions", "tail_type": "NA", "tail_pos": [112, 138]}, {"head": "accuracy", "head_type": "NA", "head_pos": [142, 150], "relation": "evaluate for", "tail": "triangulation and LAGC algorithms", "tail_type": "NA", "tail_pos": [61, 94]}, {"head": "accuracy", "head_type": "NA", "head_pos": [142, 150], "relation": "evaluate for", "tail": "state-of-the-art solutions", "tail_type": "NA", "tail_pos": [106, 132]}, {"head": "visual quality", "head_type": "NA", "head_pos": [155, 169], "relation": "evaluate for", "tail": "triangulation and LAGC algorithms", "tail_type": "NA", "tail_pos": [61, 94]}, {"head": "visual quality", "head_type": "NA", "head_pos": [155, 169], "relation": "evaluate for", "tail": "state-of-the-art solutions", "tail_type": "NA", "tail_pos": [106, 132]}], "task": "RE"}
+{"text": "This paper presents a phrase-based statistical machine translation method , based on non-contiguous phrases , i.e. phrases with gaps .", "relation": [{"head": "non-contiguous phrases", "head_type": "NA", "head_pos": [94, 116], "relation": "used for", "tail": "phrase-based statistical machine translation method", "tail_type": "NA", "tail_pos": [25, 76]}], "task": "RE"}
+{"text": "A method for producing such phrases from a word-aligned corpora is proposed .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [5, 11], "relation": "used for", "tail": "phrases", "tail_type": "NA", "tail_pos": [37, 44]}, {"head": "word-aligned corpora", "head_type": "NA", "head_pos": [52, 72], "relation": "evaluate for", "tail": "method", "tail_type": "NA", "tail_pos": [5, 11]}], "task": "RE"}
+{"text": "A statistical translation model is also presented that deals such phrases , as well as a training method based on the maximization of translation accuracy , as measured with the NIST evaluation metric .", "relation": [{"head": "statistical translation model", "head_type": "NA", "head_pos": [5, 34], "relation": "used for", "tail": "phrases", "tail_type": "NA", "tail_pos": [75, 82]}, {"head": "maximization of translation accuracy", "head_type": "NA", "head_pos": [127, 163], "relation": "used for", "tail": "training method", "tail_type": "NA", "tail_pos": [92, 107]}, {"head": "NIST evaluation metric", "head_type": "NA", "head_pos": [187, 209], "relation": "evaluate for", "tail": "statistical translation model", "tail_type": "NA", "tail_pos": [5, 34]}], "task": "RE"}
+{"text": "Translations are produced by means of a beam-search decoder .", "relation": [{"head": "beam-search decoder", "head_type": "NA", "head_pos": [49, 68], "relation": "used for", "tail": "Translations", "tail_type": "NA", "tail_pos": [3, 15]}], "task": "RE"}
+{"text": "GLOSSER is designed to support reading and learning to read in a foreign language .", "relation": [{"head": "GLOSSER", "head_type": "NA", "head_pos": [3, 10], "relation": "used for", "tail": "reading and learning", "tail_type": "NA", "tail_pos": [40, 60]}], "task": "RE"}
+{"text": "There are four language pairs currently supported by GLOSSER : English-Bulgarian , English-Estonian , English-Hungarian and French-Dutch .", "relation": [{"head": "language pairs", "head_type": "NA", "head_pos": [18, 32], "relation": "used for", "tail": "GLOSSER", "tail_type": "NA", "tail_pos": [62, 69]}, {"head": "English-Bulgarian", "head_type": "NA", "head_pos": [72, 89], "relation": "hyponym of", "tail": "language pairs", "tail_type": "NA", "tail_pos": [18, 32]}, {"head": "English-Bulgarian", "head_type": "NA", "head_pos": [66, 83], "relation": "conjunction", "tail": "English-Estonian", "tail_type": "NA", "tail_pos": [92, 108]}, {"head": "English-Estonian", "head_type": "NA", "head_pos": [92, 108], "relation": "hyponym of", "tail": "language pairs", "tail_type": "NA", "tail_pos": [18, 32]}, {"head": "English-Estonian", "head_type": "NA", "head_pos": [86, 102], "relation": "conjunction", "tail": "English-Hungarian", "tail_type": "NA", "tail_pos": [111, 128]}, {"head": "English-Hungarian", "head_type": "NA", "head_pos": [111, 128], "relation": "hyponym of", "tail": "language pairs", "tail_type": "NA", "tail_pos": [18, 32]}, {"head": "English-Hungarian", "head_type": "NA", "head_pos": [105, 122], "relation": "conjunction", "tail": "French-Dutch", "tail_type": "NA", "tail_pos": [133, 145]}, {"head": "French-Dutch", "head_type": "NA", "head_pos": [133, 145], "relation": "hyponym of", "tail": "language pairs", "tail_type": "NA", "tail_pos": [18, 32]}], "task": "RE"}
+{"text": "A demonstration -LRB- in UNIX -RRB- for Applied Natural Language Processing emphasizes components put to novel technical uses in intelligent computer-assisted morphological analysis -LRB- ICALL -RRB- , including disambiguated morphological analysis and lemmatized indexing for an aligned bilingual corpus of word examples .", "relation": [{"head": "components", "head_type": "NA", "head_pos": [90, 100], "relation": "used for", "tail": "intelligent computer-assisted morphological analysis -LRB- ICALL -RRB-", "tail_type": "NA", "tail_pos": [138, 208]}, {"head": "disambiguated morphological analysis", "head_type": "NA", "head_pos": [221, 257], "relation": "hyponym of", "tail": "components", "tail_type": "NA", "tail_pos": [90, 100]}, {"head": "disambiguated morphological analysis", "head_type": "NA", "head_pos": [215, 251], "relation": "conjunction", "tail": "lemmatized indexing", "tail_type": "NA", "tail_pos": [262, 281]}, {"head": "disambiguated morphological analysis", "head_type": "NA", "head_pos": [215, 251], "relation": "used for", "tail": "aligned bilingual corpus", "tail_type": "NA", "tail_pos": [289, 313]}, {"head": "lemmatized indexing", "head_type": "NA", "head_pos": [262, 281], "relation": "hyponym of", "tail": "components", "tail_type": "NA", "tail_pos": [90, 100]}, {"head": "lemmatized indexing", "head_type": "NA", "head_pos": [256, 275], "relation": "used for", "tail": "aligned bilingual corpus", "tail_type": "NA", "tail_pos": [289, 313]}], "task": "RE"}
