kinopoisk.py

import datasets
import pandas as pd

class KinopoiskReviewsConfig(datasets.BuilderConfig):
 def __init__(self, features, **kwargs):
 super(KinopoiskReviewsConfig, self).__init__(version=datasets.Version("1.0.0", ""), **kwargs)
 self.features = features

class Kinopoisk(datasets.GeneratorBasedBuilder):
 BUILDER_CONFIGS = [
 KinopoiskReviewsConfig(
 name="simple",
 description="Simple config",
 features=["content", "title", "grade3", "movie_name", "part", "review_id", "author", "date"],
 )
 ]
 
 def _info(self):
 #features = {feature: datasets.Value("string") for feature in self.config.features}
 #if self.config.name == "simple":
 # features = {feature: datasets.Value("string") for feature in self.config.features}
 #features["Idx"] = datasets.Value("int32")
 return datasets.DatasetInfo(
 description='Kinopoisk movie reviews dataset.',
 features=datasets.Features(
 {
 "content": datasets.Value("string"),
 "title": datasets.Value("string"),
 "grade3": datasets.Value("string"),
 "movie_name": datasets.Value("string"),
 "part": datasets.Value("string"),
 "review_id": datasets.Value("string"),
 "author": datasets.Value("string"),
 "date":datasets.Value("string"),
 "grade10": datasets.Value("string"),
 }
 ),
 supervised_keys=None,
 homepage='',
 citation='',
 )
 
 def _split_generators(self, dl_manager: datasets.DownloadManager):
 urls_to_download = self._URLS
 downloaded_files = dl_manager.download_and_extract(urls_to_download)
 
 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"]}),
 ]
 
 def _generate_examples(self, filepath):
 df = pd.read_json(filepath, lines=True)
 rows = df.to_dict(orient="records")
 
 for n, row in enumerate(rows):
 example = row
 example["Idx"] = n

 yield example["Idx"], example