hipe2020 / hipe2020.py
davanstrien's picture
davanstrien HF staff
switch to using sentence segmented v1.4 datasets
0dc3436
# coding=utf-8
# Copyright 2022 HuggingFace Datasets Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# Lint as: python3
"""TODO"""
from datetime import datetime
from typing import Optional
import datasets
import re
_CITATION = """\
TODO
"""
_DESCRIPTION = """\
TODO
"""
_BASE_URL_TRAIN_DEV = "https://raw.githubusercontent.com/impresso/CLEF-HIPE-2020/master/data/v1.4/"
_URLs = {
"EN": {
"dev": _BASE_URL_TRAIN_DEV + "en/HIPE-data-v1.4-dev-en.tsv?raw=true"
}, # English only has dev
"DE": {
"dev": _BASE_URL_TRAIN_DEV + "de/HIPE-data-v1.4-dev-de.tsv?raw=true",
"train": _BASE_URL_TRAIN_DEV + "de/HIPE-data-v1.4-train-de.tsv?raw=true",
},
"FR": {
"dev": _BASE_URL_TRAIN_DEV + "fr/HIPE-data-v1.4-dev-fr.tsv?raw=true",
"train": _BASE_URL_TRAIN_DEV + "fr/HIPE-data-v1.4-train-fr.tsv?raw=true",
},
}
class HIPE2020Config(datasets.BuilderConfig):
"""BuilderConfig for HIPE2020"""
def __init__(self, data_urls,**kwargs):
"""BuilderConfig for HIPE2020.
Args:
**kwargs: keyword arguments forwarded to super.
"""
super(HIPE2020Config, self).__init__(**kwargs)
self.data_urls = data_urls
class HIPE2020(datasets.GeneratorBasedBuilder):
"""HIPE2020 dataset."""
BUILDER_CONFIGS = [
HIPE2020Config(
name="en",
data_urls=_URLs["EN"],
version=datasets.Version("1.0.0"),
description="HIPE dataset covering English",
),
HIPE2020Config(
name="de",
data_urls=_URLs["DE"],
version=datasets.Version("1.0.0"),
description="HIPE dataset covering German",
),
HIPE2020Config(
name="fr",
data_urls=_URLs["FR"],
version=datasets.Version("1.0.0"),
description="HIPE dataset covering French",
),
]
def _info(self):
return datasets.DatasetInfo(
description=_DESCRIPTION,
features=datasets.Features(
{
"id": datasets.Value("string"),
"tokens": datasets.Sequence(datasets.Value("string")),
"NE_COARSE_LIT": datasets.Sequence(
datasets.features.ClassLabel(
names=[
"O",
"B-comp",
"B-loc",
"B-org",
"B-pers",
"B-prod",
"B-time",
"I-loc",
"I-org",
"I-pers",
"I-prod",
"I-time",
"_",
]
)
),
"NE_COARSE_METO_tags": datasets.Sequence(
datasets.features.ClassLabel(
names=[
"O",
"B-loc",
"B-org",
"B-pers",
"B-prod",
"I-loc",
"I-org",
"I-pers",
"_",
]
)
),
"NE_FINE_LIT_tags": datasets.Sequence(
datasets.features.ClassLabel(
names=[
"O",
"B-comp.name",
"B-loc",
"B-loc.add.elec",
"B-loc.add.phys",
"B-loc.adm.nat",
"B-loc.adm.reg",
