gabrielaltay commited on
Commit
cb56b49
1 Parent(s): 2f259f7

upload hubscripts/umnsrs_hub.py to hub from bigbio repo

Browse files
Files changed (1) hide show
  1. umnsrs.py +186 -0
umnsrs.py ADDED
@@ -0,0 +1,186 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ UMNSRS, developed by Pakhomov, et al., consists of 725 clinical term pairs whose semantic similarity and relatedness.
18
+ The similarity and relatedness of each term pair was annotated based on a continuous scale by having the resident touch
19
+ a bar on a touch sensitive computer screen to indicate the degree of similarity or relatedness. The Intraclass
20
+ Correlation Coefficient (ICC) for the reference standard tagged for similarity was 0.47, and 0.50 for relatedness.
21
+ Therefore, as suggested by Pakhomov and colleagues, the subset below consists of 401 pairs for the similarity set and
22
+ 430 pairs for the relatedness set which each have an ICC equal to 0.73.
23
+ """
24
+
25
+ from typing import Dict, List, Tuple
26
+
27
+ import datasets
28
+ import pandas as pd
29
+
30
+ from .bigbiohub import pairs_features
31
+ from .bigbiohub import BigBioConfig
32
+ from .bigbiohub import Tasks
33
+
34
+ _LANGUAGES = ['English']
35
+ _PUBMED = False
36
+ _LOCAL = False
37
+ _CITATION = """\
38
+ @inproceedings{pakhomov2010semantic,
39
+ title={Semantic similarity and relatedness between clinical terms: an experimental study},
40
+ author={Pakhomov, Serguei and McInnes, Bridget and Adam, Terrence and Liu, Ying and Pedersen, Ted and Melton, \
41
+ Genevieve B},
42
+ booktitle={AMIA annual symposium proceedings},
43
+ volume={2010},
44
+ pages={572},
45
+ year={2010},
46
+ organization={American Medical Informatics Association}
47
+ }
48
+ """
49
+
50
+ _DATASETNAME = "umnsrs"
51
+ _DISPLAYNAME = "UMNSRS"
52
+
53
+ _DESCRIPTION = """\
54
+ UMNSRS, developed by Pakhomov, et al., consists of 725 clinical term pairs whose semantic similarity and relatedness.
55
+ The similarity and relatedness of each term pair was annotated based on a continuous scale by having the resident touch
56
+ a bar on a touch sensitive computer screen to indicate the degree of similarity or relatedness.
57
+ The following subsets are available:
58
+ - similarity: A set of 566 UMLS concept pairs manually rated for semantic similarity (e.g. whale-dolphin) using a
59
+ continuous response scale.
60
+ - relatedness: A set of 588 UMLS concept pairs manually rated for semantic relatedness (e.g. needle-thread) using a
61
+ continuous response scale.
62
+ - similarity_mod: Modification of the UMNSRS-Similarity dataset to exclude control samples and those pairs that did not
63
+ match text in clinical, biomedical and general English corpora. Exact modifications are detailed in the paper (Corpus
64
+ Domain Effects on Distributional Semantic Modeling of Medical Terms. Serguei V.S. Pakhomov, Greg Finley, Reed McEwan,
65
+ Yan Wang, and Genevieve B. Melton. Bioinformatics. 2016; 32(23):3635-3644). The resulting dataset contains 449 pairs.
66
+ - relatedness_mod: Modification of the UMNSRS-Relatedness dataset to exclude control samples and those pairs that did
67
+ not match text in clinical, biomedical and general English corpora. Exact modifications are detailed in the paper
68
+ (Corpus Domain Effects on Distributional Semantic Modeling of Medical Terms. Serguei V.S. Pakhomov, Greg Finley,
69
+ Reed McEwan, Yan Wang, and Genevieve B. Melton. Bioinformatics. 2016; 32(23):3635-3644).
70
+ The resulting dataset contains 458 pairs.
