A User-Profile Based Multi Agent Information Retrieval System Sadra Abedinzadeh MAS Research Group Faculty of Electrical and Computer Engineering School of Engineering, University of Tehran, Tehran, Iran s.abedinzadeh@ece.ut.ac.ir Reza Basseda MAS Research Group Faculty of Electrical and Computer Engineering School of Engineering, University of Tehran, Tehran, Iran r.basseda@ece.ut.ac.ir Kaveh Pashaei MAS Research Group Faculty of Electrical and Computer Engineering School of Engineering, University of Tehran, Tehran, Iran k.pashaei@ece.ut.ac.ir Fattaneh Taghiyareh Faculty of Electrical and Computer Engineering School of Engineering, University of Tehran, Tehran, Iran ftaghiyar@ut.ac.ir ABSTRACT There are many centralized approaches in the field of intelligent information retrieval; how- ever, because of great grow of the information stored in the collections, and also the web, an- swering the users’ queries in a shorter time and with a greater precision can be achieved using the decentralized approaches like parallel and distributed systems instead of centralized ones. Agent-based approaches can be used in designing, and implementing the distributed informa- tion retrieval systems. In this paper, we introduce a new approach for such a system in which agents are able to cooperate with their friends to answer the queries better. Key Words: Information Retrieval, Multi-Agent Systems 1. Introduction Today, the size of information resources like files, web pages, and etc. have grown numerously. As a result, the response time and also the performance of centralized I.R. systems has been decreased. Using parallel mechanisms in different parts of an IR system, like indexing, searching, match- ing, ranking, and etc., can handle this prob- lem. Agent-oriented approaches are a common way of performing the tasks in parallel mode. In these systems, there are several agents which together can lead the system to achieve its final goals, for example an- swering the users query in an IR system. In some systems, the agents are responsible for searching the documents space and they only work in documents domain. In other words, they do not have any interac- tion with users. They get the query, search the documents for relevant ones, and, fi- nally, return those relevant documents to the user. These types of agents are called “Library Agents”. The library agents store no information about the users; however, they use other features of agents, like pro- activeness, reactivity, social interactions,