Disambiguation of Web Search Terms based on Clustering using Page Rank and Distance between Words

What is it about?

Web search is an indispensable activity in our daily lives. However, when users are not familiar with search operations, there is a problem that they cannot input appropriate search terms and cannot reach the desired information. In this study, we construct a system that extracts important words from search results, divides the words into several groups using clustering, and presents the groups to users. Experimental results show that clustering performance is improved by reflecting the rank of documents in the search results and the distance of words in the search results in judging the similarity between words. The rank weights of the proposed method showed somewhat better results than the conventional method for words that are polysemous but whose meanings are relatively easy to distinguish. As for the index weights, good results were obtained when the search phrase was an abbreviation of a word or when it was often accompanied by other words to form a noun phrase.

Featured Image