University of Michigan Digital Library Project
Quarterly Report 3
Daniel E. Atkins, Project Director
November 1995

Architecture

The major activities of the architecture group during the last quarter have been to develop and implement the Michigan/Stanford/Illinois interoperability objects; incorporating new agents developed by the Conspectus Search and Retrieval (CSR) and User Interface Agent (UIA) groups, and rebuilding parts of the prototype system.

We have developed, in conjunction with Stanford, interfaces to allow the exchange of information between the two digital libraries. Details are given below in the User Interface Agent report.

In terms of incorporating new agents, we now have running a variety of agents that expand the services provided by UMDL. The agents communicate using CORBA (ILU) as the transfer protocol and a pseudo-KQML performative list as the language. The protocols used by these agents have been updated to be compatible with the underlying transport mechanisms now being used to support the interoperability work.

The agents that we are running include the following types:

Further details about the agents, and the registry (a "yellow pages" for the UMDL) are described in the Conspectus Search and Retrieval report.

All the agents in this list are functioning and communicating, but are still being tested and debugged. Both the conspectus and several individual collections are searchable. There are on the order of 10 searchable collections and 100 web sites registered in the system. The system can search across multiple collections and can interoperate with Stanford's system to search Dialog.

Conspectus Search And Retrieval

Over the past quarter, our emphasis has been on generating an initial prototype of the agent-based architecture for identifying and retrieving collections based on user needs. We have concentrated on developing robust and scalable components of our architecture that allow us to have end-to-end functionality for a restricted type of collection retrieval (focusing, for now, on topical search). Our activities have included:

The Conspectus Group has concentrated on 3 areas: refining the previously existing ontology; expanding the ontology into the areas of Service and relationships; and operationalizing the exisiting ontology intop the registry system. In refining the system, we have created a finer ganularity of description for intellectual workand have reached an understanding on the level of granularity we are seeking. We have also expanded the ontology to include a brief description of service, as a first step in fully developing the second of our (so far) four branches of the ontology, and developed a first pass at relationships between resources containing intellectiual work. Finally, we have used the intellectual work ontological description in creating attributes and domains for the registry, in order to profile information resources. Librarians, programmers, and professors were all closely involved in this last process.

Intellectual Property and Economic Issues

Over the past few months, we have made some progress on the development of commerce mechanisms within the UMDL. The core auction algorithm for our market facilitator agent has been implemented. As our first test of market allocation of resources within the distributed environment, we have produced an initial design for the UMDL market for event-notification services (the Remora agent).

User Interface Agent Working Group

The User Interface Agent (UIA) Working Group has worked toward the following goals: (1) demonstrating cross-collection searching of bibliographic databases using Z39.50, (2) improving the usability of the Broad System of Ordering, the subject terminology connected with collection and user profiling, and (3) improving the UMDLÕs user interface in response to enhanced system functionality.

The ZClient is a C++ class, designed to be plugged into any C++ code to enable UMDL users to search Z39.50-accessible databases. During development, the ZClient class was tested using the World-Wide Web. An HTML interface along with C++ code to parse and display HTML data was developed and tested. The ZClient class is now a UMDL agent. A ZAgent is now under development. It will allow UMDL Agents and StanfordÕs digital library to search Z39.50-accessible collections, particularly MIRLYNÕs MCAT database. It will search the collection unique to the agent, and return only the properties that the other agents request.

Also under development and testing is the Z39.50 interface to FTL. (FTL is the retrieval engine specific to UMDL.) Currently, this interface is capable of taking a query in RPN (Reverse Polish Notation) from a Z39.50 client and passing it to the Z39.50-to-FTL interface server. The Z39.50 interface server parses the RPN query and queries the FTL collection. FTL returns retrieved titles as file-path names. The interface concatenates these path names and returns the complete record of retrieved titles to the Z39.50 client in HTML format.

The UMDL uses the Broad System of Ordering (BSO) as a controlled subject vocabulary to enable information providers to characterize the subject contents of their collections and to enable UMDL users to describe the specific topics that interest them in rather broad subject terms. UIA staff have added definitions of earth and space science terms from the McGraw-Hill Encyclopedia and Dictionary of Science and Technology and have derived a list of related words from the BSO terms and their definitions. With these related words, UIA staff have formulated a set of parameters for searching BSO terms and demonstrated browsing term relationships in BSO in preparation for implementation in January 1996.

