An Indexing Technique for Storage and Retrieval of Unstructured Verbal Data.
Abstract
Interim results are presented of an investigation to develop techniques for storing and retrieving unstructured narrative data (for example, messages) by automated techniques, an area not yet satisfactorily within the state of the computer art. Undertaken as part of the MAFTEP (Methods for Analysis of Fleet Tactical Effectiveness and Performance) Program, this research responds to a specific Naval requirement for improved techniques to analyze the command decision process and to support command and control. The technique utilizes semantic abstracts of source data as the basis for indexing and retrieval. Syntactical analysis or keyword search are not utilized. Each semantic abstract forms a data surrogate suitable for automatic retrieval. When a search locates a relevant abstract or logical chain of abstracts the corresponding sentences can be retrieved. The key to the abstraction scheme is a unique conceptualization of information as a transformational process, wherein a prior state of knowledge concerning 'participants' to the message subject matter is assumed to exist. Sentences are then shown to define transformations of characteristics of the participants. The abstraction process categorizes and indexes these transformations, and subsequent retrieval is based upon the indices. Brief indexing and retrieval experiments performed on a Naval message data base have indicated feasibility of the technique. Work is proceeding on complex retrieval problems and on specific problems of future automation. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1969
- Accession Number
- AD0867183
Entities
People
- Fredric M. Blum
- Stephen W. Leibholz