+{"text": "We present a new part-of-speech tagger that demonstrates the following ideas : -LRB- i -RRB- explicit use of both preceding and following tag contexts via a dependency network representation , -LRB- ii -RRB- broad use of lexical features , including jointly conditioning on multiple consecutive words , -LRB- iii -RRB- effective use of priors in conditional loglinear models , and -LRB- iv -RRB- fine-grained modeling of unknown word features .", "relation": [{"head": "tag contexts", "head_type": "NA", "head_pos": [147, 159], "relation": "used for", "tail": "part-of-speech tagger", "tail_type": "NA", "tail_pos": [20, 41]}, {"head": "dependency network representation", "head_type": "NA", "head_pos": [166, 199], "relation": "used for", "tail": "tag contexts", "tail_type": "NA", "tail_pos": [141, 153]}, {"head": "lexical features", "head_type": "NA", "head_pos": [230, 246], "relation": "used for", "tail": "part-of-speech tagger", "tail_type": "NA", "tail_pos": [20, 41]}, {"head": "priors in conditional loglinear models", "head_type": "NA", "head_pos": [345, 383], "relation": "used for", "tail": "part-of-speech tagger", "tail_type": "NA", "tail_pos": [20, 41]}, {"head": "fine-grained modeling of unknown word features", "head_type": "NA", "head_pos": [405, 451], "relation": "used for", "tail": "part-of-speech tagger", "tail_type": "NA", "tail_pos": [20, 41]}], "task": "RE"}
+{"text": "Using these ideas together , the resulting tagger gives a 97.24 % accuracy on the Penn Treebank WSJ , an error reduction of 4.4 % on the best previous single automatically learned tagging result .", "relation": [{"head": "accuracy", "head_type": "NA", "head_pos": [75, 83], "relation": "evaluate for", "tail": "tagger", "tail_type": "NA", "tail_pos": [46, 52]}, {"head": "Penn Treebank WSJ", "head_type": "NA", "head_pos": [91, 108], "relation": "evaluate for", "tail": "tagger", "tail_type": "NA", "tail_pos": [46, 52]}, {"head": "error", "head_type": "NA", "head_pos": [114, 119], "relation": "evaluate for", "tail": "tagger", "tail_type": "NA", "tail_pos": [46, 52]}], "task": "RE"}
+{"text": "Owing to these variations , the pedestrian data is distributed as highly-curved manifolds in the feature space , despite the current convolutional neural networks -LRB- CNN -RRB- 's capability of feature extraction .", "relation": [{"head": "highly-curved manifolds", "head_type": "NA", "head_pos": [75, 98], "relation": "used for", "tail": "pedestrian data", "tail_type": "NA", "tail_pos": [35, 50]}, {"head": "feature space", "head_type": "NA", "head_pos": [106, 119], "relation": "feature of", "tail": "highly-curved manifolds", "tail_type": "NA", "tail_pos": [69, 92]}, {"head": "convolutional neural networks -LRB- CNN -RRB-", "head_type": "NA", "head_pos": [136, 181], "relation": "used for", "tail": "feature extraction", "tail_type": "NA", "tail_pos": [205, 223]}], "task": "RE"}
+{"text": "In practice , the current deep embedding methods use the Euclidean distance for the training and test .", "relation": [{"head": "Euclidean distance", "head_type": "NA", "head_pos": [66, 84], "relation": "used for", "tail": "deep embedding methods", "tail_type": "NA", "tail_pos": [29, 51]}], "task": "RE"}
+{"text": "On the other hand , the manifold learning methods suggest to use the Euclidean distance in the local range , combining with the graphical relationship between samples , for approximating the geodesic distance .", "relation": [{"head": "Euclidean distance", "head_type": "NA", "head_pos": [78, 96], "relation": "used for", "tail": "manifold learning methods", "tail_type": "NA", "tail_pos": [27, 52]}, {"head": "Euclidean distance", "head_type": "NA", "head_pos": [72, 90], "relation": "conjunction", "tail": "graphical relationship", "tail_type": "NA", "tail_pos": [137, 159]}, {"head": "Euclidean distance", "head_type": "NA", "head_pos": [72, 90], "relation": "used for", "tail": "geodesic distance", "tail_type": "NA", "tail_pos": [200, 217]}, {"head": "local range", "head_type": "NA", "head_pos": [104, 115], "relation": "feature of", "tail": "Euclidean distance", "tail_type": "NA", "tail_pos": [72, 90]}, {"head": "graphical relationship", "head_type": "NA", "head_pos": [131, 153], "relation": "used for", "tail": "geodesic distance", "tail_type": "NA", "tail_pos": [200, 217]}], "task": "RE"}
+{"text": "From this point of view , selecting suitable positive -LRB- i.e. intra-class -RRB- training samples within a local range is critical for training the CNN embedding , especially when the data has large intra-class variations .", "relation": [{"head": "intra-class variations", "head_type": "NA", "head_pos": [210, 232], "relation": "feature of", "tail": "data", "tail_type": "NA", "tail_pos": [189, 193]}], "task": "RE"}
