TRANSLATION ALGORITHMS FOR NATURAL LANGUAGE ACCESS CAPABILITY.
Abstract
The report describes two approaches to solving the problem of natural language access to mechanically stored data by translation algorithms. The first approach operates on the output of an existing parsing algorithm for natural language sentences. When a base P marker for an input sentence has been produced by this algorithm, interpretive transformations on this marker produce a formalization of the sentence in the quasi-language SNORP, which may then be compared with the information in the data base to produce an appropriate answer. The alternative approach defines certain semantic markers from the nature of the data base and, by a multipass algorithm, extracts from the input sentence those semantic units necessary to supply a satisfactory answer. The quasi-language MOL provides expression for both the data base and the semantic content of the questions to it. The third section of the report deals with the more general problem of man-machine interface in terms of communication via a restricted subset of natural language sentences. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1967
- Accession Number
- AD0690598
Entities
People
- R. Jackendoff
- T. J. Monks
- W. C. Watt