Visualizing search results based on multi-label classification

Search engine has played an important role in information society. However, it is not very easy to find interest information from too much returned search results. Web search visualization system aims at helping users to locate interest documents rapidly from a great amount of returned search results. This paper explores visualization of Web search results based on multi-label text classification method. It conducts a multi-label classification process on the results from search engine. In this framework, users could browse interest information according to category label added by our algorithm. A paralleled Naïve Bayes multi-label classification algorithm is proposed for this application. A two-step feature selection algorithm is constructed to reduce the effect on Naïve Bayes classifier resulted from feature correlation and feature redundancy. A prototype system, named TJ-MLWC, is developed, which has the function of browsing search results by one or several categories.