Jean-Marc Andreoli, Uwe Borghoff, Remo Pareschi, Stefano Bistarello, Ugo Montanari, Francesca Rossi
Proc. of the AAAI'97 workshop on Constraints and Agents, July 27, 1997, Providence, RI, U.S.A., pp. 39-44
The irreversible trend toward decentralized information sources makes less and less adequate the current
approaches to search mostly based on a centralized index. The Constraint Based Knowledge Brokers (CBKB)
model is a first attempt to provide an efficient decentralized search engine, and it does so by exploiting the
power of constraints. One of the reasons of its efficiency comes from sending the information providers
requests based on very specific and pre-filtered information. This is done via a method which splits the scope
of the broker agents as the requests come. However, an exaggerated use of scope splitting may lead to an
unmanageable quantity of agents. Techniques to avoid such situations may be based on several approaches,
which we briefly describe in this paper, like scope merging based on anti-unification, organizing agents in
disjunctive structures, and limiting scope splitting on the basis of the capabilities of the knowledge bases.
Report number: