initial release

README.md

@@ -0,0 +1,93 @@
+---
+language:
+- "lzh"
+tags:
+- "classical chinese"
+- "literary chinese"
+- "ancient chinese"
+- "question-answering"
+- "dependency-parsing"
+datasets:
+- "universal_dependencies"
+license: "apache-2.0"
+pipeline_tag: "question-answering"
+widget:
+- text: "穴"
+  context: "不入虎穴不得虎子"
+- text: "子"
+  context: "不入虎穴不得虎子"
+- text: "不"
+  context: "[MASK]入虎穴不得虎子"
+---
+
+# bert-ancient-chinese-base-ud-head
+
+## Model Description
+
+This is a BERT model pre-trained on Classical Chinese texts for dependency-parsing (head-detection on long-unit-words) as question-answering, derived from [bert-ancient-chinese](https://huggingface.co/Jihuai/bert-ancient-chinese) and [UD_Classical_Chinese-Kyoto](https://github.com/UniversalDependencies/UD_Classical_Chinese-Kyoto). Use [MASK] inside `context` to avoid ambiguity when specifying a multiple-used word as `question`.
+
+## How to Use
+
+```py
+from transformers import AutoTokenizer,AutoModelForQuestionAnswering,QuestionAnsweringPipeline
+tokenizer=AutoTokenizer.from_pretrained("KoichiYasuoka/bert-ancient-chinese-base-ud-head")
+model=AutoModelForQuestionAnswering.from_pretrained("KoichiYasuoka/bert-ancient-chinese-base-ud-head")
+qap=QuestionAnsweringPipeline(tokenizer=tokenizer,model=model)
+print(qap(question="穴",context="不入虎穴不得虎子"))
+```
+
+or (with [ufal.chu-liu-edmonds](https://pypi.org/project/ufal.chu-liu-edmonds/))
+
+```py
+class TransformersUD(object):
+  def __init__(self,bert):
+    import os
+    from transformers import (AutoTokenizer,AutoModelForQuestionAnswering,
+      AutoModelForTokenClassification,AutoConfig,TokenClassificationPipeline)
+    self.tokenizer=AutoTokenizer.from_pretrained(bert)
+    self.model=AutoModelForQuestionAnswering.from_pretrained(bert)
+    x=AutoModelForTokenClassification.from_pretrained
+    if os.path.isdir(bert):
+      d,t=x(os.path.join(bert,"deprel")),x(os.path.join(bert,"tagger"))
+    else:
+      from transformers.file_utils import hf_bucket_url
+      c=AutoConfig.from_pretrained(hf_bucket_url(bert,"deprel/config.json"))
+      d=x(hf_bucket_url(bert,"deprel/pytorch_model.bin"),config=c)
+      s=AutoConfig.from_pretrained(hf_bucket_url(bert,"tagger/config.json"))
+      t=x(hf_bucket_url(bert,"tagger/pytorch_model.bin"),config=s)
+    self.deprel=TokenClassificationPipeline(model=d,tokenizer=self.tokenizer,
+      aggregation_strategy="simple")
+    self.tagger=TokenClassificationPipeline(model=t,tokenizer=self.tokenizer)
+  def __call__(self,text):
+    import numpy,torch,ufal.chu_liu_edmonds
+    w=[(t["start"],t["end"],t["entity_group"]) for t in self.deprel(text)]
+    z,n={t["start"]:t["entity"].split("|") for t in self.tagger(text)},len(w)
+    r,m=[text[s:e] for s,e,p in w],numpy.full((n+1,n+1),numpy.nan)
+    v,c=self.tokenizer(r,add_special_tokens=False)["input_ids"],[]
+    for i,t in enumerate(v):
+      q=[self.tokenizer.cls_token_id]+t+[self.tokenizer.sep_token_id]
+      c.append([q]+v[0:i]+[[self.tokenizer.mask_token_id]]+v[i+1:]+[[q[-1]]])
+    b=[[len(sum(x[0:j+1],[])) for j in range(len(x))] for x in c]
+    d=self.model(input_ids=torch.tensor([sum(x,[]) for x in c]),
+      token_type_ids=torch.tensor([[0]*x[0]+[1]*(x[-1]-x[0]) for x in b]))
+    s,e=d.start_logits.tolist(),d.end_logits.tolist()
+    for i in range(n):
+      for j in range(n):
+        m[i+1,0 if i==j else j+1]=s[i][b[i][j]]+e[i][b[i][j+1]-1]
+    h=ufal.chu_liu_edmonds.chu_liu_edmonds(m)[0]
+    if [0 for i in h if i==0]!=[0]:
+      i=([p for s,e,p in w]+["root"]).index("root")
+      j=i+1 if i<n else numpy.nanargmax(m[:,0])
+      m[0:j,0]=m[j+1:,0]=numpy.nan
+      h=ufal.chu_liu_edmonds.chu_liu_edmonds(m)[0]
+    u="# text = "+text.replace("\n"," ")+"\n"
+    for i,(s,e,p) in enumerate(w,1):
+      p="root" if h[i]==0 else "dep" if p=="root" else p
+      u+="\t".join([str(i),r[i-1],"_",z[s][0][2:],"_","|".join(z[s][1:]),
+        str(h[i]),p,"_","_" if i<n and w[i][0]<e else "SpaceAfter=No"])+"\n"
+    return u+"\n"
+
+nlp=TransformersUD("KoichiYasuoka/bert-ancient-chinese-base-ud-head")
+print(nlp("不入虎穴不得虎子"))
+```
+

