Search

This report presents the principal objectives and results of the Context Information Processing Project and discusses the nature of context and its relationship to conventional models of information storage and retrieval. In this approach literature searching involves prediction, and context represents a new class of clues to be obtained and processed for the overall improvement of automatic literature search systems. Context clues are those items of information that describe various objective properties and relationships that hold for individual documents: authors and reviewers, professional societies, journals, etc. Traditionally, literature searching systems have assumed that every document has a content that can be identified properly in terms of one or more index terms, describing what that document is "about." This project suggests that the fundamental concepts of subject content and of "aboutness" are imprecise and confused, in comparison to the naming of individual things or observable properties in context language. There is no implication that content processing should be eliminated; instead, the project hopes to establish context information as relevant for literature searching systems, and, furthermore, to discover the correlation, if any, between context and content information and put this correlation to use. (Author/RM)