holylovenia
commited on
Upload idner_news_2k.py with huggingface_hub
Browse files- idner_news_2k.py +219 -0
idner_news_2k.py
ADDED
@@ -0,0 +1,219 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# coding=utf-8
|
2 |
+
# Copyright 2022 The HuggingFace Datasets Authors and the current dataset script contributor.
|
3 |
+
#
|
4 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
5 |
+
# you may not use this file except in compliance with the License.
|
6 |
+
# You may obtain a copy of the License at
|
7 |
+
#
|
8 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
9 |
+
#
|
10 |
+
# Unless required by applicable law or agreed to in writing, software
|
11 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
12 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
13 |
+
# See the License for the specific language governing permissions and
|
14 |
+
# limitations under the License.
|
15 |
+
|
16 |
+
"""
|
17 |
+
A dataset of Indonesian News for Named-Entity Recognition task.
|
18 |
+
This dataset re-annotated the dataset previously provided by Syaifudin & Nurwidyantoro (2016)
|
19 |
+
(https://github.com/yusufsyaifudin/Indonesia-ner) with a more standardized NER tags.
|
20 |
+
There are three subsets, namely train.txt, dev.txt, and test.txt.
|
21 |
+
Each file consists of three columns which are Tokens, PoS Tag, and NER Tag respectively.
|
22 |
+
The format is following CoNLL dataset. The NER tag use the IOB format.
|
23 |
+
The PoS tag using UDPipe (http://ufal.mff.cuni.cz/udpipe),
|
24 |
+
a pipeline for tokenization, tagging, lemmatization and dependency parsing
|
25 |
+
whose model is trained on UD Treebanks.
|
26 |
+
"""
|
27 |
+
|
28 |
+
from pathlib import Path
|
29 |
+
from typing import Dict, List, Tuple
|
30 |
+
|
31 |
+
import datasets
|
32 |
+
import pandas as pd
|
33 |
+
|
34 |
+
from seacrowd.utils import schemas
|
35 |
+
from seacrowd.utils.configs import SEACrowdConfig
|
36 |
+
from seacrowd.utils.constants import Licenses, Tasks
|
37 |
+
|
38 |
+
_CITATION = """\
|
39 |
+
@inproceedings{khairunnisa-etal-2020-towards,
|
40 |
+
title = "Towards a Standardized Dataset on {I}ndonesian Named Entity Recognition",
|
41 |
+
author = "Khairunnisa, Siti Oryza and
|
42 |
+
Imankulova, Aizhan and
|
43 |
+
Komachi, Mamoru",
|
44 |
+
editor = "Shmueli, Boaz and
|
45 |
+
Huang, Yin Jou",
|
46 |
+
booktitle = "Proceedings of the 1st Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics
|
47 |
+
and the 10th International Joint Conference on Natural Language Processing: Student Research Workshop",
|
48 |
+
month = dec,
|
49 |
+
year = "2020",
|
50 |
+
address = "Suzhou, China",
|
51 |
+
publisher = "Association for Computational Linguistics",
|
52 |
+
url = "https://aclanthology.org/2020.aacl-srw.10",
|
53 |
+
pages = "64--71",
|
54 |
+
abstract = "In recent years, named entity recognition (NER) tasks in the Indonesian language
|
55 |
+
have undergone extensive development. There are only a few corpora for Indonesian NER;
|
56 |
+
hence, recent Indonesian NER studies have used diverse datasets. Although an open dataset is available,
|
57 |
+
it includes only approximately 2,000 sentences and contains inconsistent annotations,
|
58 |
+
thereby preventing accurate training of NER models without reliance on pre-trained models.
|
59 |
+
Therefore, we re-annotated the dataset and compared the two annotations{'} performance
|
60 |
+
using the Bidirectional Long Short-Term Memory and Conditional Random Field (BiLSTM-CRF) approach.
|
61 |
+
Fixing the annotation yielded a more consistent result for the organization tag and improved the prediction score
|
62 |
+
by a large margin. Moreover, to take full advantage of pre-trained models, we compared different feature embeddings
|
63 |
+
to determine their impact on the NER task for the Indonesian language.",
|
64 |
+
}
|
65 |
+
"""
|
66 |
+
|
67 |
+
_DATASETNAME = "idner_news_2k"
|
68 |
+
|
69 |
+
_DESCRIPTION = """\
|
70 |
+
A dataset of Indonesian News for Named-Entity Recognition task.
