Understanding digital-system specifications written in natural language
This thesis concerns itself with the specification of digital systems. The specific focus of the work described here was on understanding system specifications written in natural language. The long-term goals of the research are to provide methods and software to assure that the specifications are consistent, correct, and complete. The research described differs from previous research in several ways. First, the natural language input is used to construct an internal design representation, rather than just to query about existing design data. Second, using natural language allows a generality of expression not found in formal models. Finally, the natural language is not overly restricted. A major part of the research involves formally modeling the information found in system specifications. An extension of the USC Design Data Structure is described, with emphasis on timing and control flow. A semantic parser, PHRAN,is used as the basis for the actual interface software. PHRAN contains a knowledge base of sentence patterns along with associated concepts. PHRAN inputs English sentences and looks for patterns in the sentences. When it finds a pattern match, the concept associated with the pattern is particularized with the information found in the sentence.
- Research Organization:
- University of Southern California, Los Angeles (USA)
- OSTI ID:
- 7043171
- Resource Relation:
- Other Information: Thesis (Ph. D.)
- Country of Publication:
- United States
- Language:
- English
Similar Records
English-language processing system that learns about new domains
Retaining Systems Engineering Model Meaning Through Transformation: Demo 2