Mining Product Reviews On The View

As today’s manufacturers and customers are paying more attention to product reviews, product reviews mining is becoming more significant. Generally, valuable information is ex-tracted manually by manufacturers and customers, which is extremely inefficient. By doing researches on automatic reviews mining, this dissertation shows a way to pick up featured words automatically and estimate the emotional tendency of the reviews at different granular-ities. Based on which, a reviews mining system is designed and proved in this dissertation. The process of the research is shown as below:1.According to customer reviews, this dissertation illustrates a self-extension method to extract features of products, which means to extract the products’ features that customers pre-ferred from reviews. In this algorithm, the goal of featured words mining is realized by seed words mining and modified by words distance.2. To resolve the problem of not including all the emotional details of text by traditional feature-opinion words, this dissertation defines an emotional description item on the base of feature-opinion words to express emotions in text. In addition, an algorithm of emotional de-scription item extracting is explained in this dissertation. It will dig out the emotional descrip-tion items in text based on featured words of products.3. Researches on the algorithm of text orientation at different granularities. This disserta-tion presents a way to estimate text overall orientation based on improved mutual information, which will make the text orientation estimation independent from emotional resources and therefore make the algorithm more applicable and more accurate. As to the emotional ten-dency estimation at granular level, this algorithm can find customers’ emotional tendency to-wards different products’ characters with concern of words similarity and emotional descrip-tion items.4. In order to realize automatic customer reviews mining, a customer-reviews-based re-views mining system is designed and proved based on the researches mentioned above and the requirements of reviews mining system customers. The system can extract products’ re-views from the Internet automatically and preprocess, find out the featured words from the comment text, estimate the emotional tendency of reviews as a whole or opinion words that related to some product features, and finally present the mining results intuitively.