Information Agents for Building Hyperlinks

In 13] we describe an architecture for building special-purpose agents for solving the information capture and access problem in large, multi-media data environments with uncertainty in data representation and interpretation. The key idea is to recognize the underlying structure that exists at various levels of gran-ularity in such environments, and to use it to build partial models (that can serve as indices) which encode information about content. In this paper we apply this computational paradigm to create and maintain a rich hierarchy of indices that link documents in an electronic repository of technical reports (TR) located over a wide-area network. We describe issues in the acquisition and use of these indices in answering a wide range of queries in the TR data environment.