Part-Of-Speech Tagging for News Corpus Based on Deep Learning
Used Keras to implement POS tagging on Chinese news corpus
Compare different models, and explored different skills to improve model performance
Base model structure
Models comparison
| Model/strategy | Accuracy on test sets (descending order) |
|---|---|
| Bert | 0.9915 |
| Bi-LSTM with attention | 0.9878 |
| Bi-LSTM | 0.9867 |
| LSTM (Linear learning rate) | 0.9808 |
| LSTM (Step learning rate) | 0.9804 |
| LSTM | 0.9799 |
| RNN | 0.9787 |
| GRU | 0.9773 |
Confusion matrix
Functional interface
