Use of textual and conceptual profiles for personalized retrieval of political documents

The amount of information we are exposed to on a daily basis is increasing exponentially. Besides, Information Retrieval Systems (IRSs) return the same results for a given query regardless of who submitted it. In order to address the problems of finding useful, relevant information and adapting the results to the user, the use of personalization techniques is now more necessary than ever. They are not, however, particularly popular in live environments as users remain unconvinced about their reliability and, more importantly, are concerned about privacy issues. We have developed and compared six generic user profile representations in order to improve the personalization process and address the problem of privacy. We propose a new weighting scheme to build the profiles and a new personalization technique to join the advantages of some of the previous profiles. A comprehensive evaluation study of the proposed generic user profiles was performed and this revealed very good personalization performance results and some interesting conclusions about their use in a political context, more specifically with official documents from the Andalusian Parliament.