|
71 |
+
This dataset re-annotated the dataset previously provided by Syaifudin & Nurwidyantoro (2016)
|
72 |
+
(https://github.com/yusufsyaifudin/Indonesia-ner) with a more standardized NER tags.
|
73 |
+
There are three subsets, namely train.txt, dev.txt, and test.txt.
|
74 |
+
Each file consists of three columns which are Tokens, PoS Tag, and NER Tag respectively.
|
75 |
+
The format is following CoNLL dataset. The NER tag use the IOB format.
|
76 |
+
The PoS tag using UDPipe (http://ufal.mff.cuni.cz/udpipe),
|
77 |
+
a pipeline for tokenization, tagging, lemmatization and dependency parsing
|
78 |
+
whose model is trained on UD Treebanks.
|
79 |
+
"""
|
80 |
+
|
81 |
+
_HOMEPAGE = "https://github.com/khairunnisaor/idner-news-2k"
|
82 |
+
|
83 |
+
_LANGUAGES = ["ind"]
|
84 |
+
|
85 |
+
_LICENSE = Licenses.MIT.value
|
86 |
+
|
87 |
+
_LOCAL = False
|
88 |
+
|
89 |
+
_URLS = {
|
90 |
+
_DATASETNAME: {
|
91 |
+
"train": "https://raw.githubusercontent.com/khairunnisaor/idner-news-2k/main/train.txt",
|
92 |
+
"dev": "https://raw.githubusercontent.com/khairunnisaor/idner-news-2k/main/dev.txt",
|
93 |
+
"test": "https://raw.githubusercontent.com/khairunnisaor/idner-news-2k/main/test.txt",
|
94 |
+
},
|
95 |
+
}
|
96 |
+
|
97 |
+
_SUPPORTED_TASKS = [Tasks.NAMED_ENTITY_RECOGNITION]
|
98 |
+
|
99 |
+
_SOURCE_VERSION = "1.0.0"
|
100 |
+
|
101 |
+
_SEACROWD_VERSION = "2024.06.20"
|
102 |
+
|
103 |
+
|
104 |
+
class IdNerNews2kDataset(datasets.GeneratorBasedBuilder):
|
105 |
+
"""This dataset is designed for Named-Entity Recognition NLP task in Indonesian,
|
106 |
+
consisting of train, dev, and test files in CoNLL format. The NER tag in IOB format."""
|
107 |
+
|
108 |
+
SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
|
109 |
+
SEACROWD_VERSION = datasets.Version(_SEACROWD_VERSION)
|
110 |
+
|
111 |
+
SEACROWD_SCHEMA_NAME = "seq_label"
|
112 |
+
|
113 |
+
BUILDER_CONFIGS = [
|
114 |
+
SEACrowdConfig(
|
115 |
+
name=f"{_DATASETNAME}_source",
|
116 |
+
version=SOURCE_VERSION,
|
117 |
+
description=f"{_DATASETNAME} source schema",
|
118 |
+
schema="source",
|
119 |
+
subset_id=f"{_DATASETNAME}",
|
120 |
+
),
|
121 |
+
SEACrowdConfig(
|
122 |
+
name=f"{_DATASETNAME}_seacrowd_{SEACROWD_SCHEMA_NAME}",
|
123 |
+
version=SEACROWD_VERSION,
|
124 |
+
description=f"{_DATASETNAME} SEACrowd schema",
|
125 |
+
schema=f"seacrowd_{SEACROWD_SCHEMA_NAME}",
|
126 |
+
subset_id=f"{_DATASETNAME}",
|
127 |
+
),
|
128 |
+
]
|
129 |
+
|
130 |
+
DEFAULT_CONFIG_NAME = f"{_DATASETNAME}_source"
|
131 |
+
|
132 |
+
def _info(self) -> datasets.DatasetInfo:
|
133 |
+
NAMED_ENTITIES = ["B-LOC", "I-LOC", "B-ORG", "I-ORG", "B-PER", "I-PER", "O"]
|
134 |
+
POS_TAGS = ["PROPN", "AUX", "NUM", "NOUN", "ADP", "PRON", "VERB", "ADV", "ADJ", "PUNCT", "DET", "PART", "SCONJ", "CCONJ", "SYM", "X"]
|
135 |
+
|
136 |
+
if self.config.schema == "source":
|
137 |
+
