Planning to Gather Information

We describe Occam, a query planning algorithm that determines the best way to integrate data from dierent sources. As input, Occam takes a library of site descriptions and a user query. As output, Occam automatically generates one or more plans that encode alternative ways to gather the requested information. Occam has several important features: (1) it integrates both legacy systems and full relational databases with an ecient, domain-independent, query-planning algorithm, (2) it reasons about the capabilities of dierent information sources, (3) it handles partial goal satisfaction i.e., gathers as much data as possible when it can't gather exactly all that the user requested, (4) it is both sound and complete, (5) it is ecient. We present empirical results demonstrating Occam's performance on a variety of information gathering tasks.

• Type: Conference
• Year: 1996
• Where: AAAI
• Authors: Chung T. Kwok, Daniel S. Weld
• Keywords: AAAI 1996, Goal Satisfaction I.e, Intelligent Agents, Query Planning Algorithm, User Query