config.json

@@ -0,0 +1,33 @@
+{
+  "architectures": [
+    "BertForQuestionAnswering"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "directionality": "bidi",
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-12,
+  "lstm_dropout_prob": 0.5,
+  "lstm_embedding_size": 768,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "pooler_fc_size": 768,
+  "pooler_num_attention_heads": 12,
+  "pooler_num_fc_layers": 3,
+  "pooler_size_per_head": 128,
+  "pooler_type": "first_token_transform",
+  "position_embedding_type": "absolute",
+  "tokenizer_class": "BertTokenizer",
+  "torch_dtype": "float32",
+  "transformers_version": "4.19.4",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 38208
+}

deprel/config.json

@@ -0,0 +1,170 @@
+{
+  "architectures": [
+    "BertForTokenClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "directionality": "bidi",
+  "finetuning_task": "pos",
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "B-acl",
+    "1": "B-advcl",
+    "2": "B-advmod",
+    "3": "B-amod",
+    "4": "B-aux",
+    "5": "B-case",
+    "6": "B-cc",
+    "7": "B-ccomp",
+    "8": "B-clf",
+    "9": "B-compound",
+    "10": "B-compound:redup",
+    "11": "B-conj",
+    "12": "B-cop",
+    "13": "B-csubj",
+    "14": "B-csubj:pass",
+    "15": "B-det",
+    "16": "B-discourse",
+    "17": "B-discourse:sp",
+    "18": "B-dislocated",
+    "19": "B-expl",
+    "20": "B-fixed",
+    "21": "B-flat",
+    "22": "B-flat:foreign",
+    "23": "B-flat:vv",
+    "24": "B-iobj",
+    "25": "B-list",
+    "26": "B-mark",
+    "27": "B-nmod",
+    "28": "B-nsubj",
+    "29": "B-nsubj:pass",
+    "30": "B-nummod",
+    "31": "B-obj",
+    "32": "B-obl",
+    "33": "B-obl:lmod",
+    "34": "B-obl:tmod",
+    "35": "B-orphan",
+    "36": "B-parataxis",
+    "37": "B-root",
+    "38": "B-vocative",
+    "39": "B-xcomp",
+    "40": "I-acl",
+    "41": "I-advcl",
+    "42": "I-advmod",
+    "43": "I-amod",
+    "44": "I-ccomp",
+    "45": "I-clf",
+    "46": "I-compound",
+    "47": "I-conj",
+    "48": "I-csubj",
+    "49": "I-dislocated",
+    "50": "I-flat",
+    "51": "I-flat:foreign",
+    "52": "I-iobj",
+    "53": "I-list",
+    "54": "I-nmod",
+    "55": "I-nsubj",
+    "56": "I-nsubj:pass",
+    "57": "I-nummod",
+    "58": "I-obj",
+    "59": "I-obl",
+    "60": "I-obl:lmod",
+    "61": "I-obl:tmod",
+    "62": "I-parataxis",
+    "63": "I-root",
+    "64": "I-vocative",
+    "65": "I-xcomp"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "B-acl": 0,
+    "B-advcl": 1,