+{"text": "In this paper , we propose a novel moderate positive sample mining method to train robust CNN for person re-identification , dealing with the problem of large variation .", "relation": [{"head": "moderate positive sample mining method", "head_type": "NA", "head_pos": [38, 76], "relation": "used for", "tail": "robust CNN", "tail_type": "NA", "tail_pos": [92, 102]}, {"head": "robust CNN", "head_type": "NA", "head_pos": [86, 96], "relation": "used for", "tail": "person re-identification", "tail_type": "NA", "tail_pos": [107, 131]}], "task": "RE"}
+{"text": "In addition , we improve the learning by a metric weight constraint , so that the learned metric has a better generalization ability .", "relation": [{"head": "metric weight constraint", "head_type": "NA", "head_pos": [52, 76], "relation": "used for", "tail": "learning", "tail_type": "NA", "tail_pos": [32, 40]}, {"head": "generalization ability", "head_type": "NA", "head_pos": [119, 141], "relation": "feature of", "tail": "learned metric", "tail_type": "NA", "tail_pos": [85, 99]}], "task": "RE"}
+{"text": "Experiments show that these two strategies are effective in learning robust deep metrics for person re-identification , and accordingly our deep model significantly outperforms the state-of-the-art methods on several benchmarks of person re-identification .", "relation": [{"head": "robust deep metrics", "head_type": "NA", "head_pos": [72, 91], "relation": "used for", "tail": "person re-identification", "tail_type": "NA", "tail_pos": [102, 126]}, {"head": "deep model", "head_type": "NA", "head_pos": [143, 153], "relation": "compare", "tail": "state-of-the-art methods", "tail_type": "NA", "tail_pos": [190, 214]}, {"head": "deep model", "head_type": "NA", "head_pos": [143, 153], "relation": "used for", "tail": "person re-identification", "tail_type": "NA", "tail_pos": [93, 117]}, {"head": "state-of-the-art methods", "head_type": "NA", "head_pos": [184, 208], "relation": "used for", "tail": "person re-identification", "tail_type": "NA", "tail_pos": [93, 117]}], "task": "RE"}
+{"text": "Therefore , the study presented in this paper may be useful in inspiring new designs of deep models for person re-identification .", "relation": [{"head": "deep models", "head_type": "NA", "head_pos": [91, 102], "relation": "used for", "tail": "person re-identification", "tail_type": "NA", "tail_pos": [113, 137]}], "task": "RE"}
+{"text": "Utterance Verification -LRB- UV -RRB- is a critical function of an Automatic Speech Recognition -LRB- ASR -RRB- System working on real applications where spontaneous speech , out-of-vocabulary -LRB- OOV -RRB- words and acoustic noises are present .", "relation": [{"head": "Utterance Verification -LRB- UV -RRB-", "head_type": "NA", "head_pos": [3, 40], "relation": "hyponym of", "tail": "Automatic Speech Recognition -LRB- ASR -RRB- System", "tail_type": "NA", "tail_pos": [76, 127]}], "task": "RE"}
+{"text": "In this paper we present a new UV procedure with two major features : a -RRB- Confidence tests are applied to decoded string hypotheses obtained from using word and garbage models that represent OOV words and noises .", "relation": [{"head": "Confidence tests", "head_type": "NA", "head_pos": [81, 97], "relation": "used for", "tail": "decoded string hypotheses", "tail_type": "NA", "tail_pos": [119, 144]}, {"head": "noises", "head_type": "NA", "head_pos": [218, 224], "relation": "conjunction", "tail": "OOV words", "tail_type": "NA", "tail_pos": [198, 207]}], "task": "RE"}
+{"text": "Thus the ASR system is designed to deal with what we refer to as Word Spotting and Noise Spotting capabilities .", "relation": [{"head": "ASR system", "head_type": "NA", "head_pos": [12, 22], "relation": "used for", "tail": "Word Spotting", "tail_type": "NA", "tail_pos": [74, 87]}, {"head": "ASR system", "head_type": "NA", "head_pos": [12, 22], "relation": "used for", "tail": "Noise Spotting capabilities", "tail_type": "NA", "tail_pos": [92, 119]}], "task": "RE"}
+{"text": "b -RRB- The UV procedure is based on three different confidence tests , two based on acoustic measures and one founded on linguistic information , applied in a hierarchical structure .", "relation": [{"head": "confidence tests", "head_type": "NA", "head_pos": [62, 78], "relation": "used for", "tail": "UV procedure", "tail_type": "NA", "tail_pos": [15, 27]}, {"head": "confidence tests", "head_type": "NA", "head_pos": [56, 72], "relation": "used for", "tail": "hierarchical structure", "tail_type": "NA", "tail_pos": [169, 191]}, {"head": "two", "head_type": "NA", "head_pos": [81, 84], "relation": "hyponym of", "tail": "confidence tests", "tail_type": "NA", "tail_pos": [56, 72]}, {"head": "acoustic measures", "head_type": "NA", "head_pos": [94, 111], "relation": "used for", "tail": "two", "tail_type": "NA", "tail_pos": [75, 78]}, {"head": "one", "head_type": "NA", "head_pos": [116, 119], "relation": "hyponym of", "tail": "confidence tests", "tail_type": "NA", "tail_pos": [56, 72]}, {"head": "linguistic information", "head_type": "NA", "head_pos": [131, 153], "relation": "used for", "tail": "one", "tail_type": "NA", "tail_pos": [110, 113]}], "task": "RE"}
+{"text": "Experimental results from a real telephone application on a natural number recognition task show an 50 % reduction in recognition errors with a moderate 12 % rejection rate of correct utterances and a low 1.5 % rate of false acceptance .", "relation": [{"head": "natural number recognition task", "head_type": "NA", "head_pos": [69, 100], "relation": "feature of", "tail": "telephone application", "tail_type": "NA", "tail_pos": [36, 57]}, {"head": "recognition errors", "head_type": "NA", "head_pos": [127, 145], "relation": "evaluate for", "tail": "natural number recognition task", "tail_type": "NA", "tail_pos": [63, 94]}], "task": "RE"}
