Learning and modelling user interests using user feedback : A novel approach

Abstract

User profiles and interests have become essential for personalizing information search and retrieval. Indeed, traditional Information Retrieval Systems (IRS) don't integrate the user in the search process. Also, users do not always find what they need after a single query. Instead, they often issue multiple queries, incorporating what they learned from the previous results to iterate and refine how they express their information needs. So we rely on this process to learn the user information needs without asking him explicitly. This is achieved by capturing his judgments on the retrieved results. We consider also, in the construction of the user interests, what he is looking for and what the user doesn't want to find in the future results to build interests that best match his information needs.