LIBRARY TEAM

This team is responsible for various issues relating to the nature and characterization of collection items. They include the special issues related to integration of spatially referenced information objects into the digital library. A major set of issues that will be examined and resolved during the 12-month period include:

L1

RESEARCH TASK: the levels of geographic detail required for characterizing spatially-referenced objects

PLANNED ACTIVITY

The level of geographic detail of a spatially referenced object is a fundamental property, related to traditional concepts of generalization, resolution, and scale. We know that it is a difficult property to deal with, and that many traditional metaphors of geographic detail, such as scaling from a paper map to the surface of the Earth, fail to transfer robustly to the digital environment. This research will investigate alternatives, identify research issues that need to be addressed in the area of human cognition of concepts of geographic detail, and examine the implications for Alexandria. A paper on these issues will be developed by June 30, and the library team will interact with other Alexandria groups during the year to examine implications of geographic detail, particularly in the areas of metadata and user interface design.

ACTUAL ACTIVITY

This work has progressed in a satisfactory manner, with the results being reported in two papers, one in the GIS/LIS proceedings from Denver in November, the other to appear in the journal Geographic and Environmental Modeling in the first issue of 1997. Planning is now underway on how to integrate the results into the testbed.

L2

RESEARCH TASK: the characterization of items in terms of complex ``spatial footprints'', and especially fuzzy footprints

PLANNED ACTIVITY

We have developed a series of methods for representing uncertain geographic footprints. These occur: in the context of the gazetteer, when place-names do not correspond to crisply defined areas on the ground; in the query engine, when the user wishes to search for objects but cannot define a precise area of search; and in the catalog, where information objects may be associated with fuzzy footprints. We are investigating appropriate measures of fit between fuzzily defined queries and information objects. A demonstration has been developed, and a draft paper written. The paper will be completed for submission by June 30. In the Fall we will begin a series of experiments to elicit fuzzy footprints from Alexandria users. A second paper on these experiments will be completed by February 1997.

ACTUAL ACTIVITY

Several models of fuzzy footprints have been constructed and demonstrations have been given in a number of settings. The models are being evaluated in terms of the Mojave project. A paper is currently in progress, and will be completed by 6/30/97.

L3

RESEARCH TASK: The extraction of ``implicit footprints'' from current (traditional) library catalog records

PLANNED ACTIVITY

Library catalogs contain a rich source of information on geographic place-name referents, notably in particular subject categories. We plan to ``mine'' large-scale digital catalogs, such as Melvyl and OCLC, for place-name referents. We will then examine various techniques for displaying and using such information in a digital library context. For example, we will create a ``map'' of the place-name references in Melvyl, and of interesting subsets of records selected based on a variety of rules. One motivation for this research is the need to build a better source of feature extents for gazetteers. This research will continue through the year, and we plan to have sufficient results to support at least one publication by February 1997.

ACTUAL ACTIVITY

This project was not undertaken in the current year, owing to constraints on resources available to perform the analysis in the UC Office of Library Automation. The project is now scheduled for mid-1997.

METADATA AND CATALOG INTEROPERABILITY SUBTEAM

L4

RESEARCH TASK: The development of ``lightweight'' catalog views

PLANNED ACTIVITY

In relation to the need for more complex metadata representations, the team will incorporate extensions to current metadata schemes (e.g. USMARC, FGDC) and also develop various views of this metadata, such as the proposed ``Dublin Core'', for users.

ACTUAL ACTIVITY

We have used the Dublin core attributes in supporting the concept of lightweight catalog views, and have actively participated in all of the Dublin Core sequence of research and development workshops.

L5

RESEARCH TASK: The development of a ``top-down'' framework for metadata modeling

PLANNED ACTIVITY

The development of a general conceptual model of metadata for spatially-referenced materials and the conceptual and logical designs for a catalog based on this model. In particular, various knowledge representation languages (such as KIF, ONTOLINGUA, CONCEPTUAL GRAPHS, R-STRUCTURES) together with current standards (such as USMARC, FGDC, and SDTS) will be examined for their value in representing concepts that are required for characterizing library items and queries. In particular, we will examine the integration of various catalog tools, such as subject heading catalogs, thesauri, gazetteers, and bibliographies, within this framework. A prototype catalog system will be implemented and investigated.