+{"text": "A critical step in encoding sound for neuronal processing occurs when the analog pressure wave is coded into discrete nerve-action potentials .", "relation": [{"head": "encoding sound", "head_type": "NA", "head_pos": [22, 36], "relation": "used for", "tail": "neuronal processing", "tail_type": "NA", "tail_pos": [47, 66]}, {"head": "discrete nerve-action potentials", "head_type": "NA", "head_pos": [118, 150], "relation": "used for", "tail": "analog pressure wave", "tail_type": "NA", "tail_pos": [77, 97]}], "task": "RE"}
+{"text": "Recent pool models of the inner hair cell synapse do not reproduce the dead time period after an intense stimulus , so we used visual inspection and automatic speech recognition -LRB- ASR -RRB- to investigate an offset adaptation -LRB- OA -RRB- model proposed by Zhang et al. -LSB- 1 -RSB- .", "relation": [{"head": "pool models", "head_type": "NA", "head_pos": [10, 21], "relation": "used for", "tail": "inner hair cell synapse", "tail_type": "NA", "tail_pos": [35, 58]}, {"head": "visual inspection", "head_type": "NA", "head_pos": [130, 147], "relation": "conjunction", "tail": "automatic speech recognition -LRB- ASR -RRB-", "tail_type": "NA", "tail_pos": [158, 202]}, {"head": "visual inspection", "head_type": "NA", "head_pos": [130, 147], "relation": "used for", "tail": "offset adaptation -LRB- OA -RRB- model", "tail_type": "NA", "tail_pos": [221, 259]}, {"head": "automatic speech recognition -LRB- ASR -RRB-", "head_type": "NA", "head_pos": [152, 196], "relation": "used for", "tail": "offset adaptation -LRB- OA -RRB- model", "tail_type": "NA", "tail_pos": [221, 259]}], "task": "RE"}
+{"text": "OA improved phase locking in the auditory nerve -LRB- AN -RRB- and raised ASR accuracy for features derived from AN fibers -LRB- ANFs -RRB- .", "relation": [{"head": "OA", "head_type": "NA", "head_pos": [3, 5], "relation": "used for", "tail": "phase locking in the auditory nerve -LRB- AN -RRB-", "tail_type": "NA", "tail_pos": [21, 71]}, {"head": "OA", "head_type": "NA", "head_pos": [3, 5], "relation": "used for", "tail": "features", "tail_type": "NA", "tail_pos": [100, 108]}, {"head": "ASR accuracy", "head_type": "NA", "head_pos": [77, 89], "relation": "evaluate for", "tail": "features", "tail_type": "NA", "tail_pos": [100, 108]}, {"head": "AN fibers -LRB- ANFs -RRB-", "head_type": "NA", "head_pos": [122, 148], "relation": "used for", "tail": "features", "tail_type": "NA", "tail_pos": [94, 102]}], "task": "RE"}
+{"text": "We also found that OA is crucial for auditory processing by onset neurons -LRB- ONs -RRB- in the next neuronal stage , the auditory brainstem .", "relation": [{"head": "OA", "head_type": "NA", "head_pos": [22, 24], "relation": "used for", "tail": "auditory processing", "tail_type": "NA", "tail_pos": [46, 65]}, {"head": "onset neurons -LRB- ONs -RRB-", "head_type": "NA", "head_pos": [69, 98], "relation": "used for", "tail": "OA", "tail_type": "NA", "tail_pos": [22, 24]}], "task": "RE"}
+{"text": "Multi-layer perceptrons -LRB- MLPs -RRB- performed much better than standard Gaussian mixture models -LRB- GMMs -RRB- for both our ANF-based and ON-based auditory features .", "relation": [{"head": "Multi-layer perceptrons -LRB- MLPs -RRB-", "head_type": "NA", "head_pos": [3, 43], "relation": "compare", "tail": "Gaussian mixture models -LRB- GMMs -RRB-", "tail_type": "NA", "tail_pos": [86, 126]}, {"head": "Multi-layer perceptrons -LRB- MLPs -RRB-", "head_type": "NA", "head_pos": [3, 43], "relation": "used for", "tail": "ANF-based and ON-based auditory features", "tail_type": "NA", "tail_pos": [140, 180]}, {"head": "Gaussian mixture models -LRB- GMMs -RRB-", "head_type": "NA", "head_pos": [80, 120], "relation": "used for", "tail": "ANF-based and ON-based auditory features", "tail_type": "NA", "tail_pos": [140, 180]}], "task": "RE"}
+{"text": "Recent progress in computer vision has been driven by high-capacity models trained on large datasets .", "relation": [{"head": "high-capacity models", "head_type": "NA", "head_pos": [63, 83], "relation": "used for", "tail": "computer vision", "tail_type": "NA", "tail_pos": [22, 37]}, {"head": "large datasets", "head_type": "NA", "head_pos": [95, 109], "relation": "used for", "tail": "high-capacity models", "tail_type": "NA", "tail_pos": [57, 77]}], "task": "RE"}
+{"text": "Unfortunately , creating large datasets with pixel-level labels has been extremely costly due to the amount of human effort required .", "relation": [{"head": "pixel-level labels", "head_type": "NA", "head_pos": [54, 72], "relation": "feature of", "tail": "large datasets", "tail_type": "NA", "tail_pos": [28, 42]}], "task": "RE"}
+{"text": "In this paper , we present an approach to rapidly creating pixel-accurate semantic label maps for images extracted from modern computer games .", "relation": [{"head": "approach", "head_type": "NA", "head_pos": [33, 41], "relation": "used for", "tail": "pixel-accurate semantic label maps", "tail_type": "NA", "tail_pos": [68, 102]}, {"head": "pixel-accurate semantic label maps", "head_type": "NA", "head_pos": [62, 96], "relation": "used for", "tail": "images", "tail_type": "NA", "tail_pos": [107, 113]}, {"head": "images", "head_type": "NA", "head_pos": [101, 107], "relation": "part of", "tail": "modern computer games", "tail_type": "NA", "tail_pos": [129, 150]}], "task": "RE"}
