Query Generation for Personalized Tracking of Information

We describe in this paper a method for generating queries from both user’s profile description and users’s community of interest description. These queries are used for a monitoring task, so that those documents which are not already known by the user are retrieved. This method was implemented and integrated to the prototype of GALILEI, an open source Internet/Intranet project developed for GNU/Linux. GALILEI learns a user’s profile [1], corresponding to an area of interest, by analyzing documents assessed by the user. Then, communities of interests are derived from those profiles using a specific genetic algorithm. Both profiles and groups are described by a set of terms. We use these terms to construct queries which are submitted to search engines to retrieve yet unkown documents. We conducted some experiments in order to evaluate our approach. Preliminary results from these experiments show that the suggested method can achieve an appreciable retrieval effectiveness.