holylovenia
commited on
Commit
·
e0f3f6d
1
Parent(s):
e78cde4
Upload wikiann.py with huggingface_hub
Browse files- wikiann.py +197 -0
wikiann.py
ADDED
@@ -0,0 +1,197 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from pathlib import Path
|
2 |
+
from typing import List
|
3 |
+
|
4 |
+
import datasets
|
5 |
+
from datasets import NamedSplit
|
6 |
+
|
7 |
+
from nusacrowd.utils import schemas
|
8 |
+
from nusacrowd.utils.configs import NusantaraConfig
|
9 |
+
from nusacrowd.utils.constants import (DEFAULT_NUSANTARA_VIEW_NAME,
|
10 |
+
DEFAULT_SOURCE_VIEW_NAME, Tasks)
|
11 |
+
|
12 |
+
_DATASETNAME = "wikiann"
|
13 |
+
_SOURCE_VIEW_NAME = DEFAULT_SOURCE_VIEW_NAME
|
14 |
+
_UNIFIED_VIEW_NAME = DEFAULT_NUSANTARA_VIEW_NAME
|
15 |
+
|
16 |
+
_LANGUAGES = ["ind", "eng", "jav", "min", "sun", "ace", "mly", "map-bms"]
|
17 |
+
_LOCAL = False
|
18 |
+
_CITATION = """\
|
19 |
+
@inproceedings{pan-etal-2017-cross,
|
20 |
+
title = "Cross-lingual Name Tagging and Linking for 282 Languages",
|
21 |
+
author = "Pan, Xiaoman and
|
22 |
+
Zhang, Boliang and
|
23 |
+
May, Jonathan and
|
24 |
+
Nothman, Joel and
|
25 |
+
Knight, Kevin and
|
26 |
+
Ji, Heng",
|
27 |
+
booktitle = "Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
|
28 |
+
month = jul,
|
29 |
+
year = "2017",
|
30 |
+
address = "Vancouver, Canada",
|
31 |
+
publisher = "Association for Computational Linguistics",
|
32 |
+
url = "https://www.aclweb.org/anthology/P17-1178",
|
33 |
+
doi = "10.18653/v1/P17-1178",
|
34 |
+
pages = "1946--1958",
|
35 |
+
abstract = "The ambitious goal of this work is to develop a cross-lingual name tagging and linking framework
|
36 |
+
for 282 languages that exist in Wikipedia. Given a document in any of these languages, our framework is able
|
37 |
+
to identify name mentions, assign a coarse-grained or fine-grained type to each mention, and link it to
|
38 |
+
an English Knowledge Base (KB) if it is linkable. We achieve this goal by performing a series of
|
39 |
+
new KB mining methods: generating {``}silver-standard{''} annotations by
|
40 |
+
transferring annotations from English to other languages through cross-lingual links and KB properties,
|
41 |
+
refining annotations through self-training and topic selection,
|
42 |
+
deriving language-specific morphology features from anchor links, and mining word translation pairs from
|
43 |
+
cross-lingual links. Both name tagging and linking results for 282 languages are promising
|
44 |
+
on Wikipedia data and on-Wikipedia data.",
|
45 |
+
}
|
46 |
+
@inproceedings{rahimi-etal-2019-massively,
|
47 |
+
title = "Massively Multilingual Transfer for {NER}",
|
48 |
+
author = "Rahimi, Afshin and
|
49 |
+
Li, Yuan and
|
50 |
+
Cohn, Trevor",
|
51 |
+
booktitle = "Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics",
|
52 |
+
month = jul,
|
53 |
+
year = "2019",
|
54 |
+
address = "Florence, Italy",
|
55 |
+
publisher = "Association for Computational Linguistics",
|
56 |
+
url = "https://www.aclweb.org/anthology/P19-1015",
|
57 |
+
pages = "151--164",
|
58 |
+
}
|
59 |
+
"""
|
60 |
+
|
61 |
+
_DESCRIPTION = """\
|
62 |
+
The wikiann dataset contains NER tags with labels from O (0), B-PER (1), I-PER (2), B-ORG (3), I-ORG (4), B-LOC (5), I-LOC (6). The Indonesian subset is used.