+{"text": "We propose a novel step toward the unsupervised seg-mentation of whole objects by combining '' hints '' of partial scene segmentation offered by multiple soft , binary mattes .", "relation": [{"head": "partial scene segmentation", "head_type": "NA", "head_pos": [116, 142], "relation": "used for", "tail": "unsupervised seg-mentation of whole objects", "tail_type": "NA", "tail_pos": [38, 81]}, {"head": "soft , binary mattes", "head_type": "NA", "head_pos": [163, 183], "relation": "used for", "tail": "partial scene segmentation", "tail_type": "NA", "tail_pos": [110, 136]}], "task": "RE"}
+{"text": "These mattes are implied by a set of hypothesized object boundary fragments in the scene .", "relation": [{"head": "hypothesized object boundary fragments", "head_type": "NA", "head_pos": [46, 84], "relation": "used for", "tail": "mattes", "tail_type": "NA", "tail_pos": [9, 15]}], "task": "RE"}
+{"text": "This reflects contemporary methods for unsupervised object discovery from groups of images , and it allows us to define intuitive evaluation met-rics for our sets of segmentations based on the accurate and parsimonious delineation of scene objects .", "relation": [{"head": "contemporary methods", "head_type": "NA", "head_pos": [17, 37], "relation": "used for", "tail": "unsupervised object discovery", "tail_type": "NA", "tail_pos": [48, 77]}], "task": "RE"}
+{"text": "Our proposed approach builds on recent advances in spectral clustering , image matting , and boundary detection .", "relation": [{"head": "spectral clustering", "head_type": "NA", "head_pos": [60, 79], "relation": "used for", "tail": "approach", "tail_type": "NA", "tail_pos": [16, 24]}, {"head": "spectral clustering", "head_type": "NA", "head_pos": [54, 73], "relation": "conjunction", "tail": "image matting", "tail_type": "NA", "tail_pos": [82, 95]}, {"head": "image matting", "head_type": "NA", "head_pos": [82, 95], "relation": "used for", "tail": "approach", "tail_type": "NA", "tail_pos": [16, 24]}, {"head": "image matting", "head_type": "NA", "head_pos": [76, 89], "relation": "conjunction", "tail": "boundary detection", "tail_type": "NA", "tail_pos": [102, 120]}, {"head": "boundary detection", "head_type": "NA", "head_pos": [102, 120], "relation": "used for", "tail": "approach", "tail_type": "NA", "tail_pos": [16, 24]}], "task": "RE"}
+{"text": "It is demonstrated qualitatively and quantitatively on a dataset of scenes and is suitable for current work in unsupervised object discovery without top-down knowledge .", "relation": [{"head": "It", "head_type": "NA", "head_pos": [3, 5], "relation": "used for", "tail": "unsupervised object discovery", "tail_type": "NA", "tail_pos": [120, 149]}, {"head": "dataset of scenes", "head_type": "NA", "head_pos": [66, 83], "relation": "evaluate for", "tail": "It", "tail_type": "NA", "tail_pos": [3, 5]}], "task": "RE"}
+{"text": "Language resource quality is crucial in NLP .", "relation": [{"head": "Language resource quality", "head_type": "NA", "head_pos": [3, 28], "relation": "feature of", "tail": "NLP", "tail_type": "NA", "tail_pos": [49, 52]}], "task": "RE"}
+{"text": "Many of the resources used are derived from data created by human beings out of an NLP context , especially regarding MT and reference translations .", "relation": [{"head": "MT", "head_type": "NA", "head_pos": [127, 129], "relation": "hyponym of", "tail": "NLP", "tail_type": "NA", "tail_pos": [86, 89]}, {"head": "MT", "head_type": "NA", "head_pos": [121, 123], "relation": "conjunction", "tail": "reference translations", "tail_type": "NA", "tail_pos": [134, 156]}, {"head": "reference translations", "head_type": "NA", "head_pos": [134, 156], "relation": "hyponym of", "tail": "NLP", "tail_type": "NA", "tail_pos": [86, 89]}], "task": "RE"}
+{"text": "Indeed , automatic evaluations need high-quality data that allow the comparison of both automatic and human translations .", "relation": [{"head": "high-quality data", "head_type": "NA", "head_pos": [45, 62], "relation": "evaluate for", "tail": "automatic evaluations", "tail_type": "NA", "tail_pos": [12, 33]}], "task": "RE"}
+{"text": "This paper describes the impact of using different-quality references on evaluation .", "relation": [{"head": "different-quality references", "head_type": "NA", "head_pos": [44, 72], "relation": "used for", "tail": "evaluation", "tail_type": "NA", "tail_pos": [82, 92]}], "task": "RE"}
+{"text": "Thus , the limitations of the automatic metrics used within MT are also discussed in this regard .", "relation": [{"head": "automatic metrics", "head_type": "NA", "head_pos": [33, 50], "relation": "evaluate for", "tail": "MT", "tail_type": "NA", "tail_pos": [69, 71]}], "task": "RE"}
+{"text": "This poster paper describes a full scale two-level morphological description -LRB- Karttunen , 1983 ; Koskenniemi , 1983 -RRB- of Turkish word structures .", "relation": [{"head": "full scale two-level morphological description", "head_type": "NA", "head_pos": [33, 79], "relation": "used for", "tail": "Turkish word structures", "tail_type": "NA", "tail_pos": [139, 162]}], "task": "RE"}
