Sentiment analysis based on CNN - LSTM hotel reviews

Abstract

Abstract In order to obtain the implied semantic information of hotel reviews for emotional analysis, the correlation between discontinuous words is ignored in the traditional convolutional neural network (CNN) emotional analysis. Therefore, a novel sentiment analysis method based on CNN - LSTM model is proposed. In this method, CNN is used to extract semantic features from hotel review texts, and LSTM is used to add sentence structure features to enhance deep semantic learning. This model improves the accuracy and F1 value on the CHN senticorp-HTOL-BA-6000 hotel review data set, and can better solve the task of text sentiment analysis and discover the emotional orientation of text information.