Following is a brief summary of activities and progress of the University of Michigan Digital Library project from August - December, 1994. There are several researchers and teams working concurrently to build the digital library. This report has been divided into sub-sections which loosely correspond to these teams. Provisions have been made to maintain and update the web document which describes the activities and personnel involved in the project. The University of Michigan has also begun discussions with researchers, journalists, and community members interested in participating in the digital library project.
The architecture group began the development of the architectural specification of the UMDL. A draft of this specification (Version 1.0) is now available. The specification describes a general definition of types of software agents in the UMDL (including a description of their capabilities and detailed specifications of four prototypical agents discussed below) and a detailed specification of several protocols, using KQML, that define basic operations (behaviors) of agents necessary for rudimentary library operation .
To federate heterogeneous collections, we are developing a normative description of the information in these collections called the Conspectus. We have developed an initial draft of the Conspectus (Version 0.1) for two of our collections (Tulip collection and Blue Skies). The group of researchers developing the conspectus have begun examination of the definition, scope, physical architecture, and implementation of the conspectus. We are presently implementing Version 1.0 of the architecture specification and Version 0.1 of the Conspectus.
Working with the Intellectual Properties and Economics group, architectural requirements for managing intellectual property rights and economic resource allocation, such as security and authentication, are being examined. In addition, this group is currently evaluating billing mechanisms, designing protocols for transactions and economic negotiations among agents, and developing a prototype computational market model.
The mediation agent group has been working with NSF-UMDL teams to articulate more fully the expected activities of the mediating agents responsible for maintaining the conspectus (registration agent(s)) and for task processing (initially, query planning). This has led to refinements to the architectural specifications. This group has also been exploring the use of a locally developed agent architecture, UM-PRS, based on the procedural reasoning system, as a vehicle for the early development of mediating agents, and have begun a process of instantiating such agents.
The User Interface Agent working group has arrived at consensus regarding the submission of "Conspectus" and "collection" queries. The UIA solicits the "Conspectus query" from the end user and presents users with a form and pop-up menus into which they select entries to describe themselves, and previous use of the UMDL. The UIA submits the prospective-collections list to the CIA and displays the prospective-collections list or a profiled and clustered list to the end users. At the present time, the User Interface Agent working group is examining the handling of search results.
The Collection Interface Agents (CIA) group has considered a variety of issues relating to both short and long term implementation strategies for the NSF-UMDL. One important issue concerns which search engines (and corresponding query languages) should be supported by the UMDL. In the short run, we have decided to implement and support five different systems, including a local system (called FTL), WAIS, PAT, Blue Skys, and an SQL-type system. For the longer term, we are evaluating a wide variety of retrieval engines, with the view to selecting a subset of them for the UMDL. Furthermore, we are also investigating strategies for adding and testing the effectiveness of locally developed IR techniques in the local FTL system. A second issue concerns whether or not the UMDL should support searches in multiple, native query languages, or whether we should build a facility which would map among query languages. We have decided to connect to the native search engines via the CIA, but we are investigating the possibility of providing some sort of interlingua (or interlinguas) which would allow the system to translate among various query syntaxs.
This team, charged with making the NSF-UMDL available to end users, is pursuing two goals: to establish an "Internet culture" at the school and library sites, and to establish a framework for assessing the impact of the digital library on the user populations. Two local high schools and a local library have been selected as initial test sites. They are Community High School, Pioneer High School, and the Ann Arbor Public Library. Assessment of the technical and personnel requirements for each site has been completed, and discussions with librarians and schools teachers has to review training plans has begun. Equipment and networking is currently being deployed to these sites. Evaluation components include: the search engine, the interface, the information retrieved, the impact of the presence of the library, and changes in professional practice.
Several publishers have been contacted, and negotiations are underway to receive content to populate the digital library. Elsevier will begin delivery of content in December, 1994. McGraw Hill has delivered the Encyclopedia of Science and Technology on CD-ROM, and we are currently negotiating for additional content. Negotiations are being continued with UMI. In addition, initial discussions are underway with both Cambridge University Press and Groliers.
Presentations and discussions between the UMDL team and researchers at Xerox PARC and at Apple's Advanced Technology Group occurred in September. These exchanges provided an important link of the UMDL to research at both corporations. Ongoing discussions between individual corporate and university scientists are expected.
Bellcore researches have begun their activity with the Michigan project teams. Activities include agent and user interface research as well as assistance with the conspectus.
Equipment has been shipped by Apple and IBM.
Meetings with many of the corporate sponsors have been initiated with the goal of defining requests for additional resources for the UMDL project. These include equipment, cash, advanced technology, content and human resources. Specific discussions with the publishers about the evaluation and user studies components of the project have been begun.
Comments or questions may be sent to: UMDL.INFO@umich.edu