Sentiment Analysis of Reviews Based on Deep Learning Model

As known as opinion mining, sentiment analysis is a work by using the "natural language processing" method to find out the author's attitude, emotion or evaluation on certain topics. This paper using a dataset by Mass et al from its original Stanford AI Repository, and a commonly pre-processing method--word embedding, and establish a deep learning model for sentiment analysis. From the perspective of data analysis, learn about the movie preferences and cultural characteristics of audiences from domestic and foreign. In the experiment, we compared the performance of RNN, LSTM and GRU in natural language processing, and improve the efficiency and accuracy of sentiment analysis by a fusion model which integrating recurrent neural networks variant at the output of convolutional neural network.