Text Sentiment Analysis Based On Long Short-term Memory

At present,some web like social networking sites, Internet e-commerce site, BBS or other sites, have already gotten rapid development. A large number of users can publish, share and updates their comments anytime and anytime. In a wide variety of information,goods review information is a very important category. The rapid development of electronic commerce, drive many customers used to online shopping. At the same time, the sites also provide comments on platform for consumers. Businesses through a series of analysis of comments information, provide better service to the customers. And the text sentiment analysis is a significant part to get essencial part.Understand the emotion trend of goods comments could improve the quality of the goods or the sales strategy, in order to meet the user’s preferences and promote the product distribution.Although the traditional Recurrent neural network model is theoretically possible to cover the temporal information of the whole sentence in the text, the practice shows that the RNN language model can not recognize longer sentences due to disappearance of gradients. In this paper, we try to improve the traditional RNN model by using the idea of Long Short-Tern Memory model.Since the LSTM model can preserve the information of the long text and play the advantage of the language model in the Chinese text emotion analysis, the LSTM model can be used as the language model to analyze the Chinese text, and LSTM language model can effectively access the text of the complete timing information,By adding the door structure,LSTM can remove or increase the ability of information exchange.Compared with the traditional RNN model, the LSTM model has a certain improvement in the experimental results. Finally, this paper integrates the LSTM model with the other two models to further improve the accuracy of the model for text emotion analysis.