The aim of this task is to classify texts into Cantonese and Mandarin. It is extremely useful for filtering Cantonese text from large-scale web crawling-based corpus.
| Model | Author | Accuracy |
|---|---|---|
| Cantonese text classifier | CanCLID | 82.49% |
Please update this list if you have built your own model.
python compute_accuracy.py output.txtThe Cantonese test data are extracted from 粵語對話語料.
The Mandarin test data are extracted from PTT 八卦版問答中文語料.