"B-loc.adm.sup",
"B-loc.adm.town",
"B-loc.fac",
"B-loc.oro",
"B-loc.phys.astro",
"B-loc.phys.geo",
"B-loc.phys.hydro",
"B-loc.unk",
"B-org",
"B-org.adm",
"B-org.ent",
"B-org.ent.pressagency",
"B-pers",
"B-pers.coll",
"B-pers.ind",
"B-pers.ind.articleauthor",
"B-prod",
"B-prod.doctr",
"B-prod.media",
"B-time",
"B-time.date.abs",
"I-loc",
"I-loc.add.elec",
"I-loc.add.phys",
"I-loc.adm.nat",
"I-loc.adm.reg",
"I-loc.adm.sup",
"I-loc.adm.town",
"I-loc.fac",
"I-loc.oro",
"I-loc.phys.geo",
"I-loc.phys.hydro",
"I-loc.unk",
"I-org",
"I-org.adm",
"I-org.ent",
"I-org.ent.pressagency",
"I-pers",
"I-pers.coll",
"I-pers.ind",
"I-pers.ind.articleauthor",
"I-prod",
"I-prod.doctr",
"I-prod.media",
"I-time",
"I-time.date.abs",
"_",
]
)
),
"NE_FINE_METO_tags": datasets.Sequence(
datasets.features.ClassLabel(
names=[
"O",
"B-loc",
"B-loc.adm.reg",
"B-loc.adm.town",
"B-loc.fac",
"B-loc.oro",
"B-org",
"B-org.adm",
"B-org.ent",
"B-pers.coll",
"B-pers.ind",
"B-prod.media",
"I-loc",
"I-loc.adm.reg",
"I-loc.fac",
"I-loc.oro",
"I-org",
"I-org.adm",
"I-org.ent",
"I-pers",
"I-pers.ind",
"_",
]
)
),
"NE_FINE_COMP_tags": datasets.Sequence(
datasets.features.ClassLabel(
names=[
"O",
"B-comp.demonym",
"B-comp.function",
"B-comp.name",
"B-comp.qualifier",
"B-comp.title",
"I-comp.demonym",
"I-comp.function",
"I-comp.name",
"I-comp.qualifier",
"I-comp.title",
"_",
]
)
),
"NE_NESTED_tags": datasets.Sequence(
datasets.features.ClassLabel(
names=[
"O",
"B-loc",
"B-loc.adm.nat",
"B-loc.adm.reg",
"B-loc.adm.sup",
"B-loc.adm.town",
"B-loc.fac",
"B-loc.oro",
"B-loc.phys.geo",
"B-loc.phys.hydro",
"B-org",
"B-org.adm",
"B-org.ent",
"B-pers.coll",
"B-pers.ind",
"B-prod.media",
"B-time.date.abs",
"I-loc",
"I-loc.adm.nat",
"I-loc.adm.reg",
"I-loc.adm.town",
"I-loc.fac",
"I-loc.oro",
"I-loc.phys.geo",
"I-loc.phys.hydro",
"I-org",
"I-org.adm",
"I-org.ent",
"I-pers.ind",
"_",
]
)
),
"NEL_LIT_ID": datasets.Sequence(datasets.Value("string")),
"NEL_METO_ID": datasets.Sequence(datasets.Value("string")),
"no_space_after": datasets.Sequence(datasets.Value("bool")),
"end_of_line": datasets.Sequence(datasets.Value("bool")),
"PySBDSegment":datasets.Sequence(datasets.Value("bool")),
"date": datasets.Value("timestamp[s]"),
"title": datasets.Value("string"),
"document_id": datasets.Value("string"),
}
),
supervised_keys=None,
homepage="TODO",
citation=_CITATION,
)
def _split_generators(self, dl_manager):
"""Returns SplitGenerators."""
