from curses.ascii import isalpha import os import csv import re from typing import Sequence import json import ast import datasets _DESCRIPTION = """\ Example dataset toxic """ _DATA_URL = "https://drive.google.com/uc?id=1Ldnn3YYt_ErYq4ZGSon1MvcP3uJO0_PX" _DATA_ENG = "https://drive.google.com/uc?id=1p-iyKTRhUXaDmqsx69Zvb4ivjaCmVVr8" _TEXT = { "sen_vi": [" thất vọng", " bình thường", " hài lòng"], "sen_en": [" negative", " neutral", " positive"], "top_vi": [" giảng viên", " môn học", " phòng học", " tổng thể"], "top_en": [" lecturer", " curriculum", " facility", " general"], "top_en_": ["lecturer", "curriculum", "facility", "general"], "sen_en_": ["negative", "neutral", "positive"], "sen_vi_": ["thất vọng", "bình thường", "hài lòng"], "top_vi_": ["giảng viên", "môn học", "phòng học", "tổng thể"], } class Config(datasets.BuilderConfig): """BuilderConfig for GLUE.""" def __init__(self, data_url, **kwargs): """BuilderConfig Args: data_url: `string`, url to the dataset (word or raw level) **kwargs: keyword arguments forwarded to super. """ super(Config, self).__init__( version=datasets.Version( "1.0.0", ), **kwargs, ) self.data_url = data_url class Guess(datasets.GeneratorBasedBuilder): VERSION = datasets.Version("0.1.0") BUILDER_CONFIGS = [ Config( name="top_vi", data_url=_DATA_URL, description="data", ), Config( name="top_en", data_url=_DATA_ENG, description="data", ), Config( name="sen_vi", data_url=_DATA_URL, description="data", ), Config( name="sen_en", data_url=_DATA_ENG, description="data", ), Config( name="sen_en_", data_url=_DATA_ENG, description="data", ), Config( name="top_en_", data_url=_DATA_ENG, description="data", ), Config( name="top_vi_", data_url=_DATA_URL, description="data", ), Config( name="sen_vi_", data_url=_DATA_URL, description="data", ), ] def _info(self): # TODO(wikitext): Specifies the datasets.DatasetInfo object return datasets.DatasetInfo( # This is the description that will appear on the datasets page. description=_DESCRIPTION, # datasets.features.FeatureConnectors features=datasets.Features( { "text": datasets.Value("string"), "classes": datasets.Sequence(datasets.Value("string")), "target": datasets.Value("int8") } ), # If there's a common (input, target) tuple from the features, # specify them here. They'll be used if as_supervised=True in # builder.as_dataset. supervised_keys=None, ) def _split_generators(self, dl_manager): """Returns SplitGenerators.""" # TODO(wikitext): Downloads the data and defines the splits # dl_manager is a datasets.download.DownloadManager that can be used to # download and extract URLs data_file = dl_manager.download(self.config.data_url) return [ datasets.SplitGenerator( name=datasets.Split.TEST, gen_kwargs={"data_file": data_file, "type": self.config.name}, ), ] def _generate_examples(self, data_file, type): """Yields examples.""" # TODO(wikitext): Yields (key, example) tuples from the dataset with open(data_file, 'r') as f: lines = list(f) if type[:3] == 'sen': _CLASS = { "negative": 0, "neutral": 1, "positive": 2, } else: _CLASS = { "lecturer": 0, "curriculum": 1, "facility": 2, "others": 3 } TEXT_ = _TEXT[type] for idx, line in enumerate(lines): json_object = ast.literal_eval(line) if type[:3] == 'top': label = json_object['topic'] else: label = json_object['sentiment'] if label not in _CLASS: continue _text = json_object['text'] _classes = [] _PROMPT = { "sen_vi": f'{_text} Cảm thấy ', "sen_en": f'{_text} The sentiment of this sentence is ', "top_vi": f'Nói về ', "top_en": f'Comment about ', "sen_en_": f'{_text} The sentiment of this sentence is ', "top_en_": f'Comment about ', "sen_vi_": f'{_text} Cảm thấy ', "top_vi_": f'Nói về ', } for _cl in TEXT_: if type[:3] == 'sen': _classes.append(_cl) else: _classes.append(f'{_cl}. {_text}') yield idx, { "text" : _PROMPT[type], "classes" : _classes, "target" : _CLASS[label] }