+    "B-advmod": 2,
+    "B-amod": 3,
+    "B-aux": 4,
+    "B-case": 5,
+    "B-cc": 6,
+    "B-ccomp": 7,
+    "B-clf": 8,
+    "B-compound": 9,
+    "B-compound:redup": 10,
+    "B-conj": 11,
+    "B-cop": 12,
+    "B-csubj": 13,
+    "B-csubj:pass": 14,
+    "B-det": 15,
+    "B-discourse": 16,
+    "B-discourse:sp": 17,
+    "B-dislocated": 18,
+    "B-expl": 19,
+    "B-fixed": 20,
+    "B-flat": 21,
+    "B-flat:foreign": 22,
+    "B-flat:vv": 23,
+    "B-iobj": 24,
+    "B-list": 25,
+    "B-mark": 26,
+    "B-nmod": 27,
+    "B-nsubj": 28,
+    "B-nsubj:pass": 29,
+    "B-nummod": 30,
+    "B-obj": 31,
+    "B-obl": 32,
+    "B-obl:lmod": 33,
+    "B-obl:tmod": 34,
+    "B-orphan": 35,
+    "B-parataxis": 36,
+    "B-root": 37,
+    "B-vocative": 38,
+    "B-xcomp": 39,
+    "I-acl": 40,
+    "I-advcl": 41,
+    "I-advmod": 42,
+    "I-amod": 43,
+    "I-ccomp": 44,
+    "I-clf": 45,
+    "I-compound": 46,
+    "I-conj": 47,
+    "I-csubj": 48,
+    "I-dislocated": 49,
+    "I-flat": 50,
+    "I-flat:foreign": 51,
+    "I-iobj": 52,
+    "I-list": 53,
+    "I-nmod": 54,
+    "I-nsubj": 55,
+    "I-nsubj:pass": 56,
+    "I-nummod": 57,
+    "I-obj": 58,
+    "I-obl": 59,
+    "I-obl:lmod": 60,
+    "I-obl:tmod": 61,
+    "I-parataxis": 62,
+    "I-root": 63,
+    "I-vocative": 64,
+    "I-xcomp": 65
+  },
+  "layer_norm_eps": 1e-12,
+  "lstm_dropout_prob": 0.5,
+  "lstm_embedding_size": 768,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "pooler_fc_size": 768,
+  "pooler_num_attention_heads": 12,
+  "pooler_num_fc_layers": 3,
+  "pooler_size_per_head": 128,
+  "pooler_type": "first_token_transform",
+  "position_embedding_type": "absolute",
+  "tokenizer_class": "BertTokenizer",
+  "torch_dtype": "float32",
+  "transformers_version": "4.19.4",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 38208
+}

deprel/pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69e2b53116cbea8d40e7d0d33c23e21e7fa7954d8abc0eef70174acbf8afefee
+size 459454608

deprel/special_tokens_map.json

@@ -0,0 +1 @@
+{"unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]"}

deprel/tokenizer_config.json

@@ -0,0 +1 @@
+{"do_lower_case": true, "unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]", "tokenize_chinese_chars": true, "strip_accents": null, "do_basic_tokenize": true, "never_split": null, "special_tokens_map_file": null, "tokenizer_class": "BertTokenizer"}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:467e6868c713cca9f8de470c77e86349ee08a6c58eca584d1ccea735ee15fb07
+size 459254577

