On combining the knowledge of heterogeneous information repositories

The Internet facilitates access to a large amount of electronic information. However, in order to exploit the flood of information, sophisticated search facilities are needed which convert the inundation of electronic data coming from numerous sources into real knowledge. From this knowledge a whole range of users will benefit, from business people to casual surfers and shoppers on the Internet. Intelligent agents or knowledge brokers play a vital role in realizing this vision. This paper presents a framework for knowledge brokers who search for information which is potentially available but stored in a way not always foreseen how the information will be exploited. More striking, the paper presents an architectural framework where the user can retrieve and combine knowledge uniformly, irrespective of where or how the knowledge-representing information is stored. Lessons learned from a prototype implementation allow a discussion of shortcomings due to the emphasis of current information repositories and their interfaces, above all their poor support for knowledge combination and the difficulty of localizing the appropriate information repositories.