The Distributed Agent Architecture of the University of Michigan Digital Library

We provide an overview of a distributed-agent architecture design for the large-scale digital library environment currently under development at the University of Michigan. This paper presents some of the design principles and techniques underlying our approach, az well as some preliminary observations from our first steps in developing the system. a language and protocol for communicating about informational or processing capabilities and interests, so that interfaces and collections can be brought together appropriately. At that point, however, what they do with that connection (how they interact to accomplish their task) is not a concern at the level of our architecture. Overview: Diversity in the UMDL One of the most interesting and technically challenging features of digital libraries is diversity--in users, in information sources, and in many other features. For example, user skills, information demands, and level of professional competence will vary greatly among users of digital libraries. Similarly, the collections will vary over many parameters, including breadth and depth of subject, medium, and format. We expect that diversity will continue to increase and even accelerate, especially as digital libraries facilitate the spread of the publishing function from traditional publishing organizations to individuals. Diversity in large-scale information systems is also manifest across time. The types of tasks users perform and the tools available for performing these tasks are constantly evolving. Because the technology is changing at such a rapid pace, we believe that it is inadvisable to provide an all-encompassing set of standards for both user interfaces and collections. Thus, we do not advocate standards for describing and performing tasks, such as standard query languages or search engines. Instead, we have chosen to concentrate on defining an architecture that performs task-management tasks (meta-tasks), such as allocating resources to userrequested tasks and brokering connections among collaborating system modules. For example, we define Authors alphabetical. We gratefully acknowledge the assistance of many others participating the UMDL project, especially Gene Alloway. This work is supported in part by the National Science Foundation Digital Library Initiative. The advantage of avoiding general standards regarding information storage, retrieval, representation, or processing is that it makes incremental growth of our system both possible and appealing to developers. It certainly does not preclude the creation of partial standards, which is in itself a worthy goal, but it does not rely on their universal adoption. In fact, a goal of our endeavor is to provide an infrastructure where new capabilities--abilities to provide a better user interface, search more efficiently, translate among languages or representations, and so on--can be incorporated easily and can participate in tasks for which they are especially well suited without having to overhaul the existing components. The challenge that we face, therefore, is to provide an infrastucture that supports interoperability among potentially widely diverse collections, user interfaces, and other information providing and processing components. In building the University of Michigan Digital Library (UMDL) (Birmingham el. 1994), our approach is to distribute information-retrieval tasks to highly specialized agents. Large numbers of finegrained agents promote modularity, flexibility, and incrementality of the system, and provide a framework for tackling the interoperability problems. As new services become technologically feasible (e.g., information filters), agents can be added to the architecture; similarly, as new bodies of information become available we can likewise create new agents to manage these collections. From: AAAI Technical Report SS-95-08. Compilation copyright © 1995, AAAI (www.aaai.org). All rights reserved.