We have a small group here at Siemens looking at the problem of using
agents to do information retrieval tasks in large networks. In our
prototype system, each each user and each information repository is
represented by an agent, called a userbot and a corpusbot, respectively.
A userbot is the system's model of its user; over time,
it evolves into an expert on that user's preferences and tastes.
Similarly, a corpusbot evolves into an expert on the best way
to access the information contained within its resource.
Agents cooperate to support both browsing and explicit content-based
searches.
Userbots possess "scripts" --- arbitrary, parameterized program segments
written in a language that contains primitives for interacting
with the data environment --- that users can use to customize and extend
the functionality of the generic userbot. A userbot can
execute multiple scripts concurrently and individual scripts
can execute on multiple machines during the course of a single
invocation. Both scripts and profile information contained in one userbot
can be imported into another userbot, facilitating data discovery and
leveraging the costs associated with writing scripts.
Please contact me if you would like further information.
Ellen Voorhees
Siemens Corporate Research, Inc.
755 College Road East
Princeton, NJ 08540
USA
ellen@scr.siemens.com