Information Integration Agents

With the vast and expanding amount of interconnected information resources available today, the problem of how to pool them to obtain useful information is becoming more and more critical. Information gathering agents are computer systems that given a high level information request by a user use search, inference and knowledge to determine how to satisfy the request. Speaker presents an information gethering agent architecture called SIMS developed at the Information Sciences Institute (http://www.isi.edu) that uses the LOOMS knowledge representation scheme to build models of a number of distributed, autonomous, and heterogeneous information resources (such as data, databases and webpages). It treats a high level user request as a planning goal and uses artificial intelligence planning techniques along with some manually encoded heuristics to compute a good quality query-plan that can be efficiently executed to retrieve and integrate information from a number of diverse sources. I will also present the work that I have done to automatically learn the heuristics that can be used by the SIMS planner to efficiently compute high quality query resolution.