Natural language front-ends for expert systems
The development and implementation of NL-front-ends for expert systems has become one of the major concerns for ai research. Although this paper aims at a general discussion of the issues involved, its main points are based on the author's work on the NL-front-end for MDX, a medical expert system developed at the Ohio State University, as well as from READ (reader for event analysis and detection) and the currently being developed NL-front-end for BASIS, both projects of Battelle Memorial Institute. Information extracted from linguistic structures is subject to conventional and conversational implicatures, whereas at a later processing stage, conceptually motivated rules take effect in the reasoning process that is the domain of the expert system proper. The following five criteria are proposed as part of a design procedure to ensure the effectiveness and future productivity of a system: (1) modularity; (2) transparency; (3) extendability; (4) transportability; and (5) robustness. 17 references.
- OSTI ID:
- 5082378
- Resource Relation:
- Conference: Sponsored by Online Rev, New York, NY, USA, 10 Apr 1984
- Country of Publication:
- United States
- Language:
- English
Similar Records
FRONT-END ASIC FOR A SILICON COMPTON TELESCOPE.
Natural Language interactions with artificial experts