Constructing a Facet-Based and Personalized Just-in-Time Web Information Recommendation Application in a Multi-agent Environment

High precision is vital to the success of just-in-time information retrieval system. This paper attempts to improve it from two aspects: better understanding the user´s current need and providing a highly relevant information source. For the former, an algorithm that can model a user´s need in current context based on his behavior automatically is proposed, and for the latter, a mechanism is provided for the user to choose the information source he need by organizing all Web resources involved in a faceted way. All the algorithms and functions are encapsulated in several agents to constitute a multi-agent just-in-time Web information recommending application that help people write, and all these agent are constructed and managed by a multi-agent system middleware named ldquoMAGErdquo.