+{"text": "The description has been implemented using the PC-KIMMO environment -LRB- Antworth , 1990 -RRB- and is based on a root word lexicon of about 23,000 roots words .", "relation": [{"head": "PC-KIMMO environment", "head_type": "NA", "head_pos": [56, 76], "relation": "used for", "tail": "description", "tail_type": "NA", "tail_pos": [7, 18]}, {"head": "root word lexicon", "head_type": "NA", "head_pos": [123, 140], "relation": "used for", "tail": "description", "tail_type": "NA", "tail_pos": [7, 18]}], "task": "RE"}
+{"text": "Turkish is an agglutinative language with word structures formed by productive affixations of derivational and inflectional suffixes to root words .", "relation": [{"head": "Turkish", "head_type": "NA", "head_pos": [3, 10], "relation": "hyponym of", "tail": "agglutinative language", "tail_type": "NA", "tail_pos": [23, 45]}, {"head": "word structures", "head_type": "NA", "head_pos": [51, 66], "relation": "feature of", "tail": "agglutinative language", "tail_type": "NA", "tail_pos": [17, 39]}, {"head": "productive affixations of derivational and inflectional suffixes", "head_type": "NA", "head_pos": [77, 141], "relation": "part of", "tail": "word structures", "tail_type": "NA", "tail_pos": [45, 60]}], "task": "RE"}
+{"text": "The surface realizations of morphological constructions are constrained and modified by a number of phonetic rules such as vowel harmony .", "relation": [{"head": "phonetic rules", "head_type": "NA", "head_pos": [109, 123], "relation": "used for", "tail": "surface realizations of morphological constructions", "tail_type": "NA", "tail_pos": [7, 58]}, {"head": "vowel harmony", "head_type": "NA", "head_pos": [132, 145], "relation": "hyponym of", "tail": "phonetic rules", "tail_type": "NA", "tail_pos": [103, 117]}], "task": "RE"}
+{"text": "This paper deals with the problem of generating the fundamental frequency -LRB- F0 -RRB- contour of speech from a text input for text-to-speech synthesis .", "relation": [{"head": "fundamental frequency -LRB- F0 -RRB- contour of speech", "head_type": "NA", "head_pos": [55, 109], "relation": "used for", "tail": "text-to-speech synthesis", "tail_type": "NA", "tail_pos": [138, 162]}, {"head": "text input", "head_type": "NA", "head_pos": [123, 133], "relation": "used for", "tail": "fundamental frequency -LRB- F0 -RRB- contour of speech", "tail_type": "NA", "tail_pos": [55, 109]}], "task": "RE"}
+{"text": "We have previously introduced a statistical model describing the generating process of speech F0 contours , based on the discrete-time version of the Fujisaki model .", "relation": [{"head": "statistical model", "head_type": "NA", "head_pos": [35, 52], "relation": "used for", "tail": "speech F0 contours", "tail_type": "NA", "tail_pos": [96, 114]}, {"head": "Fujisaki model", "head_type": "NA", "head_pos": [159, 173], "relation": "used for", "tail": "statistical model", "tail_type": "NA", "tail_pos": [35, 52]}], "task": "RE"}
+{"text": "One remarkable feature of this model is that it has allowed us to derive an efficient algorithm based on powerful statistical methods for estimating the Fujisaki-model parameters from raw F0 contours .", "relation": [{"head": "remarkable feature", "head_type": "NA", "head_pos": [7, 25], "relation": "feature of", "tail": "model", "tail_type": "NA", "tail_pos": [40, 45]}, {"head": "algorithm", "head_type": "NA", "head_pos": [89, 98], "relation": "used for", "tail": "Fujisaki-model parameters", "tail_type": "NA", "tail_pos": [162, 187]}, {"head": "statistical methods", "head_type": "NA", "head_pos": [123, 142], "relation": "used for", "tail": "algorithm", "tail_type": "NA", "tail_pos": [89, 98]}, {"head": "raw F0 contours", "head_type": "NA", "head_pos": [193, 208], "relation": "used for", "tail": "Fujisaki-model parameters", "tail_type": "NA", "tail_pos": [156, 181]}], "task": "RE"}
+{"text": "To associate a sequence of the Fujisaki-model parameters with a text input based on statistical learning , this paper proposes extending this model to a context-dependent one .", "relation": [{"head": "text input", "head_type": "NA", "head_pos": [73, 83], "relation": "used for", "tail": "Fujisaki-model parameters", "tail_type": "NA", "tail_pos": [34, 59]}, {"head": "statistical learning", "head_type": "NA", "head_pos": [93, 113], "relation": "used for", "tail": "Fujisaki-model parameters", "tail_type": "NA", "tail_pos": [34, 59]}], "task": "RE"}
+{"text": "We further propose a parameter training algorithm for the present model based on a decision tree-based context clustering .", "relation": [{"head": "parameter training algorithm", "head_type": "NA", "head_pos": [24, 52], "relation": "used for", "tail": "model", "tail_type": "NA", "tail_pos": [75, 80]}, {"head": "decision tree-based context clustering", "head_type": "NA", "head_pos": [92, 130], "relation": "used for", "tail": "parameter training algorithm", "tail_type": "NA", "tail_pos": [24, 52]}], "task": "RE"}
+{"text": "We introduce a method to accelerate the evaluation of object detection cascades with the help of a divide-and-conquer procedure in the space of candidate regions .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [18, 24], "relation": "used for", "tail": "evaluation of object detection cascades", "tail_type": "NA", "tail_pos": [49, 88]}, {"head": "divide-and-conquer procedure", "head_type": "NA", "head_pos": [108, 136], "relation": "used for", "tail": "method", "tail_type": "NA", "tail_pos": [18, 24]}, {"head": "space of candidate regions", "head_type": "NA", "head_pos": [144, 170], "relation": "feature of", "tail": "divide-and-conquer procedure", "tail_type": "NA", "tail_pos": [102, 130]}], "task": "RE"}
