Enhancing Metadata Assignment in a Digital Library with Background Knowledge

Traditionally, information retrieval experts have addressed the problem of information overload. This pursuit has yielded automated electronic bibliographic card catalogs as well as effective search engines for enormous collections including the World Wide Web. Despite advances in this area, librarians lament the ambiguity or absence of metadata required to effectively communicate content. In the 20th century, technocrats such as Vanevar Bush and J. C. R. Licklider foresaw a future in which automatic interactive systems aid humans in navigating and understanding the knowledge of science and humanities. Eventually, the Web and Hypertext Transfer Protocol presented a medium for the digitization and dissemination of knowledge in textual and multimedia forms. More recently, the Semantic Web has provided a framework to encode the metadata about such corpora and thereby automate information discovery at the finest levels of semantic granularity. Yet metadata of such scope and quality are seldom realized. Various obstacles inhibit the acquisition, encoding, and sharing of metadata. In the context of institutional repositories, users’ ignorance of the technical language of bibliographic cataloging inhibits knowledge capture at the time of document ingestion. Furthermore, the technical languages of digital bibliographic catalogs suffer from the same semantic problems as other formal knowledge representation languages, including context-specific definitions and conflicting representations. The difficulties of large-scale knowledge formalization have been documented by scholars like Marvin Minsky and Lucy Suchman and realized all too well during largescale projects such as Doug Lenat’s Cyc. However, the institutional repository offers structures to deal with the obstacles to capturing, encoding, and sharing metadata. The institutional repository not only encompasses schemas and standards for encoding information and metadata, but it is also oriented toward the intellectual domains of its host institution. This situation is ripe for knowledge-based approaches to repository management. The work proposed here aims to enhance knowledge cap-