Sentiment Classification Incorporating User Profile

With the emergence of the Internet social shopping platform, a large quantity of sentiment corpus is accumulating rapidly. Sentiment classification, which is a specific application of sentiment analysis, has received a lot of attention from researchers in the fields of natural language processing. The traditional method to classify sentiment text is usually limited to the content of text. However, the corpus on the Internet is attached with a large amount of additional information now, which may contribute to sentiment classification. Based on that situation, this paper develops a sentiment classification model which synthesizes text content and user profile to improve the classification accuracy. Concretely, we first calculate the similarity of both reviewers and content between reviews, then use label propagation algorithm to predict sentiment polarity of unlabeled reviews based on labeled reviews. The sentiment classification model incorporating user profile proved to have a higher accuracy compared to models exploiting only text content.