Online Query Grouping With Collaborative Ranking Based On Search History

Users are depending on the web to pursue complex tasks and to achieve broader information. Search of complex tasks usually breaks down into co-dependent steps and issue multiple queries. Query Grouping is used to collect related queries which need common information. Query groups are used to support user in their long term information search. Online query groups are created in an automated and dynamic fashion. Keyword based similarity suffers from the problem of synonymic and polysomic queries. In time based similarity, related queries may not appear close to one another in search history. In KMeans algorithm, number of cluster and cluster means are decided initially. If data increases dynamically, there is no flexibility to increase clusters in K-Means algorithm. In the proposed system, the relevance algorithm resolves problem of polysomic and synonymic queries. Also related queries appear close to each other in query group. The proposed system resolves problem of K-Means algorithm by providing facility to increase the number of clusters (groups) if data increases. Relevance algorithm is based on a query fusion graph, where each edge represents either common clicks or consecutive count. In previous methodology, queries are added randomly into related group. In proposed system collaborative ranking is applied on each query. Each newly inserted query is added into its related groups according to its ranked relevance value. Keywords— Search history, query reformulation graph, query fusion graph, query click graph, query clustering, collaborative ranking.