ACTUAL ACTIVITY

We developed a top-down framework for metadata modeling, that was reported in articles in the IEEE Conference on Metadata and in DLIB magazine. The framework is being used in part as a basis for the next version of ADL.

L6

RESEARCH TASK: Semantic interoperability for metadata

PLANNED ACTIVITY

In relation to the framework for metadata, we will study the issue of metadata interoperability, and in particular provide a resolution to the issue of appropriate inter-catalog exchanges of semantic information about metadata.

ACTUAL ACTIVITY

This task has led to the development of a simple model of semantic interoperability based on the use of a small number of standards for geospatial information. In relation to the joint DLI supplement project between Illinois and UCSB, successful experiments relating to concept spaces and self-organizing maps techniques have been carried out in relation to the generation of a visual thesaurus. A large testbed of aerial photos has been created and is being analyzed using the NCSA SGI Power Challenge and Convex Exemplar supercomputers. A Java based interface for viewing visual thesaurus is also under development.

L7

RESEARCH TASK: Metadata for procedures

PLANNED ACTIVITY

Since information retrieval is probably more important than collection item retrieval in DL's, users will need to apply information-retrieving procedures to items. Hence there is a need to resolve the issue of representing such procedures and their use in terms of catalog metadata.

ACTUAL ACTIVITY

Resources were not available to pursue this issue, which we believe to be of great significance.

ALEXANDRIA ATLAS SUBTEAM

L8

RESEARCH AND DEVELOPMENT TASK: Develop an ``Alexandria Atlas''

PLANNED ACTIVITY

The Alexandria Atlas is to serve three purposes:

1. provide a ``live'' geographic background map for spatial searching of the library;
2. incorporate a comprehensive gazetteer, user tools and thesaurus;
3. provide an on-line browseable atlas for users, much like current analog equivalents.

The mapping data will be constructed to allow users to access increasing amounts of geographic detail and procedures by a process of ``drilling down'' through various thematic layers.

Key aspects of the Atlas are that it will:

1. support geographical browsing, search, and retrieval in terms of a graphical ``background map'' interface;
2. allow tying new information resources to the Alexandria database at ingest time;
3. provide geographic proximity search expansion;
4. permit display of the gazetteer's geographic names and allow features to be queried for name and class and to be selected for display;
5. will include both feature-based (vector) and image-based (raster) data in the various layers;
6. have an interface compatible with ADL browser;
7. support variety of U.S. data layers, that may include geodetic control network, DOQs, DLGs, DEMs, center-line street network, watershed ID system of the US, postal code systems (ZIPs in the US), phone codes, the political hierarchy and international datasets that may include DCW, Landsat/SPOT, AVHRR, thematic (e.g. world soils) and global DEMs.
   

ACTUAL ACTIVITY

During the past year, the Atlas and Implementation Teams examined two of the four options identified as possible atlas interfaces. One of these options, the ESRI Map Objects, is currently operational and is being tested. The Digital Chart of the World (DCW) and AVHRR imagery will form the base level of the Atlas together with larger scale datasets such as TIGER Files, Digital Orthrophoto Quads, and Digital Line Graphs for the United States. While the data is available within ADL to construct these layers, the press of other activities has prevented us from assigning qualified staff to complete the compilation of the layers. A related project, however, that may lead to an atlas interface is currently underway with Microsoft. Using their Encarta product as the initial mapping layers, we believe that we may be able to construct the additional Atlas functionality that is laid out in our requirements. An extended model of a gazetteer has been constructed and is currently being tested within a database engine. This extended model should provide a vehicle for the additional types of spatial information that are often found in currently published atlases. To date we have not tested the connection between the atlas and the gazetteer.