Domain-Specific Intelligent Search with Case-based BDI Agents

Cindy Olivia, Carlos F. Enguix, Chee-Fon Chang and Aditya K. Ghose Decision Systems Laboratory University of Wollongong, NSW, Australia 2522 Abstract This paper presents the Web CBR-BDI agent architecture for effective and intelligent real-time search on welldemarcated domains on the WWW. The proposed architecture is based upon the integration of case-based reasoning (CBR) with the BDI agent architecture. The approach demonstrates the intelligent search capabilities of such agent based on the ability to learn from previous cases stored in a case memory, coupled with a domainspecific knowledge-base. 1. Architecture of Web CBR-BDI Agents This research aims to demonstrate the efficacy of CBR-BDI agents in a well-demarcated domain by supporting the following features: • Efficient crawling: ability to reuse knowledge from previous traversals, avoiding redundant and exhaustive search • Focused search: utilization of domain-specific knowledge associated to well-demarcated logical domains on the WWW to support automated search guidance towards promising sites • Concept spanning ability: capability of retrieving concepts that span more than a single Web page. The proposed architecture (Olivia et al. 1999) not only accelerates the search process but also enhances resource discovery, finding a wider range of related pages that might not be retrieved using current Web-IR systems. Web CBR-BDI agents are designed to locate and extract topic-related information from target Web pages. The Web-CBR-BDI agent is implemented in a client-server architecture (fig. 1).