File size: 9,406 Bytes
dab57fa |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 |
import os
import datasets
from bs4 import ResultSet, BeautifulSoup
from datasets import DownloadManager
_CITATION = """\
@report{Magnini2021,
author = {Bernardo Magnini and Begoña Altuna and Alberto Lavelli and Manuela Speranza
and Roberto Zanoli and Fondazione Bruno Kessler},
keywords = {Clinical data,clinical enti-ties,corpus,multilingual,temporal information},
title = {The E3C Project:
European Clinical Case Corpus El proyecto E3C: European Clinical Case Corpus},
url = {https://uts.nlm.nih.gov/uts/umls/home},
year = {2021},
}
"""
_DESCRIPTION = """\
The European Clinical Case Corpus (E3C) project aims at collecting and \
annotating a large corpus of clinical documents in five European languages (Spanish, \
Basque, English, French and Italian), which will be freely distributed. Annotations \
include temporal information, to allow temporal reasoning on chronologies, and \
information about clinical entities based on medical taxonomies, to be used for semantic reasoning.
"""
_URL = "https://github.com/hltfbk/E3C-Corpus/archive/refs/tags/v2.0.0.zip"
class E3CConfig(datasets.BuilderConfig):
"""BuilderConfig for SQUAD."""
def __init__(self, **kwargs):
"""BuilderConfig for SQUAD.
Args:
**kwargs: keyword arguments forwarded to super.
"""
self.layer = kwargs.pop("layer")
super(E3CConfig, self).__init__(**kwargs)
class E3C(datasets.GeneratorBasedBuilder):
VERSION = datasets.Version("1.1.0")
BUILDER_CONFIGS = [
E3CConfig(
name="en",
version=VERSION,
description="this is the split of the layer 1 for English of E3C dataset",
layer="1",
),
E3CConfig(
name="es",
version=VERSION,
description="this is the split of the layer 1 for Spanish of E3C dataset",
layer="1",
),
E3CConfig(
name="eu",
version=VERSION,
description="this is the split of the layer 1 for Basque of E3C dataset",
layer="1",
),
E3CConfig(
name="fr",
version=VERSION,
description="this is the split of the layer 1 for French of E3C dataset",
layer="1",
),
E3CConfig(
name="it",
version=VERSION,
description="this is the split of the layer 1 for Italian of E3C dataset",
layer="1",
),
]
def _info(self):
"""This method specifies the DatasetInfo which contains information and typings."""
features = datasets.Features(
{
"tokens": datasets.Sequence(datasets.Value("string")),
"ner_tags": datasets.Sequence(
datasets.features.ClassLabel(
names=[
"O",
"CLINENTITY",
"EVENT",
"ACTOR",
"BODYPART",
"TIMEX3",
"RML",
],
),
),
}
)
return datasets.DatasetInfo(
description=_DESCRIPTION,
features=features,
citation=_CITATION,
supervised_keys=None,
)
def _split_generators(self, dl_manager: DownloadManager) -> list[datasets.SplitGenerator]:
"""Returns SplitGenerators who contains all the difference splits of the dataset.
Each language has its own split and each split has 3 different layers (sub-split):
- layer 1: full manual annotation of clinical entities, temporal information and
factuality, for benchmarking and linguistic analysis.
- layer 2: semi-automatic annotation of clinical entities
- layer 3: non-annotated documents
Args:
dl_manager: A `datasets.utils.DownloadManager` that can be used to download and
extract URLs.
Returns:
A list of `datasets.SplitGenerator`. Contains all subsets of the dataset depending on
the language and the layer.
"""
url = _URL
data_dir = dl_manager.download_and_extract(url)
language = {
"en": "English",
"es": "Spanish",
"eu": "Basque",
"fr": "French",
"it": "Italian",
}[self.config.name]
return [
datasets.SplitGenerator(
name=self.config.name,
gen_kwargs={
"filepath": os.path.join(
data_dir,
"E3C-Corpus-2.0.0/data_annotation",
language,
f"layer{self.config.layer}",
),
},
),
]
@staticmethod
def get_annotations(entities: ResultSet, text: str) -> list:
"""Extract the offset, the text and the type of the entity.
Args:
entities: The entities to extract.
text: The text of the document.
Returns:
A list of list containing the offset, the text and the type of the entity.
"""
return [
[
int(entity.get("begin")),
int(entity.get("end")),
text[int(entity.get("begin")) : int(entity.get("end"))],
]
for entity in entities
]
def get_parsed_data(self, filepath: str):
"""Parse the data from the E3C dataset and store it in a dictionary.
Iterate over the files in the dataset and parse for each file the following entities:
- CLINENTITY
- EVENT
- ACTOR
- BODYPART
- TIMEX3
- RML
for each entity, we extract the offset, the text and the type of the entity.
Args:
filepath: The path to the folder containing the files to parse.
"""
for root, _, files in os.walk(filepath):
for file in files:
with open(f"{root}/{file}") as soup_file:
soup = BeautifulSoup(soup_file, "xml")
text = soup.find("cas:Sofa").get("sofaString")
yield {
"CLINENTITY": self.get_annotations(
soup.find_all("custom:CLINENTITY"), text
),
"EVENT": self.get_annotations(soup.find_all("custom:EVENT"), text),
"ACTOR": self.get_annotations(soup.find_all("custom:ACTOR"), text),
"BODYPART": self.get_annotations(soup.find_all("custom:BODYPART"), text),
"TIMEX3": self.get_annotations(soup.find_all("custom:TIMEX3"), text),
"RML": self.get_annotations(soup.find_all("custom:RML"), text),
"SENTENCE": self.get_annotations(soup.find_all("type4:Sentence"), text),
"TOKENS": self.get_annotations(soup.find_all("type4:Token"), text),
}
def _generate_examples(self, filepath) -> tuple[str, dict]:
"""Yields examples as (key, example) tuples.
Args:
filepath: The path to the folder containing the files to parse.
Yields:
The unique id of an example and the example itself containing tokens and ner_tags in
IOB format.
"""
guid = 0
for content in self.get_parsed_data(filepath):
for sentence in content["SENTENCE"]:
filtered_tokens = list(
filter(
lambda token: token[0] >= sentence[0] and token[1] <= sentence[1],
content["TOKENS"],
)
)
labels = ["O"] * len(filtered_tokens)
for entity_type in [
"CLINENTITY",
"EVENT",
"ACTOR",
"BODYPART",
"TIMEX3",
"RML",
]:
if len(content[entity_type]) != 0 and sentence[1] >= content[entity_type][0][0]:
for entities in list(
filter(
lambda entity: sentence[0] <= entity[0] <= sentence[1],
content[entity_type],
)
):
annotated_tokens = [
idx_token
for idx_token, token in enumerate(filtered_tokens)
if token[0] >= entities[0] and token[1] <= entities[1]
]
for idx_token in annotated_tokens:
if idx_token == annotated_tokens[0]:
labels[idx_token] = f"{entity_type}"
else:
labels[idx_token] = f"{entity_type}"
guid += 1
yield guid, {
"tokens": list(map(lambda tokens: tokens[2], filtered_tokens)),
"ner_tags": labels,
}
|