Evolutionary reinforcement of user models in an adaptive search engine

The volume and variety of the Internet information is exponentially grows and therefore causes difficulties for a user to obtain information that accurately matches of the user interested. Several combination techniques are used to achieve the precise goal. This is due, firstly, to the fact that users often do not present queries to information retrieval systems that optimally represent the information they want, and secondly, the measure of a document´s relevance is highly subjective and variable between different users. We address this problem with an approach that relies on evolutionary user-modelling, in order to retrieve domain-specific information. We describe an adaptive information retrieval system that learns user needs from user-provided relevance feedback. The method combines qualitative feedback measures using fuzzy inference, and quantitative feedback using genetic algorithms (GA) fitness measures. We utilise the multiagent design approach for designing an information retrieval system (IRS). The system consists of following combination of complex processes: document indexing, learning strategic for relevant feedback and user modelling using genetic algorithm, filtering and ranking the retrieve documents based on the user model. We show the design of the IRS consists of several agents that cooperate with each other and may perform in parallel to achieve the system goal.