ON THE BEST WAY OF DESIGNATING OBJECTS IN SENTENCE GENERATION

Abstract

This paper deals with a problem encountered in natural language generation which seems to have been largely ignored in the literature, that of generating non‐ambiguous (i.e. discriminating) designations of objects in a given context, from a knowledge basis, which associates the properties and relations, concerning the objects present in the environment, with their respective formal labels. A search algorithm of type A is proposed, which always generates a discriminating designation when such a designation exists in terms of the available knowledge; for the evaluation the algorithm uses a subjective length function which takes into account the “intelligibility” of the designation. This work takes place in the SYROCO system, a dialogue interface for limited domains of discourse; the sentence interpretation as well as the sentence generation in SYROCO are briefly presented in the first part of this paper.

Citation

FARRENY, H. and PRADE, H. (1984), "ON THE BEST WAY OF DESIGNATING OBJECTS IN SENTENCE GENERATION", Kybernetes, Vol. 13 No. 1, pp. 43-46. https://doi.org/10.1108/eb005673