Commit
·
106cf39
1
Parent(s):
7f31ee9
upload hubscripts/scicite_hub.py to hub from bigbio repo
Browse files- scicite.py +235 -0
scicite.py
ADDED
@@ -0,0 +1,235 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 loader for the SciCite dataset.
|
18 |
+
|
19 |
+
SciCite is a dataset of 11K manually annotated citation intents based on
|
20 |
+
citation context in the computer science and biomedical domains.
|
21 |
+
|
22 |
+
Some of the code in this module is based on the corresponding module in the
|
23 |
+
datasets library.
|
24 |
+
https://github.com/huggingface/datasets/blob/master/datasets/scicite/scicite.py
|
25 |
+
|
26 |
+
In the source schema, we follow the datasets implementation and replace
|
27 |
+
missing values.
|
28 |
+
TODO: Use standard BigBio missing values.
|
29 |
+
"""
|
30 |
+
|
31 |
+
import json
|
32 |
+
import os
|
33 |
+
from typing import Dict, List, Tuple
|
34 |
+
|
35 |
+
import datasets
|
36 |
+
import numpy as np
|
37 |
+
|
38 |
+
from .bigbiohub import text.features
|
39 |
+
from .bigbiohub import BigBioConfig
|
40 |
+
from .bigbiohub import Tasks
|
41 |
+
|
42 |
+
_LANGUAGES = ['English']
|
43 |
+
_PUBMED = False
|
44 |
+
_LOCAL = False
|
45 |
+
_CITATION = """\
|
46 |
+
@inproceedings{cohan:naacl19,
|
47 |
+
author = {Arman Cohan and Waleed Ammar and Madeleine van Zuylen and Field Cady},
|
48 |
+
title = {Structural Scaffolds for Citation Intent Classification in Scientific Publications},
|
49 |
+
booktitle = {Conference of the North American Chapter of the Association for Computational Linguistics},
|
50 |
+
year = {2019},
|
51 |
+
url = {https://aclanthology.org/N19-1361/},
|
52 |
+
doi = {10.18653/v1/N19-1361},
|
53 |
+
}
|
54 |
+
"""
|
55 |
+
|
56 |
+
_DATASETNAME = "scicite"
|
57 |
+
_DISPLAYNAME = "SciCite"
|
58 |
+
|
59 |
+
_DESCRIPTION = """\
|
60 |
+
SciCite is a dataset of 11K manually annotated citation intents based on
|
61 |
+
citation context in the computer science and biomedical domains.
|
62 |
+
"""
|
63 |
+
|
64 |
+
_HOMEPAGE = "https://allenai.org/data/scicite"
|
65 |
+
|
66 |
+
_LICENSE = 'License information unavailable'
|
67 |
+
|
68 |
+
_URLS = {
|
69 |
+
_DATASETNAME: "https://s3-us-west-2.amazonaws.com/ai2-s2-research/scicite/scicite.tar.gz",
|
70 |
+
}
|
71 |
+
|
72 |
+
_SUPPORTED_TASKS = [Tasks.TEXT_CLASSIFICATION]
|
73 |
+
|
74 |
+
_SOURCE_VERSION = "1.0.0"
|
75 |
+
|
76 |
+
_BIGBIO_VERSION = "1.0.0"
|
77 |
+
|
78 |
+
|
79 |
+
class SciciteDataset(datasets.GeneratorBasedBuilder):
|
80 |
+
"""SciCite is a dataset of 11K manually annotated citation intents based on
|
81 |
+
citation context in the computer science and biomedical domains."""
|
82 |
+
|
83 |
+
SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
|
84 |
+
BIGBIO_VERSION = datasets.Version(_BIGBIO_VERSION)
|
85 |
+
|
86 |
+
# You will be able to load the "source" or "bigbio" configurations with
|
87 |
+
# ds_source = datasets.load_dataset('scicite', name='source')
|
88 |
+
# ds_bigbio = datasets.load_dataset('scicite', name='bigbio')
|
89 |
+
|
90 |
+
BUILDER_CONFIGS = [
|
91 |
+
BigBioConfig(
|
92 |
+
name="scicite_source",
|
93 |
+
version=SOURCE_VERSION,
|
94 |
+
description="SciCite source schema",
|
95 |
+
schema="source",
|
96 |
+
subset_id="scicite",
|
97 |
+
),
|
98 |
+
BigBioConfig(
|
99 |
+
name="scicite_bigbio_text",
|
100 |
+
version=BIGBIO_VERSION,
|
101 |
+
description="SciCite BigBio schema",
|
102 |
+
schema="bigbio_text",
|
103 |
+
subset_id="scicite",
|
104 |
+
),
|
105 |
+
]
|
106 |
+
|
107 |
+
DEFAULT_CONFIG_NAME = "scicite_source"
|
108 |
+
|
109 |
+
def _info(self) -> datasets.DatasetInfo:
|
110 |
+
if self.config.schema == "source":
|
111 |
+
features = datasets.Features(
|
112 |
+
{
|
113 |
+
"source": datasets.Value("string"),
|
114 |
+
"citeStart": datasets.Value("int64"),
|
115 |
+
"sectionName": datasets.Value("string"),
|
116 |
+
"string": datasets.Value("string"),
|
117 |
+
"citeEnd": datasets.Value("int64"),
|
118 |
+
"label": datasets.features.ClassLabel(
|
119 |
+
names=["method", "background", "result"]
|
120 |
+
),
|
121 |
+
"label_confidence": datasets.Value("float"),
|
122 |
+
"label2": datasets.features.ClassLabel(
|
123 |
+
names=["supportive", "not_supportive", "cant_determine", "none"]
|
124 |
+
),
|
125 |
+
"label2_confidence": datasets.Value("float"),
|
126 |
+
"citingPaperId": datasets.Value("string"),
|
127 |
+
"citedPaperId": datasets.Value("string"),
|
128 |
+
"isKeyCitation": datasets.Value("bool"),
|
129 |
+
"id": datasets.Value("string"),
|
130 |
+
"unique_id": datasets.Value("string"),
|
131 |
+
"excerpt_index": datasets.Value("int64"),
|
132 |
+
}
|
133 |
+
)
|
134 |
+
elif self.config.schema == "bigbio_text":
|
135 |
+
features = text.features
|
136 |
+
else:
|
137 |
+
raise ValueError("Unrecognized schema: %s" % self.config.schema)
|
138 |
+
|
139 |
+
return datasets.DatasetInfo(
|
140 |
+
description=_DESCRIPTION,
|
141 |
+
features=features,
|
142 |
+
homepage=_HOMEPAGE,
|
143 |
+
license=str(_LICENSE),
|
144 |
+
citation=_CITATION,
|
145 |
+
)
|
146 |
+
|
147 |
+
def _split_generators(self, dl_manager) -> List[datasets.SplitGenerator]:
|
148 |
+
"""Returns SplitGenerators."""
