Semantic Information Retrieval for Verbal Data.
Abstract
The development of a method for storing and retrieving unstructured narrative data by semantically-sensitive techniques suitable for automation is presented. This research responds to a specific Naval requirement for techniques to improve the command decision process via capture and analysis of Fleet exercise command-and-control data. The technique utilizes semantic surrogates or abstracts of source data as the basis for indexing. Retrieval incorporates inferential techniques designed to identify data relevant to, but not rigidly specified by, a query. When a search locates a relevant abstract or logical chain of abstracts the corresponding sentences or messages are retrieved. The key to the abstraction scheme is a unique concept of information as a transformational process, wherein the existence of prior states of knowledge concerning entities in the text is exploited. Sentences define transformations of characteristics of these entities. The abstraction process categorizes and indexes the transformations, and subsequent retrieval is based upon these indices, each compacted into only 200 bits of memory.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1972
- Accession Number
- AD0893749
Entities
People
- Fredric M. Blum
- Regina M. Harris
- Samuel D. Epstein
- Stephen W. Leibholz