In response to enhanced functionality, UIA staff have developed several iterations of a user interface that is intended for its first user group Ñ high school students in Ann Arbor, Michigan. We have called on the UMDLÕs Education Group to assist us in critiquing various user interface iterations which have required much work on graphical icons and representations. In addition, UIA staff have drafted an interviewer-administered questionnaire to gather end-user reactions to and experiences using the UMDL and especially its interface in walkthroughs with high school students and teachers at Ann Arbor high schools. Prior to data collection, both questionnaire and data collection procedures must be submitted for review by the University of MichiganÕs Human Subject Review Committee.

Evaluation

In the past quarter a group of UMDL personnel concerned about evaluation have met to discuss plans. Outcomes of these meetings can be loosely grouped into actions relevant to public school deployment and actions relevant to campus deployment of the UMDL.

Deployment in the Ann Arbor Public Schools for Fall, 1995 was scaled back to include a single experimental classroom at Pioneer High School and existing computer classrooms at Community High School. A prototype UMDL-based curriculum, which was developed throughout the summer and early fall, is currently being tested in science courses at Pioneer and at Community. As this curriculum matures and as the user-interface of the UMDL becomes more attractive and appropriate for secondary school users, deployment will be spread into other classrooms at Pioneer and will make an initial deployment at Huron High School. Lack of adequate Internet infrastrucutre within the Ann Arbor Public Schools (with the exception of Community) remains a significant barrier to faster and broader deployment of UMDL-based curricula. Evaluation of the Fall, 1995 testbed classrooms is being supervised.

Deployment at the University of Michigan was postponed until January, 1996 in order to better assess user needs on campus. Interviews with science librarians, combined with bilbiometric analysis of publication and reading patterns among space scientists, suggested that existing UMDL content was not adequate to support a campus deployment in Fall, 1995. Specifically, analyses showed that demand was highest for publications sponsored by the American Geophysical Union (AGU) -- and none of these journals are currently represented in the UMDL. In fact, a survey of reading and publishing patterns among space physicists revealed that only a small fraction of the current UMDL content is currently useful to this community (i.e., the journal Science). Based on these findings, we have initiated a renewed effort to obtain permissions from the AGU, and the science librarians have begun an effort to gauge journal use among University of Michigan space scientists.

Additional analysis of space scientists' demands showed a strong need for reliable and efficient access to archived data collected by ground-based and spacecraft-based instruments (e.g., the National Center for Atmospheric Research archives in Boulder, CO). Based on this finding, an aggressive effort was launched to asses the existence and demand for archived scientific data. Experience in a parallel project [UARC - the Upper Atmospheric Research Collaboratory] strongly suggested the desirability of real-time collaboration over images visualized from archived data. Creating a mechanism to obtain these data via the UMDL was identified as a high priority for successful deployment at the University of Michigan. Along these lines, we began a series of interviews with space scientists to determine the feasibility of offering archived data through the UMDL interface, possibly as a front-end to archival playback of key atmospheric events in UARC (vis the December, 1993 sub-storm).

In summary, the evaluation effort is at an early stage and is focused on assessing user needs with the hope that changes in the UMDL interface and system can reflect these needs. Unfortunately, the overall unreliability of the February, 1995 UMDL prototype and the primitiveness of that prototype's user interface severely limit user evaluation. Assuming dramatic improvements in the next generation of the UMDL system and interface, deployment and evaluation will be more ambitious in the next quarter.

Education and Deployment

The overarching goal of this group is to create learning experiences for high school science students in which the resources in the NSF-UMDL play a critical role. The pedagogy that we have adopted in this effort is that of inquiry-based learning: students pursue a question that is meaningful to them and develop various artifacts that represent their evolving understanding. For example, we are developing a curriculum unit that explores the content area of "ozone." Potentially meaningful questions that involve this content are area: "Should I wear sunscreen?" or "In what ways will the food chain be disrupted by the greenhouse effect?"

Currently, we are developing online curriculum materials that help structure students' investigations of the above sorts of questions. We are also developing a tutorial unit on how to do "online research." The investigations will be conducted online: the "ozone unit" is a five day unit in which students are online for those five days. We are working with teachers in three Ann Arbor high schools; we are developing implementation models that are tuned to the different needs of each school. The difference in teaching/learning among these schools represent a sub-set of schools across the nation. Thus, the models we develop should have wide applicability.

This past quarater has been one of planning for classroom implementation in the next quarter. By the end of this fall semester, over 500 students will have used our online curriculum materials and carried out a range of investigations using the NSF-UMDL and Internet resources.

Collection Development

Collection negotiations are continuing process. This quarter, we have accomplished the following: