The Study Of Sentiment Analysis For Online Review Texts

With the rapid development of electronic commerce in recent years, more and more consumers have offered comments for the goods or services via online shopping, and the amount of comments keeps growing. The comments data are not only useful for consumers as reference information, but also instructive for merchants to improve the marketing strategies. Therefore, it is significant to conduct research on the text data of the comments. As an important technique for text analysis, the sentiment analysis has attracted extensive attentions of the researchers with many achievements. The main task of sentiment analysis is to determine the sentiment orientations of opinions. Typically,the sentiment analysis methods can be divided into two categories as below: sentiment lexicon-based methods and machine learning. The first method is easy to be implemented but lacks pertinence, the second one has better effects but dependents on a large number of training data.In the thesis, the background, state of arts and key techniques of sentiment analysis are introduced firstly. Secondly, the respective advantages and disadvantages of lexicon-based methods and machine learning methods are discussed. The Long Short Term Memory(LSTM) model is utilized as the classification model because of its effectiveness of deep learning in the machine learning. The sentiment analysis is performed by combining lexicon-based method with LSTM method. The experimental comparison shows that our new method, which utilizes both lexicon and LSTM, can generate the results with higher precision than the existing methods when there is a lack of training data.