Clustering Method Based on Label Hierarchical Search Results
Boqin Feng · Xi'an Jiaotong Daxue xuebao · 2009
A novel clustering method based on hierarchical search results is proposed to facilitate users browsing web search results produced by search engines and to locate the interesting information quickly and efficiently.The snippets are collected and preprocessed.Frequent bigrams are identified based on term co-occurrence information,from which n-grams are obtained.After filtering out the redundant phrases and sorting by significance,candidate cluster labels are obtained.Finally,the snippets are grouped into clusters based on the candidate cluster labels,and a hierarchical result is generated.Experimental results show that the proposed method can generate accurate and highly readable cluster labels,which can help users effectively browse through the search results returned by search engine,and locating their interesting information.The method outperforms Vivisimo,Lingo and STC algorithms on different indexes.A comparison on Chinese dataset further illustrates the validity of the method.