A Multi-Layer Neural Network Model Integrating BiLSTM and CNN for Chinese Sentiment Recognition
Shanliang Yang, Qi Long Sun, Huyong Zhou, Zhengjie Gong · 2018
Technology of artificial intelligent has become research focus. Natural language understanding (NLU) is regarded as core technology of AI. Sentiment recognition is a difficult task in NLU; however it is advantageous to business market and public opinion analysis. We proposed a multi-layer neural network model through integrating LSTM and CNN to improve the performance of sentiment recognition. The structure of LSTM is appropriate to storage text sequence information, and CNN has ability to extract salient features for sentiment recognition task. We implemented models of LSTM-CNN and BiLSTM-CNN, and conduct experiments on different dataset. In the end, we contrast our proposed method with certain baseline methods. The result shows that the proposed method outperforms single layer model and other statistic learnint method.