Enhancing the degree of personalization through Vector Space Model and Profile Ontology

Web browsers need to match the users' queries to the information available data bases. However, matching the users' needs with their interests and preferences to provide personalized search results in a ranked order of relevance entails a complex interaction of information attributes and as such, it remains one of the main challenges researchers face. Information Retrieval (IR) techniques focusing specifically on using Vector Space Model (VSM) with Profile Ontology (PO) hybridizationproved an improvement on personalized search results. We improve the degree of personalization by incorporating a new metric, the Dwell Time of each search session to optimize a learned re-ranked model. For a longitudinal naturalistic study of Web interactions, search logs were gathered as stimuli for the ranking algorithms of our personalized search engine. The performance of our re-ranking mechanism using Discounted Cumulative Gain (DCG) and F-measurewas tested. The scheme devised in this study was compared with the Google search engine. It was shown that, at the 10 top ranks of our personalized search engine, 14% improvement in the relevance is achieved.