Attention-based BiLSTM Network for Chinese Simile Recognition

Simile is a figure of speech that connects two different things with a common characteristic, which is expressed as "A is like B". In this paper, we focused on the Chinese Simile recognition which means classify the sentences between the simile and non-simile. In past works, models with higher classification results usually focus on the Ilarticular syntactic structures, while models with better generalization ability have lower classification accuracy. Therefore, our goal is not only to improve the classification results of the model, but also to take into account its generalization ability. We propose to use the Bidirectional LSTM network with attention mechanism to recognize the simile sentences and then obtained a higher Fl value. In order to better describe our model, we gives a detailed introduction to the bidirectional LSTM and attention mechanism works. In addition, we discussed the role of attention mechanism in this simile sentence recognition task.