71
+ """
72
+
73
+ _HOMEPAGE = "https://conservancy.umn.edu/handle/11299/196265/"
74
+
75
+ _LICENSE = 'Creative Commons Zero v1.0 Universal'
76
+
77
+ _BASE_URL = "https://conservancy.umn.edu/bitstream/handle/11299/196265/"
78
+
79
+ _URLS = {
80
+ "umnsrs_similarity": _BASE_URL + "UMNSRS_similarity.csv",
81
+ "umnsrs_relatedness": _BASE_URL + "UMNSRS_relatedness.csv",
82
+ "umnsrs_similarity_mod": _BASE_URL + "UMNSRS_similarity_mod449_word2vec.csv",
83
+ "umnsrs_relatedness_mod": _BASE_URL + "UMNSRS_relatedness_mod458_word2vec.csv",
84
+ }
85
+
86
+ _SUPPORTED_TASKS = [Tasks.SEMANTIC_SIMILARITY]
87
+
88
+ _SOURCE_VERSION = "1.0.0"
89
+
90
+ _BIGBIO_VERSION = "1.0.0"
91
+
92
+
93
+ class UmnsrsDataset(datasets.GeneratorBasedBuilder):
94
+ """UMNSRS, developed by Pakhomov, et al., contains clinical term pairs whose semantic similarity and
95
+ relatedness were scored by experts."""
96
+
97
+ SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
98
+ BIGBIO_VERSION = datasets.Version(_BIGBIO_VERSION)
99
+
100
+ BUILDER_CONFIGS = []
101
+
102
+ for subset in ["similarity", "relatedness"]:
103
+ for mod in ["_mod", ""]:
104
+ BUILDER_CONFIGS.append(
105
+ BigBioConfig(
106
+ name=f"umnsrs_{subset}{mod}_source",
107
+ version=SOURCE_VERSION,
108
+ description=f"UMNSRS {subset}{mod} source schema",
109
+ schema="source",
110
+ subset_id=f"umnsrs_{subset}{mod}",
111
+ )
112
+ )
113
+ BUILDER_CONFIGS.append(
114
+ BigBioConfig(
115
+ name=f"umnsrs_{subset}{mod}_bigbio_pairs",
116
+ version=BIGBIO_VERSION,
117
+ description=f"UMNSRS {subset}{mod} BigBio schema",
118
+ schema="bigbio_pairs",
119
+ subset_id=f"umnsrs_{subset}{mod}",
120
+ )
121
+ )
122
+
123
+ DEFAULT_CONFIG_NAME = "umnsrs_similarity_source"
124
+
125
+ def _info(self) -> datasets.DatasetInfo:
126
+
127
+ if self.config.schema == "source":
128
+ features = datasets.Features(
129
+ {
130
+ "mean_score": datasets.Value("float32"),
131
+ "std_score": datasets.Value("float32"),
132
+ "text_1": datasets.Value("string"),
133
+ "text_2": datasets.Value("string"),
134
+ "code_1": datasets.Value("string"),
135
+ "code_2": datasets.Value("string"),
136
+ }
137
+ )
138
+
139
+ elif self.config.schema == "bigbio_pairs":
140
+ features = pairs_features
141
+
142
+ return datasets.DatasetInfo(
143
+ description=_DESCRIPTION,
144
+ features=features,
145
+ homepage=_HOMEPAGE,
146
+ license=str(_LICENSE),
147
+ citation=_CITATION,
148
+ )
149
+
150
+ def _split_generators(self, dl_manager) -> List[datasets.SplitGenerator]:
151
+ """Returns SplitGenerators."""
152
+
153
+ urls = _URLS[self.config.subset_id]
154
+ filepath = dl_manager.download_and_extract(urls)
155
+
156
+ return [
157
+ datasets.SplitGenerator(
158
+ name=datasets.Split.TRAIN,
159
+ gen_kwargs={
160
+ "filepath": filepath,
161
+ },
162
+ )
163
+ ]
164
+
165
+ def _generate_examples(self, filepath) -> Tuple[int, Dict]:
166
+ """Yields examples as (key, example) tuples."""
167
+ print(filepath)
168
+ data = pd.read_csv(
169
+ filepath,
170
+ sep=",",
171
+ header=0,
172
+ names=["mean_score", "std_score", "text_1", "text_2", "code_1", "code_2"],
173
+ )
174
+
175
+ if self.config.schema == "source":
176
+ for id_, row in data.iterrows():
177
+ yield id_, row.to_dict()
178
+ elif self.config.schema == "bigbio_pairs":
179
+ for id_, row in data.iterrows():
180
+ yield id_, {
181
+ "id": id_,
182
+ "document_id": id_,
183
+ "text_1": row["text_1"],
184
+ "text_2": row["text_2"],
185
+ "label": row["mean_score"],
186
+ }