ShkalikovOleh commited on
Commit
9bd4e59
1 Parent(s): a69690d

Add loader script

Browse files

This commit adds a loader script for the dataset which download data from the original GitHub repository and generate HF dataset

Files changed (1) hide show
  1. europarl_ner_loader.py +137 -0
europarl_ner_loader.py ADDED
@@ -0,0 +1,137 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+
3
+ """The HF Datasets adapter for Evaluation Corpus for Named Entity Recognition using Europarl"""
4
+
5
+ import datasets
6
+
7
+ _CITATION = """@inproceedings{agerri-etal-2018-building,
8
+ title = "Building Named Entity Recognition Taggers via Parallel Corpora",
9
+ author = "Agerri, Rodrigo and
10
+ Chung, Yiling and
11
+ Aldabe, Itziar and
12
+ Aranberri, Nora and
13
+ Labaka, Gorka and
14
+ Rigau, German",
15
+ editor = "Calzolari, Nicoletta and
16
+ Choukri, Khalid and
17
+ Cieri, Christopher and
18
+ Declerck, Thierry and
19
+ Goggi, Sara and
20
+ Hasida, Koiti and
21
+ Isahara, Hitoshi and
22
+ Maegaard, Bente and
23
+ Mariani, Joseph and
24
+ Mazo, H{\'e}l{\`e}ne and
25
+ Moreno, Asuncion and
26
+ Odijk, Jan and
27
+ Piperidis, Stelios and
28
+ Tokunaga, Takenobu",
29
+ booktitle = "Proceedings of the Eleventh International Conference on Language Resources and Evaluation ({LREC} 2018)",
30
+ month = may,
31
+ year = "2018",
32
+ address = "Miyazaki, Japan",
33
+ publisher = "European Language Resources Association (ELRA)",
34
+ url = "https://aclanthology.org/L18-1557",
35
+ }"""
36
+
37
+ _DESCRIPTION = """This dataset contains a gold-standard test set created from the
38
+ Europarl corpus. The test set consists of 799 sentences manually annotated using
39
+ four entity types and following the CoNLL 2002 and 2003 guidelines for 4 languages:
40
+ English, German, Italian and Spanish."""
41
+
42
+ _DATA_URLs = {
43
+ "en": "https://github.com/ixa-ehu/ner-evaluation-corpus-europarl/raw/master/en-europarl.test.conll02",
44
+ "de": "https://github.com/ixa-ehu/ner-evaluation-corpus-europarl/raw/master/de-europarl.test.conll02",
45
+ "es": "https://github.com/ixa-ehu/ner-evaluation-corpus-europarl/raw/master/es-europarl.test.conll02",
46
+ "it": "https://github.com/ixa-ehu/ner-evaluation-corpus-europarl/raw/master/it-europarl.test.conll02",
47
+ }
48
+ _HOMEPAGE = "https://github.com/ixa-ehu/ner-evaluation-corpus-europarl"
49
+ _VERSION = "1.0.0"
50
+ _LANGS = ["en", "de", "es", "it"]
51
+
52
+
53
+ class EuroparlNERConfig(datasets.BuilderConfig):
54
+ def __init__(self, **kwargs):
55
+ super(EuroparlNERConfig, self).__init__(
56
+ version=datasets.Version(_VERSION, ""), **kwargs
57
+ )
58
+
59
+
60
+ class EuroparlNER(datasets.GeneratorBasedBuilder):
61
+ """EuroparlNER is a multilingual named entity recognition dataset consisting of
62
+ manualy anotated part of the European Parliament Proceedings Parallel Corpus
63
+ 1996-2011 with LOC, PER, ORG and MISC tags"""
64
+
65
+ VERSION = datasets.Version(_VERSION)
66
+ BUILDER_CONFIGS = [
67
+ EuroparlNERConfig(
68
+ name=lang, description=f"EuroparlNER examples in language {lang}"
69
+ )
70
+ for lang in _LANGS
71
+ ]
72
+ DEFAULT_CONFIG_NAME = "en"
73
+
74
+ def _info(self):
75
+ features = datasets.Features(
76
+ {
77
+ "tokens": datasets.Sequence(datasets.Value("string")),
78
+ "ner_tags": datasets.Sequence(
79
+ datasets.features.ClassLabel(
80
+ names=[
81
+ "O",
82
+ "B-PER",
83
+ "I-PER",
84
+ "B-ORG",
85
+ "I-ORG",
86
+ "B-LOC",
87
+ "I-LOC",
88
+ "B-MISC",
89
+ "I-MISC",
90
+ ]
91
+ )
92
+ ),
93
+ }
94
+ )
95
+ return datasets.DatasetInfo(
96
+ description=_DESCRIPTION,
97
+ features=features,
98
+ supervised_keys=None,
99
+ homepage=_HOMEPAGE,
100
+ citation=_CITATION,
101
+ )
102
+
103
+ def _split_generators(self, dl_manager):
104
+ lang = self.config.name
105
+ dl_dir = dl_manager.download(_DATA_URLs[lang])
106
+
107
+ return [
108
+ datasets.SplitGenerator(
109
+ name=datasets.Split.TEST,
110
+ gen_kwargs={"filepath": dl_dir},
111
+ ),
112
+ ]
113
+
114
+ def _generate_examples(self, filepath):
115
+ guid_index = 1
116
+ with open(filepath, encoding="utf-8") as f:
117
+ tokens = []
118
+ ner_tags = []
119
+ for line in f:
120
+ if line == "" or line == "\n":
121
+ if tokens:
122
+ yield guid_index, {
123
+ "tokens": tokens,
124
+ "ner_tags": ner_tags,
125
+ }
126
+ guid_index += 1
127
+ tokens = []
128
+ ner_tags = []
129
+ else:
130
+ # EuroparlNER data is tab separated
131
+ splits = line.split("\t")
132
+ tokens.append(splits[0])
133
+ if len(splits) > 1:
134
+ ner_tags.append(splits[1].replace("\n", ""))
135
+ else:
136
+ # examples have no label in test set
137
+ ner_tags.append("O")