Intelligent Agents in Digital Libraries
CS6210 Class Presentation
Qiu Long and Tok Wee Hyong
(qiul,tokwh}@comp.nus.edu.sg

Overview
Motivation
Intelligent Agents in action!
UMDL
Zuno
What’s the next step?

Motivation
Be able to handle large number of content providers
Rich collection of information that could be stored in digital libraries
Large collection spanning multiple distributed databases
Need to query information across several digital library
Need for a scalable digital library architecture
Remove single point of failure or bottleneck
Knowbots (Kahn, R and Cerf, V – 1988)
Primary medium of communication and interaction between various major components in the digital library

Intelligent Agents
Software components
Communication
Interoperable with other agents
Communicate with their peers via message exchanges
Autonomous (=> local or decentralized control)
Process knowledge and draw inferences
Reactive
Proactive
Adaptive

University of Michigan
Digital Library (UMDL)
Information space divided into 2 levels
Conspectus: Stores a set of collection descriptors
(E.g Search engine capabilities, document structure)
Collection: Stores the actual documents in its native format

UMDL – Intelligent Agents
An agent in UMDL
Represent an element of the digital library
Communicate and negotiate with other agents via KQML
Consists of 3 main types of agents
User Interface Agents (UIAs)
Collection Interface Agents (CIAs)
Mediator Agents

UMDL –Agents in Action!

ZunoDL (Zuno Digital Library)
Agent-based
Decentralized framework
Domain independence
WWW front-end
Information economy
Decentralized, dynamic supply & demand, efficiency desired
Consumer, producer and broker

ZunoDL: Agents
Producer Agents
Library Service Agents (LSAs)
Catalogue Agents(CAs)
Facilitator Agents
Search Agents(SAs)
Consumer Agents
User Interface Agents (UIAs)

ZunoDL System Structure

Conclusion
Digital Libraries and Agent-based System
UMDL
Zuno
Issues addressed
Distributed data & control
Lack of global viewpoint
Dynamic environment
Efficiency

Agent commonalities –
What’s the next Step?
Moving from task-oriented agents towards intelligent, learning agents
Continuous learning
Mimicking expert users
Improving on the exploration process
Security

Q&A