special_tokens_map.json

@@ -0,0 +1 @@
+{"unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]"}

tagger/config.json

@@ -0,0 +1,174 @@
+{
+  "architectures": [
+    "BertForTokenClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "directionality": "bidi",
+  "finetuning_task": "pos",
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "B-ADP|Degree=Equ",
+    "1": "B-ADP|_",
+    "2": "B-ADV|AdvType=Cau",
+    "3": "B-ADV|AdvType=Deg|Degree=Cmp",
+    "4": "B-ADV|AdvType=Deg|Degree=Pos",
+    "5": "B-ADV|AdvType=Deg|Degree=Sup",
+    "6": "B-ADV|AdvType=Tim",
+    "7": "B-ADV|AdvType=Tim|Aspect=Perf",
+    "8": "B-ADV|AdvType=Tim|Tense=Fut",
+    "9": "B-ADV|AdvType=Tim|Tense=Past",
+    "10": "B-ADV|AdvType=Tim|Tense=Pres",
+    "11": "B-ADV|Degree=Equ|VerbForm=Conv",
+    "12": "B-ADV|Degree=Pos|VerbForm=Conv",
+    "13": "B-ADV|Polarity=Neg",
+    "14": "B-ADV|Polarity=Neg|VerbForm=Conv",
+    "15": "B-ADV|VerbForm=Conv",
+    "16": "B-ADV|_",
+    "17": "B-AUX|Mood=Des",
+    "18": "B-AUX|Mood=Nec",
+    "19": "B-AUX|Mood=Pot",
+    "20": "B-AUX|VerbType=Cop",
+    "21": "B-AUX|Voice=Pass",
+    "22": "B-CCONJ|_",
+    "23": "B-INTJ|_",
+    "24": "B-NOUN|Case=Loc",
+    "25": "B-NOUN|Case=Tem",
+    "26": "B-NOUN|NounType=Clf",
+    "27": "B-NOUN|_",
+    "28": "B-NUM|NumType=Ord",
+    "29": "B-NUM|_",
+    "30": "B-PART|_",
+    "31": "B-PRON|Person=1|PronType=Prs",
+    "32": "B-PRON|Person=2|PronType=Prs",
+    "33": "B-PRON|Person=3|PronType=Prs",
+    "34": "B-PRON|PronType=Dem",
+    "35": "B-PRON|PronType=Int",
+    "36": "B-PRON|PronType=Prs",
+    "37": "B-PRON|PronType=Prs|Reflex=Yes",
+    "38": "B-PROPN|Case=Loc|NameType=Geo",
+    "39": "B-PROPN|Case=Loc|NameType=Nat",
+    "40": "B-PROPN|NameType=Giv",
+    "41": "B-PROPN|NameType=Prs",
+    "42": "B-PROPN|NameType=Sur",
+    "43": "B-SCONJ|_",
+    "44": "B-SYM|_",
+    "45": "B-VERB|Degree=Equ",
+    "46": "B-VERB|Degree=Equ|VerbForm=Part",
+    "47": "B-VERB|Degree=Pos",
+    "48": "B-VERB|Degree=Pos|VerbForm=Part",
+    "49": "B-VERB|Polarity=Neg",
+    "50": "B-VERB|Polarity=Neg|VerbForm=Part",
+    "51": "B-VERB|VerbForm=Part",
+    "52": "B-VERB|_",
+    "53": "I-ADV|VerbForm=Conv",
+    "54": "I-NOUN|Case=Loc",
+    "55": "I-NOUN|Case=Tem",
+    "56": "I-NOUN|_",
+    "57": "I-NUM|NumType=Ord",
+    "58": "I-NUM|_",
+    "59": "I-PROPN|Case=Loc|NameType=Geo",
+    "60": "I-PROPN|Case=Loc|NameType=Nat",
+    "61": "I-PROPN|NameType=Giv",
+    "62": "I-PROPN|NameType=Prs",
+    "63": "I-PROPN|NameType=Sur",
+    "64": "I-VERB|Degree=Equ",
+    "65": "I-VERB|Degree=Pos",
+    "66": "I-VERB|VerbForm=Part",
+    "67": "I-VERB|_"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "B-ADP|Degree=Equ": 0,
+    "B-ADP|_": 1,
+    "B-ADV|AdvType=Cau": 2,
+    "B-ADV|AdvType=Deg|Degree=Cmp": 3,
+    "B-ADV|AdvType=Deg|Degree=Pos": 4,
+    "B-ADV|AdvType=Deg|Degree=Sup": 5,
+    "B-ADV|AdvType=Tim": 6,
+    "B-ADV|AdvType=Tim|Aspect=Perf": 7,
+    "B-ADV|AdvType=Tim|Tense=Fut": 8,
+    "B-ADV|AdvType=Tim|Tense=Past": 9,
+    "B-ADV|AdvType=Tim|Tense=Pres": 10,
+    "B-ADV|Degree=Equ|VerbForm=Conv": 11,
