Textual-Geospatial Integration Services for the National SMETE Digital Library

University Of California-Santa Barbara, Santa Barbara CA

Investigators

Abstract

The emerging National Science, Mathematics, Engineering, and Technology Education Digital Library (NSDL) includes many sets of items whose representations, while different at the user level, contain semantically related information. Consider, for example, the association between land-use maps of some region, and references to that region in a book concerning agricultural practices. The NSDL aims to support its users with services that help them find and integrate such semantically related information in such heterogeneously represented items. This project is constructing operational services to make it easy for users of the NSDL to have integrated access to information from texts, maps, and images--specifically, information that is semantically related in terms of geospatial features and regions. These services will help users find: -- maps or images containing features or places that are referred to implicitly within texts, and -- texts that implicitly reference features and places contained in maps and images. More advanced services will help users evaluate the consistency between geospatial information obtained from such varying sources. Collectively, these services will allow large collections of texts to be indexed and accessed geospatially, even if their "cataloging records" do not contain explicit geospatial references. To implement these services, the project team is integrating the geospatial search services of the Alexandria Digital Library (ADL; http://www.alexandria.ucsb.edu) at the University of California at Santa Barbara with the high-performance information retrieval (IR) services developed by the Illinois Institute of Technology's Information Retrieval Laboratory (http://www.ir.iit.edu). The key component in this integration is the 4+ million item gazetteer and associated services developed for ADL. (A gazetteer provides a mapping between textual representations of places and features, and coordinate-based representations of their geospatial "footprints," thus allowing a semantic mapping to be made in either direction between textual and map/image representations of related phenomena.) Specifically, the project team is integrating a gazetteer into an IR engine, integrating an IR engine into a geospatial digital library, and developing services that check the consistency of alternative geospatial references. These services will allow seamless geospatial queries across distributed heterogeneous (text and non-text) digital libraries and will enable geospatial referencing of texts that have not otherwise been geographically cataloged (i.e., almost all texts currently available).