|
63 |
+
WikiANN (sometimes called PAN-X) is a multilingual named entity recognition dataset consisting of Wikipedia articles
|
64 |
+
annotated with LOC (location), PER (person), and ORG (organisation)
|
65 |
+
tags in the IOB2 format. This version corresponds to the balanced train, dev, and test splits of
|
66 |
+
Rahimi et al. (2019), and uses the following subsets from the original WikiANN corpus
|
67 |
+
|
68 |
+
Language WikiAnn ISO 639-3
|
69 |
+
Indonesian id ind
|
70 |
+
Javanese jv jav
|
71 |
+
Minangkabau min min
|
72 |
+
Sundanese su sun
|
73 |
+
Acehnese ace ace
|
74 |
+
Malay ms mly
|
75 |
+
Banyumasan map-bms map-bms
|
76 |
+
|
77 |
+
|
78 |
+
"""
|
79 |
+
|
80 |
+
_HOMEPAGE = "https://github.com/afshinrahimi/mmner"
|
81 |
+
|
82 |
+
_LICENSE = "Apache-2.0 license"
|
83 |
+
|
84 |
+
_URLs = {
|
85 |
+
"wikiann": "https://s3.amazonaws.com/datasets.huggingface.co/wikiann/1.1.0/panx_dataset.zip",
|
86 |
+
}
|
87 |
+
|
88 |
+
_SUPPORTED_TASKS = [Tasks.NAMED_ENTITY_RECOGNITION]
|
89 |
+
|
90 |
+
_SOURCE_VERSION = "1.1.0"
|
91 |
+
_NUSANTARA_VERSION = "1.0.0"
|
92 |
+
|
93 |
+
|
94 |
+
def nusantara_config_constructor(lang, schema, version):
|
95 |
+
if lang == "":
|
96 |
+
raise ValueError(f"Invalid lang {lang}")
|
97 |
+
|
98 |
+
if schema != "source" and schema != "nusantara_seq_label":
|
99 |
+
raise ValueError(f"Invalid schema: {schema}")
|
100 |
+
|
101 |
+
return NusantaraConfig(
|
102 |
+
name="wikiann_{lang}_{schema}".format(lang=lang, schema=schema),
|
103 |
+
version=datasets.Version(version),
|
104 |
+
description="wikiann with {schema} schema for {lang} language".format(lang=lang, schema=schema),
|
105 |
+
schema=schema,
|
106 |
+
subset_id="wikiann",
|
107 |
+
)
|
108 |
+
|
109 |
+
|
110 |
+
LANGUAGES_MAP = {"eng": "english", "ind": "indonesian", "jav": "javanese", "min": "minangkabau", "sun": "sundanese", "ace": "acehnese", "mly": "malay", "map_bms": "banyumasan"} # Actual code is map-bms
|
111 |
+
LANG_CODES = {"eng": "en", "ind": "id", "jav": "jv", "min": "min", "sun": "su", "ace": "ace", "mly": "ms", "map_bms": "map-bms"}
|
112 |
+
|
113 |
+
|
114 |
+
class WikiAnnDataset(datasets.GeneratorBasedBuilder):
|
115 |
+
"""wikiann is an NER tagging dataset consisting of Wikipedia articles annotated with LOC, PER, and ORG tags
|
116 |
+
for multiple Indonesian language. If the language is not specified, it loads the Indonesian subset."""
