File size: 7,515 Bytes
5753b4a
 
 
ae98cf6
 
 
5753b4a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8222d6c
5753b4a
 
9ce19e5
ae98cf6
 
 
0fd9bd4
9ce19e5
5753b4a
 
c79a262
 
 
 
 
 
 
 
 
 
 
5753b4a
 
 
c79a262
c036ec5
8222d6c
9ce19e5
c79a262
5753b4a
c036ec5
8222d6c
5753b4a
 
 
 
 
c79a262
 
ae98cf6
 
c79a262
ae98cf6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
63cc6b6
 
ae98cf6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
68e94fa
ae98cf6
68e94fa
ae98cf6
 
 
 
 
 
 
5753b4a
 
 
c79a262
5753b4a
 
 
 
c79a262
ae98cf6
 
 
 
0fd9bd4
5a8f589
c7e902c
ae98cf6
0fd9bd4
 
 
 
 
 
 
9ce19e5
0fd9bd4
9ce19e5
0fd9bd4
9ce19e5
5753b4a
 
ae98cf6
75b410f
ae98cf6
 
 
a6b071e
 
 
 
 
 
 
 
 
 
 
5753b4a
9598a02
c7e902c
25ec43b
acd0763
 
74372b4
 
acd0763
c79a262
a6b071e
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
import datasets


logger = datasets.logging.get_logger(__name__)


_CITATION = """\
@InProceedings{mutinda2022pico,
  title = {PICO Corpus: A Publicly Available Corpus to Support Automatic Data Extraction from Biomedical Literature},
  author = {Mutinda, Faith and Liew, Kongmeng and Yada, Shuntaro and Wakamiya, Shoko and Aramaki, Eiji},
  booktitle = {Proceedings of the first Workshop on Information Extraction from Scientific Publications},
  pages = {26--31},
  year = {2022}
}
"""

_DESCRIPTION = """\
The corpus consists of about 1,011 PubMed abstracts which are RCTs related
to breast cancer. For each abstract, text snippets that identify the
Participants, Intervention, Control, and Outcome (PICO elements) are annotated.
The abstracts were annotated using BRAT (https://brat.nlplab.org/) and later
converted to IOB format.
"""

_URL = "https://raw.githubusercontent.com/Martin-Masson/pico-breast-cancer/main/pico_iob/"
_TRAINING_FILE = "train.txt"
_DEV_FILE = "dev.txt"
_TEST_FILE = "test.txt"
_ALL_FILE = "all.txt"
_SHORT_FILE = "all_short.txt"


class PicoBreastCancerConfig(datasets.BuilderConfig):
    """BuilderConfig for PicoBreastCancer"""

    def __init__(self, **kwargs):
        """BuilderConfig for PicoBreastCancer.
        Args:
          **kwargs: keyword arguments forwarded to super.
        """
        super(PicoBreastCancerConfig, self).__init__(**kwargs)


class PicoBreastCancer(datasets.GeneratorBasedBuilder):
    """A corpus of about 1,011 PubMed abstracts from RCTs related to breast cancer."""

    BUILDER_CONFIGS = [
        PicoBreastCancerConfig(name="split", version=datasets.Version("1.0.0"), description="Train/Validation/Test splits containing respectively 80%/10%/10% of all the abstracts."),
        PicoBreastCancerConfig(name="all", version=datasets.Version("1.0.0"), description="Single dataset containing all the absracts for custom splitting."),
        PicoBreastCancerConfig(name="short", version=datasets.Version("1.0.0"), description="Similar to 'all' but with shortened examples to account for embeddings size."),
    ]

    DEFAULT_CONFIG_NAME = "split"

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "id": datasets.Value("string"),
                    "tokens": datasets.Sequence(datasets.Value("string")),
                    "ner_tags": datasets.Sequence(
                        datasets.features.ClassLabel(
                            names=[
                                "O",
                                "B-total-participants",
                                "I-total-participants",
                                "B-intervention-participants",
                                "I-intervention-participants",
                                "B-control-participants",
                                "I-control-participants",
                                "B-age",
                                "I-age",
                                "B-eligibility",
                                "I-eligibility",
                                "B-ethinicity",
                                "I-ethinicity",
                                "B-condition",
                                "I-condition",
                                "B-location",
                                "I-location",
                                "B-intervention",
                                "I-intervention",
                                "B-control",
                                "I-control",
                                "B-outcome",
                                "I-outcome",
                                "B-outcome-measure",
                                "I-outcome-measure",
                                "B-iv-bin-abs",
                                "I-iv-bin-abs",
                                "B-cv-bin-abs",
                                "I-cv-bin-abs",
                                "B-iv-bin-percent",
                                "I-iv-bin-percent",
                                "B-cv-bin-percent",
                                "I-cv-bin-percent",
                                "B-iv-cont-mean",
                                "I-iv-cont-mean",
                                "B-cv-cont-mean",
                                "I-cv-cont-mean",
                                "B-iv-cont-median",
                                "I-iv-cont-median",
                                "B-cv-cont-median",
                                "I-cv-cont-median",
                                "B-iv-cont-sd",
                                "I-iv-cont-sd",
                                "B-cv-cont-sd",
                                "I-cv-cont-sd",
                                "B-iv-cont-q1",
                                "I-iv-cont-q1",
                                "B-cv-cont-q1",
                                "I-cv-cont-q1",
                                "B-iv-cont-q3",
                                "I-iv-cont-q3",
                                "B-cv-cont-q3",
                                "I-cv-cont-q3",
                            ]
                        )
                    ),
                }
            ),
            supervised_keys=None,
            homepage="https://github.com/Martin-Masson/pico-corpus",
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        """Returns SplitGenerators."""
        urls_to_download = {
            "train": f"{_URL}{_TRAINING_FILE}",
            "dev": f"{_URL}{_DEV_FILE}",
            "test": f"{_URL}{_TEST_FILE}",
            "all": f"{_URL}{_ALL_FILE}",
            "short": f"{_URL}{_SHORT_FILE}",
        }
        downloaded_files = dl_manager.download_and_extract(urls_to_download)
        if self.config.name == "split":
            return [
                datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": downloaded_files["train"]}),
                datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath": downloaded_files["dev"]}),
                datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": downloaded_files["test"]}),
            ]
        else:
            file = "all" if self.config.name == "all" else "short"
            return [
                datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": downloaded_files[file]}),
            ]
            

    def _generate_examples(self, filepath):
        logger.info("⏳ Generating examples from = %s", filepath)
        with open(filepath, encoding="utf-8") as f:
            idx = 0
            tokens = []
            ner_tags = []
            lines = f.read().splitlines()
            for line in lines:
                if not line:
                    yield idx, {
                        "id": str(idx),
                        "tokens": tokens,
                        "ner_tags": ner_tags,
                    }
                    idx += 1
                    tokens = []
                    ner_tags = []
                else:
                    splits = line.rstrip().rsplit(" ", 1)
                    tokens.append(splits[0])
                    ner_tags.append(splits[1])
            # last example
            if tokens:
                yield idx, {
                    "id": str(idx),
                    "tokens": tokens,
                    "ner_tags": ner_tags,
                }