|
149 |
+
urls = _URLS[_DATASETNAME]
|
150 |
+
data_dir = dl_manager.download_and_extract(urls)
|
151 |
+
|
152 |
+
return [
|
153 |
+
datasets.SplitGenerator(
|
154 |
+
name=datasets.Split.TRAIN,
|
155 |
+
gen_kwargs={
|
156 |
+
"filepath": os.path.join(data_dir, "scicite", "train.jsonl"),
|
157 |
+
"split": "train",
|
158 |
+
},
|
159 |
+
),
|
160 |
+
datasets.SplitGenerator(
|
161 |
+
name=datasets.Split.TEST,
|
162 |
+
gen_kwargs={
|
163 |
+
"filepath": os.path.join(data_dir, "scicite", "test.jsonl"),
|
164 |
+
"split": "test",
|
165 |
+
},
|
166 |
+
),
|
167 |
+
datasets.SplitGenerator(
|
168 |
+
name=datasets.Split.VALIDATION,
|
169 |
+
gen_kwargs={
|
170 |
+
"filepath": os.path.join(data_dir, "scicite", "dev.jsonl"),
|
171 |
+
"split": "dev",
|
172 |
+
},
|
173 |
+
),
|
174 |
+
]
|
175 |
+
|
176 |
+
def _generate_examples(self, filepath, split: str) -> Tuple[int, Dict]:
|
177 |
+
"""Yields examples as (key, example) tuples."""
|
178 |
+
|
179 |
+
with open(filepath, "r") as data_file:
|
180 |
+
examples = [json.loads(line) for line in data_file]
|
181 |
+
|
182 |
+
# Preprocesses examples
|
183 |
+
keys = set()
|
184 |
+
for example in examples:
|
185 |
+
# Fixes duplicate keys
|
186 |
+
if example["unique_id"] in keys:
|
187 |
+
example["unique_id"] = example["unique_id"] + "_duplicate"
|
188 |
+
else:
|
189 |
+
keys.add(example["unique_id"])
|
190 |
+
|
191 |
+
if self.config.schema == "source":
|
192 |
+
for example in examples:
|
193 |
+
yield str(example["unique_id"]), {
|
194 |
+
"string": example["string"],
|
195 |
+
"label": str(example["label"]),
|
196 |
+
"sectionName": str(example["sectionName"]),
|
197 |
+
"citingPaperId": str(example["citingPaperId"]),
|
198 |
+
"citedPaperId": str(example["citedPaperId"]),
|
199 |
+
"excerpt_index": int(example["excerpt_index"]),
|
200 |
+
"isKeyCitation": bool(example["isKeyCitation"]),
|
201 |
+
"label2": str(example.get("label2", "none")),
|
202 |
+
"citeEnd": _safe_int(example["citeEnd"]),
|
203 |
+
"citeStart": _safe_int(example["citeStart"]),
|
204 |
+
"source": str(example["source"]),
|
205 |
+
"label_confidence": float(
|
206 |
+
example.get("label_confidence", np.nan)
|
207 |
+
),
|
208 |
+
"label2_confidence": float(
|
209 |
+
example.get("label2_confidence", np.nan)
|
210 |
+
),
|
211 |
+
"id": str(example["id"]),
|
212 |
+
"unique_id": str(example["unique_id"]),
|
213 |
+
}
|
214 |
+
|
215 |
+
elif self.config.schema == "bigbio_text":
|
216 |
+
for example in examples:
|
217 |
+
if "label2" in example:
|
218 |
+
labels = [example["label"], example["label2"]]
|
219 |
+
else:
|
220 |
+
labels = [example["label"]]
|
221 |
+
|
222 |
+
yield str(example["unique_id"]), {
|
223 |
+
"id": example["unique_id"],
|
224 |
+
"document_id": example["citingPaperId"],
|
225 |
+
"text": example["string"],
|
226 |
+
"labels": labels,
|
227 |
+
}
|
228 |
+
|
229 |
+
|
230 |
+
def _safe_int(a):
|
231 |
+
try:
|
232 |
+
# skip NaNs
|
233 |
+
return int(a)
|
234 |
+
except ValueError:
|
235 |
+
return -1
|