The Recommendation Click Graph: Properties and Applications

摘要

Query recommendations help users to formulate better queries and to obtain the desired search results.Users' clicks on query recommendations contain a great deal of information about search intent,query ambiguity and search performance.We use query recommendation click information contained in search logs to construct a recommendation click graph.A directed edge in the graph connects the prior query and the clicked recommended query.By analyzing the graph,we develop methods for finding ambiguous queries and improving the search results.The experimental results show that our method for finding ambiguous queries is effective,and using recommendation click information can improve the search performance of ambiguous queries.