Research On Retrieval Method Based On Web Pages And Annotations Clustering

With the keeping development of the Internet, how to retrieve the information we focus from the huge Web resources, is becoming a key point to the research field. An effective search tool can play an important role in assisting the users to get the useful information easily.Firstly, this paper summarizes the existing Social Annotation Systems systematically, and analyses their merits and demerits. Secondly, it classifies and sets out the main clustering methods for web pages and tags, and also shows the merits and demerits of each method. Finally, it presents a new idea uses the Hypergraph Spectral Clustering method to cluster web pages and tags. Through the analysis and classification of the clustering results, the users could have a new list of search results which can play better than the Del.icio.us search results.The key point of this paper is to analyze the cluster results of the web pages and tags used the Hypergraph Spectral Clustering methods. At the meantime, this paper also compares the quality of the search results of K-means, Spectral Clustering, Ncut and Hypergraph Spectral Clustering method and demonstrates the Hypergraph Spectral Clustering method performance better than others whether in cluster precision or in the degree of relationship.In order to evaluate the effectiveness of the four cluster methods, this paper deign and compile a search system based on the Del.icio.us, focus on the web pages which have tagged on, called Search System based on the Clustering of Web pages and Tags (WTSS). WTSS uses Hypergraph Spectral Clustering method to cluster the results. The opinion of the present search method is not only based on the content of web pages but also on the interests of the public users. It is a union of social retrieval and traditional search engine. At last, the paper uses several kinds of methods to assess the return results of WTSS. The experiment shows that the methods which presented by this paper can satisfy the expect of the users for the search results'ranking.