chenyuxuan commited on
Commit
535f41f
1 Parent(s): 574afd3

Create wikigold.py

Browse files
Files changed (1) hide show
  1. wikigold.py +145 -0
wikigold.py ADDED
@@ -0,0 +1,145 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import datasets
2
+ from tqdm import tqdm
3
+
4
+ _CITATION = """
5
+ @inproceedings{balasuriya-etal-2009-named,
6
+ title = "Named Entity Recognition in Wikipedia",
7
+ author = "Balasuriya, Dominic and
8
+ Ringland, Nicky and
9
+ Nothman, Joel and
10
+ Murphy, Tara and
11
+ Curran, James R.",
12
+ booktitle = "Proceedings of the 2009 Workshop on The People{'}s Web Meets {NLP}:
13
+ Collaboratively Constructed Semantic Resources (People{'}s Web)",
14
+ month = aug,
15
+ year = "2009",
16
+ address = "Suntec, Singapore",
17
+ publisher = "Association for Computational Linguistics",
18
+ url = "https://aclanthology.org/W09-3302",
19
+ pages = "10--18",
20
+ }
21
+ """
22
+
23
+ _LICENCE = "CC-BY 4.0"
24
+
25
+ _DESCRIPTION = """
26
+ WikiGold dataset.
27
+ """
28
+
29
+ _URL = (
30
+ "https://github.com/juand-r/entity-recognition-datasets/raw/master/"
31
+ "data/wikigold/CONLL-format/data/wikigold.conll.txt"
32
+ )
33
+
34
+ # the label ids
35
+ NER_TAGS_DICT = {
36
+ "O": 0,
37
+ "PER": 1,
38
+ "LOC": 2,
39
+ "ORG": 3,
40
+ "MISC": 4,
41
+ }
42
+
43
+ NER_BIO_TAGS_DICT = {
44
+ "O": 0,
45
+ "B-PER": 1,
46
+ "I-PER": 2,
47
+ "B-LOC": 3,
48
+ "I-LOC": 4,
49
+ "B-ORG": 5,
50
+ "I-ORG": 6,
51
+ "B-MISC": 7,
52
+ "I-MISC": 8
53
+ }
54
+
55
+
56
+ class WikiGoldConfig(datasets.BuilderConfig):
57
+ """BuilderConfig for WikiGold"""
58
+
59
+ def __init__(self, **kwargs):
60
+ """BuilderConfig for WikiGold.
61
+ Args:
62
+ **kwargs: keyword arguments forwarded to super.
63
+ """
64
+ super(WikiGoldConfig, self).__init__(**kwargs)
65
+
66
+
67
+ class WikiGold(datasets.GeneratorBasedBuilder):
68
+ def _info(self):
69
+ return datasets.DatasetInfo(
70
+ description=_DESCRIPTION,
71
+ features=datasets.Features(
72
+ {
73
+ "id": datasets.Value("string"),
74
+ "tokens": datasets.features.Sequence(datasets.Value("string")),
75
+ "ner_tags": datasets.features.Sequence(
76
+ datasets.features.ClassLabel(
77
+ names=["O", "PER", "LOC", "ORG", "MISC"]
78
+ )
79
+ ),
80
+ "ner_bio_tags": datasets.features.Sequence(
81
+ datasets.features.ClassLabel(
82
+ names=["O", "B-PER", "I-PER", "B-LOC", "I-LOC",
83
+ "B-ORG", "I-ORG", "B-MISC", "I-MISC"]
84
+ )
85
+ ),
86
+ }
87
+ ),
88
+ supervised_keys=None,
89
+ citation=_CITATION,
90
+ license=_LICENCE,
91
+ )
92
+
93
+ def _split_generators(self, dl_manager):
94
+ """Returns SplitGenerators."""
95
+ urls_to_download = dl_manager.download_and_extract(_URL)
96
+ return [
97
+ datasets.SplitGenerator(
98
+ name=datasets.Split.TRAIN,
99
+ gen_kwargs={"filepath": urls_to_download},
100
+ ),
101
+ ]
102
+
103
+ def _generate_examples(self, filepath=None):
104
+ num_lines = sum(1 for _ in open(filepath))
105
+ id = 0
106
+
107
+ with open(filepath, "r") as f:
108
+ tokens, ner_tags, ner_bio_tags = [], [], []
109
+ for line in tqdm(f, total=num_lines):
110
+ line = line.strip().split()
111
+
112
+ if line:
113
+ assert len(line) == 2
114
+ token, ner_tag = line
115
+
116
+ if token == "-DOCSTART-":
117
+ continue
118
+
119
+ tokens.append(token)
120
+ ner_bio_tags.append(ner_tag)
121
+ if ner_tag != "O":
122
+ ner_tag = ner_tag.split("-")[1]
123
+ ner_tags.append(NER_TAGS_DICT[ner_tag])
124
+
125
+ elif tokens:
126
+ # organize a record to be written into json
127
+ record = {
128
+ "tokens": tokens,
129
+ "id": str(id),
130
+ "ner_tags": ner_tags,
131
+ "ner_bio_tags": ner_bio_tags,
132
+ }
133
+ tokens, ner_tags = [], []
134
+ id += 1
135
+ yield record["id"], record
136
+
137
+ # take the last sentence
138
+ if tokens:
139
+ record = {
140
+ "tokens": tokens,
141
+ "id": str(id),
142
+ "ner_tags": ner_tags,
143
+ "ner_bio_tags": ner_bio_tags,
144
+ }
145
+ yield record["id"], record