Contents

  Freedom of Information Conference 2000

Prospects for Improved Access to Large Document Sets

W. John Wilbur
National Center for Biotechnology Information National Library of Medicine

Abstract

Document retrieval from a large natural language database in answer to a natural language query is an important problem that challenges current technology and occupies a number of researchers in the field of digital libraries. The goal is more intelligent retrieval and the expectation is that the fields of artificial intelligence and knowledge bases will at some point make a substantial contribution to this effort. Because it is recognized that intelligence depends on a significant store of knowledge it is imperative that knowledge be built into systems in some way if they are to produce intelligent retrieval. All of this leads to an important question, What kind of knowledge could lead to more intelligent retrieval? We find that the answer is not common sense and not subject specific knowledge in the topical area of the database, but rather some kind of generic pattern recognition ability in comparing documents that resides not in a single person but in the whole community of database users. Based on the collective experience of the AI community and on our own experimental work we outline one path that we believe can lead to the capture of such knowledge and improved retrieval.