+    "B-ADV|Degree=Pos|VerbForm=Conv": 12,
+    "B-ADV|Polarity=Neg": 13,
+    "B-ADV|Polarity=Neg|VerbForm=Conv": 14,
+    "B-ADV|VerbForm=Conv": 15,
+    "B-ADV|_": 16,
+    "B-AUX|Mood=Des": 17,
+    "B-AUX|Mood=Nec": 18,
+    "B-AUX|Mood=Pot": 19,
+    "B-AUX|VerbType=Cop": 20,
+    "B-AUX|Voice=Pass": 21,
+    "B-CCONJ|_": 22,
+    "B-INTJ|_": 23,
+    "B-NOUN|Case=Loc": 24,
+    "B-NOUN|Case=Tem": 25,
+    "B-NOUN|NounType=Clf": 26,
+    "B-NOUN|_": 27,
+    "B-NUM|NumType=Ord": 28,
+    "B-NUM|_": 29,
+    "B-PART|_": 30,
+    "B-PRON|Person=1|PronType=Prs": 31,
+    "B-PRON|Person=2|PronType=Prs": 32,
+    "B-PRON|Person=3|PronType=Prs": 33,
+    "B-PRON|PronType=Dem": 34,
+    "B-PRON|PronType=Int": 35,
+    "B-PRON|PronType=Prs": 36,
+    "B-PRON|PronType=Prs|Reflex=Yes": 37,
+    "B-PROPN|Case=Loc|NameType=Geo": 38,
+    "B-PROPN|Case=Loc|NameType=Nat": 39,
+    "B-PROPN|NameType=Giv": 40,
+    "B-PROPN|NameType=Prs": 41,
+    "B-PROPN|NameType=Sur": 42,
+    "B-SCONJ|_": 43,
+    "B-SYM|_": 44,
+    "B-VERB|Degree=Equ": 45,
+    "B-VERB|Degree=Equ|VerbForm=Part": 46,
+    "B-VERB|Degree=Pos": 47,
+    "B-VERB|Degree=Pos|VerbForm=Part": 48,
+    "B-VERB|Polarity=Neg": 49,
+    "B-VERB|Polarity=Neg|VerbForm=Part": 50,
+    "B-VERB|VerbForm=Part": 51,
+    "B-VERB|_": 52,
+    "I-ADV|VerbForm=Conv": 53,
+    "I-NOUN|Case=Loc": 54,
+    "I-NOUN|Case=Tem": 55,
+    "I-NOUN|_": 56,
+    "I-NUM|NumType=Ord": 57,
+    "I-NUM|_": 58,
+    "I-PROPN|Case=Loc|NameType=Geo": 59,
+    "I-PROPN|Case=Loc|NameType=Nat": 60,
+    "I-PROPN|NameType=Giv": 61,
+    "I-PROPN|NameType=Prs": 62,
+    "I-PROPN|NameType=Sur": 63,
+    "I-VERB|Degree=Equ": 64,
+    "I-VERB|Degree=Pos": 65,
+    "I-VERB|VerbForm=Part": 66,
+    "I-VERB|_": 67
+  },
+  "layer_norm_eps": 1e-12,
+  "lstm_dropout_prob": 0.5,
+  "lstm_embedding_size": 768,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "pooler_fc_size": 768,
+  "pooler_num_attention_heads": 12,
+  "pooler_num_fc_layers": 3,
+  "pooler_size_per_head": 128,
+  "pooler_type": "first_token_transform",
+  "position_embedding_type": "absolute",
+  "tokenizer_class": "BertTokenizer",
+  "torch_dtype": "float32",
+  "transformers_version": "4.19.4",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 38208
+}

tagger/pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e054845e4212ceb70ee2fc8ae172efa45b284aea5abd445b6f4fd07ba8f62a00
+size 459460755

tagger/special_tokens_map.json

@@ -0,0 +1 @@
+{"unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]"}

tagger/tokenizer_config.json

@@ -0,0 +1 @@
+{"do_lower_case": true, "unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]", "tokenize_chinese_chars": true, "strip_accents": null, "do_basic_tokenize": true, "never_split": null, "special_tokens_map_file": null, "tokenizer_class": "BertTokenizer"}

tokenizer_config.json

@@ -0,0 +1 @@
+{"do_lower_case": true, "unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]", "tokenize_chinese_chars": true, "strip_accents": null, "do_basic_tokenize": true, "never_split": null, "tokenizer_class": "BertTokenizer"}