Attention-based bidirectional LSTM for Chinese punctuation prediction

Punctuation prediction is an important task in Chinese automatic proofreading system which aims to tell whether the words and punctuations we use are right or not. The proofreading problem can be transformed into a prediction task. In this study, we propose an attention-based bidirectional Long Short-Term Memory (LSTM) model to predict punctuations. We use not only the sentence before the punctuation as the model's input, but also the sentence after the punctuation and the properties of the words. Experimental results show that the proposed LSTM model can achieve a good performance in punctuation prediction.