Adaptive multi-agent systems for information retrieval

The current exponential growth of the Internet precipitates a need for improved tools to help people cope with the volurne of information available. Existing search engines such, as Yahoo, Alta vista and Excite are efficient in terms of high recall (percentage of relevant document that are retrieved from Internet), and fast response time, at the cost of poor precision (percentage of documents retrieved that are considered relevant). The problem is due to the lack of filtering, lack of specialisation, lack of relevance feedback, lack of adaptation and lack of exploration. One solution for the above problems is to use intelligent agents, which can operate autonomously and become better over time. The agents rely on a user model to improve their performance in retrieving the information. This paper presents an adaptive information retrieval (IR) that learns from the user feedback through an evolutionary method, namely, genetic algorithms (GA).