| import datasets | |
| import pandas as pd | |
| import os | |
| class Vua20(datasets.GeneratorBasedBuilder): | |
| BUILDER_CONFIG_CLASS=datasets.BuilderConfig | |
| def _info(self): | |
| if self.config.name == 'default': | |
| feature = { | |
| 'id':datasets.Value('int32'), | |
| 'sent_id':datasets.Value('string'), | |
| 'word_index': datasets.Value('int32'), | |
| 'tokens':datasets.Sequence(datasets.Value('string')), | |
| 'label':datasets.ClassLabel(num_classes=2, names=['Not Metaphor', 'Is Metaphor']), | |
| 'pos':datasets.Value('string') | |
| } | |
| elif self.config.name == 'combined': | |
| feature = { | |
| 'sent_id':datasets.Value('string'), | |
| 'is_target': datasets.Sequence(datasets.ClassLabel(num_classes=2)), | |
| 'tokens':datasets.Sequence(datasets.Value('string')), | |
| 'labels': datasets.Sequence(datasets.ClassLabel(num_classes=2)), | |
| 'pos_tags':datasets.Sequence(datasets.Value('string')) | |
| } | |
| return datasets.DatasetInfo( | |
| description='Vua metaphor detection datasets.', | |
| features=datasets.Features(feature), | |
| config_name=self.config.name | |
| ) | |
| def _split_generators(self, dl_manager): | |
| return [ | |
| datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={'filepath': dl_manager.download('train.tsv')}), | |
| datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={'filepath': dl_manager.download('test.tsv')}), | |
| ] | |
| def _generate_examples(self, filepath): | |
| df = pd.read_csv(filepath, sep='\t') | |
| if self.config.name == 'default': | |
| for index, row in df.iterrows(): | |
| yield index, { | |
| 'id': index, | |
| 'sent_id': row['index'], | |
| 'tokens': row['sentence'].split(), | |
| 'word_index': row['w_index'], | |
| 'label': row['label'], | |
| 'pos': row['POS'] | |
| } | |
| elif self.config.name == 'combined': | |
| for index, (sent_id, group) in enumerate(df.groupby('index')): | |
| tokens = group.iloc[0]['sentence'].split() | |
| is_target = [1 if i in group.w_index.values else 0 for i in range(len(tokens))] | |
| labels = [1 if is_target[i] and group.label.iloc[sum(is_target[:i+1])-1] else 0 for i in range(len(tokens))] | |
| pos_tags = [group.POS.iloc[sum(is_target[:i+1])-1] if is_target[i] else '_' for i in range(len(tokens))] | |
| yield index, { | |
| 'sent_id': sent_id, | |
| 'tokens': tokens, | |
| 'is_target': is_target, | |
| 'labels': labels, | |
| 'pos_tags': pos_tags | |
| } |