A Multiagent Framework for Collaborative Conceptual Learning Using a Dempster-Shafer Belief System

In this paper, we describe a multiagent framework for collaborative conceptual earning using a Dempster-Shafer belief system in the domain of information retrieval. In our multiagent system, each agent maintains a database of documents, entertains different queries from its users, and thus learns a unique dictionary of concepts. Filed for each concept is a set of keywords collected from the documents supporting that concept. A document may be filed under various concepts and thus concepts may share keywords. This provides for a metric to evaluate the relevance of a document to a query. Our proposed work enables the query of an agent to be composed at a conceptual evel or expanded at a conceptual level, justified by a set of keywords, and to be learned by other agents through communications. The learned concepts are stored in each agent’s unique translation table. In this manner, agents are able to evolve independently their own knowledge while maintaining translation tables through collaborative learning to help sustain the information retrieval process.