+{"text": "Compared to the exhaustive procedure that thus far is the state-of-the-art for cascade evaluation , the proposed method requires fewer evaluations of the classifier functions , thereby speeding up the search .", "relation": [{"head": "exhaustive procedure", "head_type": "NA", "head_pos": [19, 39], "relation": "used for", "tail": "cascade evaluation", "tail_type": "NA", "tail_pos": [88, 106]}, {"head": "exhaustive procedure", "head_type": "NA", "head_pos": [19, 39], "relation": "compare", "tail": "method", "tail_type": "NA", "tail_pos": [122, 128]}, {"head": "method", "head_type": "NA", "head_pos": [116, 122], "relation": "used for", "tail": "search", "tail_type": "NA", "tail_pos": [210, 216]}], "task": "RE"}
+{"text": "Furthermore , we show how the recently developed efficient subwindow search -LRB- ESS -RRB- procedure -LSB- 11 -RSB- can be integrated into the last stage of our method .", "relation": [{"head": "subwindow search -LRB- ESS -RRB- procedure", "head_type": "NA", "head_pos": [62, 104], "relation": "part of", "tail": "method", "tail_type": "NA", "tail_pos": [171, 177]}], "task": "RE"}
+{"text": "This allows us to use our method to act not only as a faster procedure for cascade evaluation , but also as a tool to perform efficient branch-and-bound object detection with nonlinear quality functions , in particular kernel-ized support vector machines .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [29, 35], "relation": "used for", "tail": "cascade evaluation", "tail_type": "NA", "tail_pos": [84, 102]}, {"head": "method", "head_type": "NA", "head_pos": [29, 35], "relation": "used for", "tail": "branch-and-bound object detection", "tail_type": "NA", "tail_pos": [145, 178]}, {"head": "nonlinear quality functions", "head_type": "NA", "head_pos": [184, 211], "relation": "used for", "tail": "branch-and-bound object detection", "tail_type": "NA", "tail_pos": [139, 172]}, {"head": "kernel-ized support vector machines", "head_type": "NA", "head_pos": [228, 263], "relation": "hyponym of", "tail": "nonlinear quality functions", "tail_type": "NA", "tail_pos": [178, 205]}], "task": "RE"}
+{"text": "Experiments on the PASCAL VOC 2006 dataset show an acceleration of more than 50 % by our method compared to standard cascade evaluation .", "relation": [{"head": "PASCAL VOC 2006 dataset", "head_type": "NA", "head_pos": [22, 45], "relation": "evaluate for", "tail": "method", "tail_type": "NA", "tail_pos": [98, 104]}, {"head": "PASCAL VOC 2006 dataset", "head_type": "NA", "head_pos": [22, 45], "relation": "evaluate for", "tail": "cascade evaluation", "tail_type": "NA", "tail_pos": [126, 144]}, {"head": "cascade evaluation", "head_type": "NA", "head_pos": [126, 144], "relation": "compare", "tail": "method", "tail_type": "NA", "tail_pos": [92, 98]}], "task": "RE"}
+{"text": "Background modeling is an important component of many vision systems .", "relation": [{"head": "Background modeling", "head_type": "NA", "head_pos": [3, 22], "relation": "part of", "tail": "vision systems", "tail_type": "NA", "tail_pos": [63, 77]}], "task": "RE"}
+{"text": "When the scene exhibits a persistent dynamic behavior in time , such an assumption is violated and detection performance deteriorates .", "relation": [{"head": "persistent dynamic behavior", "head_type": "NA", "head_pos": [35, 62], "relation": "feature of", "tail": "scene", "tail_type": "NA", "tail_pos": [12, 17]}], "task": "RE"}
+{"text": "In this paper , we propose a new method for the modeling and subtraction of such scenes .", "relation": [{"head": "method", "head_type": "NA", "head_pos": [36, 42], "relation": "used for", "tail": "modeling and subtraction of such scenes", "tail_type": "NA", "tail_pos": [57, 96]}], "task": "RE"}
+{"text": "Towards the modeling of the dynamic characteristics , optical flow is computed and utilized as a feature in a higher dimensional space .", "relation": [{"head": "optical flow", "head_type": "NA", "head_pos": [63, 75], "relation": "used for", "tail": "modeling of the dynamic characteristics", "tail_type": "NA", "tail_pos": [15, 54]}, {"head": "optical flow", "head_type": "NA", "head_pos": [57, 69], "relation": "used for", "tail": "feature", "tail_type": "NA", "tail_pos": [106, 113]}, {"head": "feature", "head_type": "NA", "head_pos": [106, 113], "relation": "used for", "tail": "modeling of the dynamic characteristics", "tail_type": "NA", "tail_pos": [15, 54]}, {"head": "higher dimensional space", "head_type": "NA", "head_pos": [119, 143], "relation": "feature of", "tail": "feature", "tail_type": "NA", "tail_pos": [100, 107]}], "task": "RE"}
+{"text": "Inherent ambiguities in the computation of features are addressed by using a data-dependent bandwidth for density estimation using kernels .", "relation": [{"head": "ambiguities", "head_type": "NA", "head_pos": [12, 23], "relation": "feature of", "tail": "computation of features", "tail_type": "NA", "tail_pos": [37, 60]}, {"head": "data-dependent bandwidth", "head_type": "NA", "head_pos": [86, 110], "relation": "used for", "tail": "ambiguities", "tail_type": "NA", "tail_pos": [12, 23]}, {"head": "data-dependent bandwidth", "head_type": "NA", "head_pos": [80, 104], "relation": "used for", "tail": "density estimation", "tail_type": "NA", "tail_pos": [115, 133]}, {"head": "kernels", "head_type": "NA", "head_pos": [140, 147], "relation": "used for", "tail": "density estimation", "tail_type": "NA", "tail_pos": [109, 127]}], "task": "RE"}