|
117 |
+
|
118 |
+
label_classes = ["B-PER", "I-PER", "B-ORG", "I-ORG", "B-LOC", "I-LOC", "O"]
|
119 |
+
|
120 |
+
BUILDER_CONFIGS = [nusantara_config_constructor(lang, "source", _SOURCE_VERSION) for lang in LANGUAGES_MAP] + [nusantara_config_constructor(lang, "nusantara_seq_label", _NUSANTARA_VERSION) for lang in LANGUAGES_MAP]
|
121 |
+
|
122 |
+
DEFAULT_CONFIG_NAME = "wikiann_ind_source"
|
123 |
+
|
124 |
+
def _info(self):
|
125 |
+
if self.config.schema == "source":
|
126 |
+
features = datasets.Features({"index": datasets.Value("string"), "tokens": [datasets.Value("string")], "ner_tag": [datasets.Value("string")]})
|
127 |
+
elif self.config.schema == "nusantara_seq_label":
|
128 |
+
features = schemas.seq_label_features(self.label_classes)
|
129 |
+
|
130 |
+
return datasets.DatasetInfo(
|
131 |
+
description=_DESCRIPTION,
|
132 |
+
features=features,
|
133 |
+
homepage=_HOMEPAGE,
|
134 |
+
license=_LICENSE,
|
135 |
+
citation=_CITATION,
|
136 |
+
)
|
137 |
+
|
138 |
+
def get_lang(self, name):
|
139 |
+
return name.removesuffix("_source").removesuffix("_nusantara_seq_label").removeprefix("wikiann_")
|
140 |
+
|
141 |
+
def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
|
142 |
+
path = Path(dl_manager.download_and_extract(_URLs["wikiann"]))
|
143 |
+
lang = LANG_CODES[self.get_lang(self.config.name)]
|
144 |
+
wikiann_dl_dir = path / f"{lang}.tar.gz"
|
145 |
+
return [
|
146 |
+
datasets.SplitGenerator(
|
147 |
+
name=datasets.Split.VALIDATION,
|
148 |
+
gen_kwargs={"split": "dev", "filepath": dl_manager.iter_archive(wikiann_dl_dir)},
|
149 |
+
),
|
150 |
+
datasets.SplitGenerator(
|
151 |
+
name=datasets.Split.TEST,
|
152 |
+
gen_kwargs={"split": "test", "filepath": dl_manager.iter_archive(wikiann_dl_dir)},
|
153 |
+
),
|
154 |
+
datasets.SplitGenerator(
|
155 |
+
name=datasets.Split.TRAIN,
|
156 |
+
gen_kwargs={"split": "train", "filepath": dl_manager.iter_archive(wikiann_dl_dir)},
|
157 |
+
),
|
158 |
+
datasets.SplitGenerator(
|
159 |
+
name=NamedSplit("extra"),
|
160 |
+
gen_kwargs={"split": "extra", "filepath": dl_manager.iter_archive(wikiann_dl_dir)},
|
161 |
+
),
|
162 |
+
]
|
163 |
+
|
164 |
+
def _generate_examples(self, filepath: Path, split):
|
165 |
+
"""Based on https://github.com/huggingface/datasets/blob/main/datasets/wikiann/wikiann.py"""
|
166 |
+
fps = filepath
|
167 |
+
tokens = []
|
168 |
+
ner_tags = []
|
169 |
+
langs = []
|
170 |
+
guid_index = 0
|
171 |
+
for k, file in fps:
|
172 |
+
if k == split:
|
173 |
+
for line in file:
|
174 |
+
line = line.decode("utf-8")
|
175 |
+
if line == "" or line == "\n":
|
176 |
+
if tokens:
|
177 |
+
if self.config.schema == "source":
|
178 |
+
yield guid_index, {"index": str(guid_index), "tokens": tokens, "ner_tag": ner_tags}
|
179 |
+
elif self.config.schema == "nusantara_seq_label":
|
180 |
+
yield guid_index, {"id": str(guid_index), "tokens": tokens, "labels": ner_tags}
|
181 |
+
else:
|
182 |
+
raise ValueError(f"Invalid config: {self.config.name}")
|
183 |
+
guid_index += 1
|
184 |
+
tokens = []
|
185 |
+
ner_tags = []
|
186 |
+
langs = []
|
187 |
+
else:
|
188 |
+
# wikiann data is tab separated
|
189 |
+
splits = line.split("\t")
|
190 |
+
# strip out en: prefix
|
191 |
+
langs.append(splits[0].split(":")[0])
|
192 |
+
tokens.append(":".join(splits[0].split(":")[1:]))
|
193 |
+
if len(splits) > 1:
|
194 |
+
ner_tags.append(splits[-1].replace("\n", ""))
|
195 |
+
else:
|
196 |
+
# examples have no label in test set
|
197 |
+
ner_tags.append("O")
|