Research On Sentiment Analysis And Its Application In Product’s Reviews Of E-commerce

With the development of information technology, the amount of user-generated content is increasing rapidly. Meanwhile, sentiment analysis is applied to automatically analyze the massive sentiment-orientation data from the Internet, so as to assist the users to better acquire and understand the information they need.This article conducts some deep research on sentiment analysis from both theoretical and applicable aspects, which includes:(1). Generative classification model affixed domain-specific sentiment lexicons.Based on the fact that sentiment knowledge is strongly domain-specific, we propose a generative sentiment classification model affixed domain-specific sentiment prior knowledge. Firstly, domain-specific sentiment lexicons, which are extracted from labeled sentiment oriented data, are used as seeds. Secondly, the sentimental seed lexicons set is extended from unlabeled data. Finally, a classification model is built incorporating such sentimental priors. The experiment results on standard datasets prove our method.(2). Design and implementation of a review analyze system (RAS).Under current sentiment analysis technology, we design and fulfill a prototype system which could analyze the product’s review data from e-commerce websites. Specifically, this system is composed of three modules:data crawling, data processing and visualization. Each module is designed independently, and owns standard inferences for data exchange, so as to ease the update and extension of the system.(3). Application of sentiment analysis.With the aid of multiple sentiment methods, we implement the data processing module of RAS. After preprocessing of review data (including sentence splitting, word segmentation and POS tagging), subjective classification is conducted on review sentences in the first step; then sentiment lexicons and aspects of product are extracted and the summary of reviews for each product is generated finally.