
ANNUAL PROGRESS REPORT AND PROGRAM PLAN
(beyond Cooperative Agreement specification)
PROJECT SUMMARY
DATE PREPARED: February 15, 1996
ORGANIZATION: The University of Michigan
PROJECT DIRECTOR:
Daniel E. Atkins
atkins@umich.edu
Telephone: 313-747-3576
Fax: 313-764-2475
TITLE OF EFFORT: University of Michigan Digital Library Project: An agent-based architecture for federating information collections, services, and users.
ACCESS INFORMATION: http://www.sils.umich.edu/UMDL/HomePage.html
OBJECTIVE:
The University of Michigan Digital Library (UMDL) is a large-scale effort to provide digital information collections and services for research and education, in university and high school environments. The objective is to define, implement, and evaluate in selected user environments an architecture for federating a large collection of heterogeneous digital information resources that are both physically and organizationally distributed.
APPROACH:
The approach to the above objective is to design an open, distributed system architecture where interacting software agents cooperate and compete to provide library services. The distributed architecture promotes modularity, flexibility, and incremental development, and accommodates diversity in current and future library environments. At the same time, distribution presents difficult problems in interoperability, coordination, search, and resource allocation. The UMDL architecture addresses these problems by dynamically forming agent teams to perform complex library tasks. A virtual economy of information goods and services directs resource allocation in the distributed system. Deployment of the UMDL in high school science classrooms provides a context for the design of specialized agent services and appropriate user interfaces.
PROGRESS:
In the first two years of the project, we wanted to establish the basic software system--the UMDL agent architecture--and get initial feedback from users on the usability and value of UMDL to them in getting work done. We have mostly met those goals now The agent architecture is running, and is ready for third parties to add either content (which will be done this year by high-school students) or to create new services (which will be done this year by students in various classes at the University). The agents in the architecture have met the goals of being locally controlled (semi autonomous), and having the ability to coordinate their actions to perform large-scale task (e.g., search for information which no single agent can do on its own).
The future activities will focus on populating the architecture with specific agents to provide a collection of services relevant to the educational user community. We will add the commerce substrate, and will continue to evolved a more integrated and transparent search and fusion of information resources. We will also continue to demonstrate interoperability with other DLI projects.
We have done well in getting UMDL into the schools; it is really there and being used by students and teachers, but we greatly underestimated the effort We have not gotten information up quickly enough for students to use, and more importantly, the interface to the system has not been well accepted; it is too hard to use. Based on this feedback, we are putting more effort into developing both new interface and new services to support students. We have also realized the need to radically rethink the way users will interact with digital libraries in the future. Based on this experience and with the addition of Professor George Furnas to the project, we will seek a more interactive interface combining browsing, search, and social filtering.
RECENT ACCOMPLISHMENTS:
- Agent architecture. We have designed, implemented, and demonstrated the basic communication protocols (and supporting "content languages") for the agent architecture. Complementing the communication protocols, we have designed, implemented, and demonstrated a variety of agent coordination mechanisms. These mechanisms allow the agents to coordinate their actions to perform value-added tasks (such as information retrieval). In addition, we have scaled the architecture by adding a variety of new services (e.g., thesaurus, notification/subscription) without having to change either the architecture or existing agents (e.g., our collection and user-interface agents).
- Addition of 180 external distributed collections and hundreds of journals.
- Production digital library use of technologies partially developed in the UMDL. Large scale deployment of Elsevier journals (1,100 from 1995 forward); 5,000 images from seven art museums; complete runs of twelve major journals in history, economics and ecology from the late 1800's to 1990 (over 1,000,000 scanned OCR and indexed pages);
- Deployment to public schools. Three schools in the Ann Arbor area are now using UMDL as an integrated part of their inquiry-based science curriculum. To support this work, we have trained teachers in how to use UMDL and have created a set of "curriculum units" (e.g., collections of related, high-school level articles and other types of information). In addition, we have received funding to deploy the UMDL in Detroit metropolitan area middle schools.
- Community impact. We have conducted over twelve interactive training sessions for students, teachers, media specialists, librarians, administrators and parents. These experiences are the driving force for the iterative design of the UMDL interface and collection development.
- Creation of a new advanced interface team under direction of George Furnas, a new faculty member at the School of Information.
- Interoperability. With Stanford, we have designed, implemented and demonstrated an interoperability mechanism for exchanging queries and records (e.g., US MARC) between library projects. We are able to "register" Stanford's digital library as a collection in our digital library, so that it is as accessible as any collection we maintain. This protocol is the basis for more elaborate and interesting interoperability mechanisms we hope to develop over the remaining years in the project. For example, we hope to have our agents "negotiate" usage rights and payment for information in Stanford's collection.
- Ontology development. We have established a "base" ontology for digital library intellectual work and library services. This ontology has defined what various standard cataloging concepts mean in the digital world. For example, we have identified a basic set of intellectual work types (e.g., text, sound, video, images) and how these relate to common concepts such as books (e.g., defining a "book" in a hyperlinked, fluid information world is a difficult task). The ontology will be available soon as a web site. We will also propose a working group to the D-LIB Forum on this topic.
PLANS:
- Design an authorization and security plan for UMDL. In addition, we plan to implement portions of the security plan in the UMDL.
- Integrate the NetBill payment system into the UMDL
* Implement market facilitators and integrate them with a set of UMDL agents. This goal depends on the development of a language to allow agents to describe goods (and thus markets) and the development of a negotiation protocol. Further, we will investigate the use of other mechanisms for resource allocation which will be complementary to the market facilitators.
- Develop a design for distributing the registry that includes incorporation of the CNRI Handle System
- Revise and update UMDL architecture design document.
- Extend interoperability protocols to include services, document types and negotiation over services.
- Develop search and retrieval capabilities (including ability to identify and add web-books to the UMDL; thesaurus development; collection query transformation and investigation of relevance feedback).
- Design and begin prototype implementation of a visually rich paradigm for information gathering and structuring.
- Revise user interface using user-centered iterative design process and continue deployment of new versions to Ann Arbor public high schools.
- Begin preparation for UMDL deployment on University of Michigan campus.
TECHNOLOGY TRANSITION, SHARING, PARTNERING:
- Interoperability with Stanford University. We have shared our Z39.50 client software with Stanford which has been incorporated into their digital library.
- New Publishing Partners: Ebsco, Cambridge University Press, Groliers (processing data received from them)
- Hewlett-Packard has contributed $450,000 in server equipment to UMDL.
- Virage, Inc. contributes content-based image retrieval software to UMDL.
- CMU NETBILL adoption.
- UMDL technology being used in Mellon JSTOR project, Cornell-UM Making of America Project, and being considered by the Digital Library Federation of the Library Congress and major academic libraries. It will also be used increasing for the UM campus digital library and academic outreach initiatives. Have also begun discussions with a European digital library consortia (ERCIM)
Return to UMDL Publications
Return to the Main Page
Comments or questions may be sent to: UMDL.INFO@umich.edu