izhx/yue-openrice-review

Openrice Review Classification dataset

From github.com/Christainx/Dataset_Cantonese_Openrice.

The dataset includes 60k instances from Cantonese reviews in Openrice.

The rating ranks from 1-star (very negative) to 5-star (very positive). The instances are shuffled in order to disperse reviews of same restaurant.

Code for the splits creation

import datasets

def load_openrice():
    # https://github.com/Christainx/Dataset_Cantonese_Openrice/blob/master/Openrice_Cantonese.7z
    with open('Openrice_Cantonese.txt') as file:
        for i, line in enumerate(file):
            label = int(line[0])
            text = line[1:].strip()
            yield {'text': text, 'label': label}


ds = datasets.Dataset.from_generator(load_openrice)
print(ds)
dsd = ds.train_test_split(0.1, seed=42)
dsd['test'].to_json('data/test.jsonl', orient='records', force_ascii=False)
dsd['train'].to_json('data/train.jsonl', orient='records', force_ascii=False)
print(dsd)

Citation

@inproceedings{xiang2019sentiment,
  title={Sentiment Augmented Attention Network for Cantonese Restaurant Review Analysis},
  author={Xiang, Rong and Jiao, Ying and Lu, Qin},
  booktitle={Proceedings of the 8th KDD Workshop on Issues of Sentiment Discovery and Opinion Mining (WISDOM)},
  pages={1--9},
  year={2019},
  organization={KDD WISDOM}
}