|
import json |
|
|
|
import datasets |
|
|
|
|
|
logger = datasets.logging.get_logger(__name__) |
|
|
|
|
|
_DESCRIPTION = """\ |
|
Ukrainian News Dataset |
|
|
|
This is a dataset of news articles downloaded from various Ukrainian websites and Telegram channels. The dataset contains approximately ~23M JSON objects (news) |
|
""" |
|
|
|
|
|
_URLS = [ |
|
"ukrainian-news-vol1.jsonl", |
|
"ukrainian-news-vol2.jsonl", |
|
"ukrainian-news-vol3.jsonl", |
|
"ukrainian-telegram.jsonl" |
|
] |
|
|
|
|
|
class UkrainianNews(datasets.GeneratorBasedBuilder): |
|
"""Ukrainian News Dataset""" |
|
VERSION = datasets.Version("0.0.2") |
|
DEFAULT_CONFIG_NAME = "default" |
|
BUILDER_CONFIGS = [ |
|
datasets.BuilderConfig(name="default", version=VERSION, description=""), |
|
] |
|
|
|
def _info(self): |
|
return datasets.DatasetInfo( |
|
description=_DESCRIPTION, |
|
features=datasets.Features( |
|
{ |
|
"id": datasets.Value("int32"), |
|
"url": datasets.Value("string"), |
|
"title": datasets.Value("string"), |
|
"text": datasets.Value("string"), |
|
"owner": datasets.Value("string"), |
|
"datetime": datasets.Value("string"), |
|
} |
|
) |
|
) |
|
|
|
def _split_generators(self, dl_manager): |
|
downloaded_files = dl_manager.download(_URLS) |
|
|
|
return [ |
|
datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepaths": downloaded_files}) |
|
] |
|
|
|
def _generate_examples(self, filepaths): |
|
"""This function returns the examples in the raw (text) form.""" |
|
logger.info("generating examples from = %s", filepaths) |
|
for path in filepaths: |
|
with open(path, encoding="utf-8") as f: |
|
for news_str in f: |
|
news = json.loads(news_str) |
|
yield news['id'], news |
|
|