Art. 05 – Vol. 25 – No. 1 – 2015
Ropardo SRL, Sibiu – Department for Research and Development
Abstract: This paper presents a personalized search system for re-ranking web search results. The system is accessible from the web browser. By rating web pages, the user may create multiple profiles that can be used to personalize the search results provided by the search engine. A User Profile algorithm is used to learn the user’s preferences using a hierarchical weighted keywords model. The potential of such personalized system is evaluated through experiments. Furthermore, two Information Retrieval algorithms, LSI and FCRN, are implemented in order to evaluate potential improvements of the retrieval process, using data from the User Profile. A hybrid algorithm, between LSI and FCRN, is also proposed and evaluated. Through experiments, which use the performance of the search engine as a baseline, the benefits of the personalized search system are presented.
Keywords: Personalized Search, User Profile, Information Retrieval, Similarity Metrics, Singular Value Decomposition, Evaluation.