+{"text": "In this paper , we present our approach for using information extraction annotations to augment document retrieval for distillation .", "relation": [{"head": "information extraction annotations", "head_type": "NA", "head_pos": [53, 87], "relation": "used for", "tail": "document retrieval for distillation", "tail_type": "NA", "tail_pos": [105, 140]}], "task": "RE"}
+{"text": "This paper presents a novel representation for three-dimensional objects in terms of affine-invariant image patches and their spatial relationships .", "relation": [{"head": "representation", "head_type": "NA", "head_pos": [31, 45], "relation": "used for", "tail": "three-dimensional objects", "tail_type": "NA", "tail_pos": [56, 81]}, {"head": "affine-invariant image patches", "head_type": "NA", "head_pos": [94, 124], "relation": "feature of", "tail": "three-dimensional objects", "tail_type": "NA", "tail_pos": [50, 75]}, {"head": "spatial relationships", "head_type": "NA", "head_pos": [135, 156], "relation": "feature of", "tail": "affine-invariant image patches", "tail_type": "NA", "tail_pos": [88, 118]}], "task": "RE"}
+{"text": "Multi-view constraints associated with groups of patches are combined with a normalized representation of their appearance to guide matching and reconstruction , allowing the acquisition of true three-dimensional affine and Euclidean models from multiple images and their recognition in a single photograph taken from an arbitrary viewpoint .", "relation": [{"head": "Multi-view constraints", "head_type": "NA", "head_pos": [3, 25], "relation": "conjunction", "tail": "normalized representation", "tail_type": "NA", "tail_pos": [86, 111]}, {"head": "Multi-view constraints", "head_type": "NA", "head_pos": [3, 25], "relation": "used for", "tail": "matching", "tail_type": "NA", "tail_pos": [141, 149]}, {"head": "Multi-view constraints", "head_type": "NA", "head_pos": [3, 25], "relation": "used for", "tail": "reconstruction", "tail_type": "NA", "tail_pos": [154, 168]}, {"head": "normalized representation", "head_type": "NA", "head_pos": [80, 105], "relation": "used for", "tail": "matching", "tail_type": "NA", "tail_pos": [141, 149]}, {"head": "normalized representation", "head_type": "NA", "head_pos": [80, 105], "relation": "used for", "tail": "reconstruction", "tail_type": "NA", "tail_pos": [154, 168]}, {"head": "matching", "head_type": "NA", "head_pos": [135, 143], "relation": "conjunction", "tail": "reconstruction", "tail_type": "NA", "tail_pos": [154, 168]}, {"head": "images", "head_type": "NA", "head_pos": [264, 270], "relation": "used for", "tail": "acquisition of true three-dimensional affine and Euclidean models", "tail_type": "NA", "tail_pos": [178, 243]}], "task": "RE"}
+{"text": "The proposed approach does not require a separate segmentation stage and is applicable to cluttered scenes .", "relation": [{"head": "approach", "head_type": "NA", "head_pos": [16, 24], "relation": "used for", "tail": "cluttered scenes", "tail_type": "NA", "tail_pos": [99, 115]}], "task": "RE"}
+{"text": "Fast algorithms for nearest neighbor -LRB- NN -RRB- search have in large part focused on 2 distance .", "relation": [{"head": "Fast algorithms", "head_type": "NA", "head_pos": [3, 18], "relation": "used for", "tail": "nearest neighbor -LRB- NN -RRB- search", "tail_type": "NA", "tail_pos": [29, 67]}], "task": "RE"}
+{"text": "Here we develop an approach for 1 distance that begins with an explicit and exactly distance-preserving embedding of the points into 2 2 .", "relation": [{"head": "approach", "head_type": "NA", "head_pos": [22, 30], "relation": "used for", "tail": "1 distance", "tail_type": "NA", "tail_pos": [41, 51]}], "task": "RE"}
+{"text": "We show how this can efficiently be combined with random-projection based methods for 2 NN search , such as locality-sensitive hashing -LRB- LSH -RRB- or random projection trees .", "relation": [{"head": "this", "head_type": "NA", "head_pos": [15, 19], "relation": "conjunction", "tail": "random-projection based methods", "tail_type": "NA", "tail_pos": [59, 90]}, {"head": "random-projection based methods", "head_type": "NA", "head_pos": [53, 84], "relation": "used for", "tail": "NN search", "tail_type": "NA", "tail_pos": [97, 106]}, {"head": "locality-sensitive hashing -LRB- LSH -RRB-", "head_type": "NA", "head_pos": [117, 159], "relation": "hyponym of", "tail": "random-projection based methods", "tail_type": "NA", "tail_pos": [53, 84]}, {"head": "locality-sensitive hashing -LRB- LSH -RRB-", "head_type": "NA", "head_pos": [111, 153], "relation": "conjunction", "tail": "random projection trees", "tail_type": "NA", "tail_pos": [163, 186]}, {"head": "random projection trees", "head_type": "NA", "head_pos": [163, 186], "relation": "hyponym of", "tail": "random-projection based methods", "tail_type": "NA", "tail_pos": [53, 84]}], "task": "RE"}
+{"text": "We rigorously establish the correctness of the methodology and show by experimentation using LSH that it is competitive in practice with available alternatives .", "relation": [{"head": "it", "head_type": "NA", "head_pos": [105, 107], "relation": "compare", "tail": "alternatives", "tail_type": "NA", "tail_pos": [156, 168]}], "task": "RE"}