Query recommendation based on improved query flow graph

Abstract

Search engines contain many user search behaviours. Based on the idea of group wisdom, analysis of the search log in a search engine to recommend a query has attracted significant attention from researchers. Mining valid information from a search log has shown good utility for constructing a query flow graph for query recommendation. In a query flow graph, nodes representing queries are connected if they have the same search intention. In the process of query recommendation, user intention is associated both with current queries and with the clicked URL. The URL can serve as important information to locate user search intentions. However, a query flow graph only considers query information in the search log to determine the relation between the queries. Accordingly, a novel method based on the improved query flow graph is proposed, which expands the query flow graph by adding the clicked URL information and semantic information. The clicked URL information and semantic information in the query flow graph can make the next query closer to user search intentions. Empirical experiments are performed in accordance with AOL log, and the results confirm the effectiveness of our approach in suggesting queries. The results demonstrate that the performance of our query recommendation algorithm is superior to those of other algorithms.