Research On Mining Techniques Of Product Reviews Base On Semantic Analysis

In the past several years, with the rapid development of Web 2.0 technologies, People have been getting used to modern life styles, such as writing blogs, surfing in the forum, leaving comments and so on. Research on the product reviews shows that the majority of online shoppers tend to refer to the feedback reviews from other users, so that they can make better judgments. In the information-explosion society, it is of great importance finding out some efficient ways to extract useful information from the huge scale of data.Product reviews mining belongs to the scope of natural language analysis. Its main purpose is to find out user-interested aspects from the huge scale of data, classify the reviews according to the different aspects, and then derive the emotional coloring from reviews. In this paper, we mainly analyze the reviews of catering industry, and then make some research on the key technoques of product reviews mining system. The creative contributions of this paper are as follows:1) Present a new method of extracting aspects and clustering meaningful words from the huge scale of data. Firstly, probabilistic latent semantic analysis (PLSA) is applied in order to find the relationship between words and latent semantics and then calculate the similarity of words. With the help of semi-supervised method and clustering algorithm, words are clustered according to the different aspects.2) Present a new method of extracting emotional coloring from reviews. Subjective review sentences can be denoted by a series of feature-opinion pairs (F-O pairs). Our method assumes that the reviews are consistent with the corresponding scores. By using the scores and the clustering results, the emotional tendency of F-O pairs can be calculated automatically by the computer. After statistic processing of huge amount reviews, the tendency database is constructed. This method, which doesn t need any prior information, is of several advantages, such as simplicity, accuracy, efficiency and so on.To Sum up, based on PLSA algorithm, an aspect extracting and reviews classifying method is proposed. And we also do some research on extracting emotional coloring from reviews. In this paper, catering industry reviews is treated as the background of study and application. It s demonstrated by the experiments that the method is well performed and also has good engineering application value.