features = datasets.Features(
|
138 |
+
{
|
139 |
+
"tokens": datasets.Sequence(datasets.Value("string")),
|
140 |
+
"pos_tags": datasets.Sequence(datasets.ClassLabel(names=POS_TAGS)),
|
141 |
+
"ner_tags": datasets.Sequence(datasets.ClassLabel(names=NAMED_ENTITIES)),
|
142 |
+
}
|
143 |
+
)
|
144 |
+
|
145 |
+
elif self.config.schema == f"seacrowd_{self.SEACROWD_SCHEMA_NAME}":
|
146 |
+
features = schemas.seq_label.features(NAMED_ENTITIES)
|
147 |
+
|
148 |
+
else:
|
149 |
+
raise ValueError(f"Invalid config: {self.config.name}")
|
150 |
+
|
151 |
+
return datasets.DatasetInfo(
|
152 |
+
description=_DESCRIPTION,
|
153 |
+
features=features,
|
154 |
+
homepage=_HOMEPAGE,
|
155 |
+
license=_LICENSE,
|
156 |
+
citation=_CITATION,
|
157 |
+
)
|
158 |
+
|
159 |
+
def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
|
160 |
+
|
161 |
+
urls = _URLS[_DATASETNAME]
|
162 |
+
train_path = dl_manager.download_and_extract(urls["train"])
|
163 |
+
dev_path = dl_manager.download_and_extract(urls["dev"])
|
164 |
+
test_path = dl_manager.download_and_extract(urls["test"])
|
165 |
+
|
166 |
+
return [
|
167 |
+
datasets.SplitGenerator(
|
168 |
+
name=datasets.Split.TRAIN,
|
169 |
+
gen_kwargs={
|
170 |
+
"filepath": train_path,
|
171 |
+
"split": "train",
|
172 |
+
},
|
173 |
+
),
|
174 |
+
datasets.SplitGenerator(
|
175 |
+
name=datasets.Split.TEST,
|
176 |
+
gen_kwargs={
|
177 |
+
"filepath": test_path,
|
178 |
+
"split": "test",
|
179 |
+
},
|
180 |
+
),
|
181 |
+
datasets.SplitGenerator(
|
182 |
+
name=datasets.Split.VALIDATION,
|
183 |
+
gen_kwargs={
|
184 |
+
"filepath": dev_path,
|
185 |
+
"split": "dev",
|
186 |
+
},
|
187 |
+
),
|
188 |
+
]
|
189 |
+
|
190 |
+
def _generate_examples(self, filepath: Path, split: str) -> Tuple[int, Dict]:
|
191 |
+
"""Yields examples as (key, example) tuples."""
|
192 |
+
|
193 |
+
df = pd.read_csv(filepath, delimiter=" ", header=None, skip_blank_lines=False)
|
194 |
+
if self.config.schema == "source":
|
195 |
+
tokens, pos_tags, ner_tags = [], [], []
|
196 |
+
|
197 |
+
for idx, row in df.iterrows():
|
198 |
+
if pd.isnull(row[0]):
|
199 |
+
if tokens:
|
200 |
+
yield idx, {"id": idx, "tokens": tokens, "pos_tags": pos_tags, "ner_tags": ner_tags}
|
201 |
+
tokens, pos_tags, ner_tags = [], [], []
|
202 |
+
else:
|
203 |
+
tokens.append(row[0])
|
204 |
+
pos_tags.append(row[1])
|
205 |
+
ner_tags.append(row[2])
|
206 |
+
|
207 |
+
elif self.config.schema == f"seacrowd_{self.SEACROWD_SCHEMA_NAME}":
|
208 |
+
tokens, ner_tags = [], []
|
209 |
+
|
210 |
+
for idx, row in df.iterrows():
|
211 |
+
if pd.isnull(row[0]):
|
212 |
+
if tokens:
|
213 |
+
yield idx, {"id": idx, "tokens": tokens, "labels": ner_tags}
|
214 |
+
tokens, ner_tags = [], []
|
215 |
+
else:
|
216 |
+
tokens.append(row[0])
|
217 |
+
ner_tags.append(row[2])
|
218 |
+
else:
|
219 |
+
raise ValueError(f"Invalid config: {self.config.name}")
|