Research And Implementaion Of Feature Based Sentiment Analysis

With the rise of Web2.0 and the rapid development of electronic commerce, shopping online has become a part of life of ordinary people.After purchasing, many cosummers tend to post their comments on the e-businness website to express their self experience and opinons about the commodity they have bought. These user reviews contain huge amounts of valuable information. For one thing, the reviews can guide other potential consumers to choose the right product. For another,manufactures can improve their products according to these comments so that to meet the market demand. However, because of the large amount of user reviews on the internet, it is unrealistic to obtain information from them manually. Hence, mining information from the review by using sentiment technology has become a popular research topic in rencent years.The traditional sentiment analysis methods fail to perform fine-grained sentiment anslysis on customer reviews. It can only provide overall sentiment tendencdies of a review. Based on previous researches,this paper focused on the method of feature based sentiment analysis and designed and implemented a sentiment analysis system to perform sentiment analysis of product features. This paper first introduces the related technology in sentiment analysis. Then we made the detail analysis and introduced overall architecture of this system. At last, we implemented the system and test it. Concretely, the main work of this paper includes following aspects:(1)By studying the existing methods of sentiment analysis, we proposed a novel method to perform feature based sentment analysis of product reviews. The algorithm base on association rules was used to extract product features from reviews. And LDA based model was proposed to judge sentiemnt tendency of the comments.(2) The feature based sentiemnt analysis method was used to implement a sentiment analysis system of product reviews. By using this system, feature level and comments level sentiment analysis can be done. Moreover, product comparision and feature search function have been added into the system so that make it closer to practical application. The system availability is vertified by system testing.The aspect level sentiment analysis method proposed in this paper has a good effect on attribute mining and sentiment classification. The sentiment analysis system built on the method can be used to feature based fentiment analysis. This system has strong practicability and broad application prospect.