Sentiment Analysis of Chinese Text Based on CNN-BiLSTM Serial Hybrid Model
Ping Huang, Lei Zheng, Ying Wang, Hui Zhu · 2021
Considering the forward and backward dependencies of words in Chinese sentences, this paper combines convolutional neural networks (CNN) and bi-directional long short-term memory (BiLSTM) according to a certain framework to form CNN- BiLSTM model, and uses this model to complete the task of emotional classification of Chinese texts. The model uses CNN to extract sentence features and then uses BiLSTM to capture two-way semantics to realize text sentiment classification. Through verification on Sina Weibo comments data set, CNN-BiLSTM can obtain better accuracy, recall and F1-score compared to CNN, RNN, LSTM and BiLSTM. It shows that CNN-BiLSTM can achieve better performance in the field of Chinese sentiment classification.