On automatic support to indexing a life sciences data base

摘要

The paper describes a technique developed as automatic support to subject heading indexing at BIOSIS. The technique is based on the use of a formalized language for semantic representation of biological texts and subject headings—the language of Concept Primitives. The structure of the language is discussed as well as the structure of the Semantic Vocabulary, in which natural language words from biological texts are described by Concept Primitives. The Semantic Vocabulary is being constructed. Approximately 8,000 entries corresponding to high frequency significant words have been compiled, comprising at least three-quarters of the final number. Results of experiments checking the approach are given, and journal/subject heading and author/subject heading correlation data are analyzed to be used as a supporting technique.