downloaded_files = dl_manager.download_and_extract(self.config.data_urls)
if self.config.name != "en":
data_files = {
"train": downloaded_files["train"],
"dev": downloaded_files["dev"],
}
else:
data_files = {"dev": downloaded_files["dev"]}
if self.config.name == "en":
return [
datasets.SplitGenerator(
name=datasets.Split.VALIDATION,
gen_kwargs={"filepath": data_files["dev"]},
),
# datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": data_files["test"]}), # TODO add test splits
]
else:
return [
datasets.SplitGenerator(
name=datasets.Split.TRAIN,
gen_kwargs={"filepath": data_files["train"]},
),
datasets.SplitGenerator(
name=datasets.Split.VALIDATION,
gen_kwargs={"filepath": data_files["dev"]},
),
]
def _generate_examples(self, filepath):
date_re = re.compile(r"# date = (\d{4}-\d{2}-\d{02})")
title_re = re.compile(r"newspaper = (\w{3})")
document_id_re = re.compile(r"document_id = (.*)")
with open(filepath, encoding="utf-8") as f:
guid = 0
tokens = []
NE_COARSE_LIT_tags = []
NE_COARSE_METO_tags = []
NE_FINE_LIT_tags = []
NE_FINE_METO_tags = []
NE_FINE_COMP_tags = []
NE_NESTED_tags = []
NEL_LIT_ID = []
NEL_METO_ID = []
no_space_after = []
end_of_line = []
pysdbsegment = []
for line in f:
if line.startswith(
"TOKEN NE-COARSE-LIT NE-COARSE-METO NE-FINE-LIT NE-FINE-METO NE-FINE-COMP NE-NESTED NEL-LIT NEL-METO MISC"
):
continue
if line.startswith("#") or line == "\n":
date_match = re.search(date_re, line)
if date_match:
date = date_match.group(1)
date = datetime.strptime(date, "%Y-%m-%d")
title_match = re.search(title_re, line)
if title_match:
title = title_match.group(1)
document_id_match = re.search(document_id_re, line)
if document_id_match:
document_id = document_id_match.group(1)
if tokens:
yield guid, {
"id": str(guid),
"tokens": tokens,
"NE_COARSE_LIT": NE_COARSE_LIT_tags,
"NE_COARSE_METO_tags": NE_COARSE_METO_tags,
"NE_FINE_LIT_tags": NE_FINE_LIT_tags,
"NE_FINE_METO_tags": NE_FINE_METO_tags,
"NE_FINE_COMP_tags": NE_FINE_COMP_tags,
"NE_NESTED_tags": NE_NESTED_tags,
"NEL_LIT_ID": NEL_LIT_ID,
"NEL_METO_ID": NEL_METO_ID,
"no_space_after": no_space_after,
"end_of_line": end_of_line,
"PySBDSegment":pysdbsegment,
"date": date,
"title": title,
"document_id": document_id,
}
guid += 1
tokens = []
NE_COARSE_LIT_tags = []
NE_COARSE_METO_tags = []
NE_FINE_LIT_tags = []
NE_FINE_METO_tags = []
NE_FINE_COMP_tags = []
NE_NESTED_tags = []
NEL_LIT_ID = []
NEL_METO_ID = []
no_space_after = []
end_of_line = []
pysdbsegment = []
else:
# HIPE 2020 tokens are tab separated
splits = line.split(
"\t"
) # TOKEN NE-COARSE-LIT NE-COARSE-METO NE-FINE-LIT NE-FINE-METO NE-FINE-COMP NE-NESTED NEL-LIT NEL-METO MISC
tokens.append(splits[0])
NE_COARSE_LIT_tags.append(splits[1])
NE_COARSE_METO_tags.append(splits[2])
NE_FINE_LIT_tags.append(splits[3])
NE_FINE_METO_tags.append(splits[4])
NE_FINE_COMP_tags.append(splits[5])
NE_NESTED_tags.append(splits[6])
NEL_LIT_ID.append(splits[7])
NEL_METO_ID.append(splits[8])
misc = splits[-1]
is_space = "NoSpaceAfter" in misc
is_end_of_line = "EndOfLine" in misc
PySBDSegment = "PySBDSegment" in misc
no_space_after.append(is_space)
end_of_line.append(is_end_of_line)
pysdbsegment.append(PySBDSegment)
# last example
yield guid, {
"id": str(guid),
"tokens": tokens,
"NE_COARSE_LIT": NE_COARSE_LIT_tags,
"NE_COARSE_METO_tags": NE_COARSE_METO_tags,
"NE_FINE_LIT_tags": NE_FINE_LIT_tags,
"NE_FINE_METO_tags": NE_FINE_METO_tags,
"NE_FINE_COMP_tags": NE_FINE_COMP_tags,
"NE_NESTED_tags": NE_NESTED_tags,
"NEL_LIT_ID": NEL_LIT_ID,
"NEL_METO_ID": NEL_METO_ID,
"no_space_after": no_space_after,
"end_of_line": end_of_line,
"PySBDSegment":pysdbsegment,
"date": date,
"title": title,
"document_id": document_id,
}