Semantic Processing For Speech Understanding

Abstract

The semantic component of the speech understanding system being developed jointly by SRI and SDC rules out phrase combinations that are not meaningful and produces semantic interpretations for combination that are. The system consists of a semantic network model and routines that interact with it. The net is partitioned into a set of hierarchically ordered subnets, facilitating the encoding of higher-order predicates and the maintenance of multiple parsing hypotheses. Composition routines, combining utterance components into phrases, consult network descriptions of prototype situations and surface-to-deep-case maps. Outputs from these routines are network fragments consisting of several subnets that in aggregate capture the interrelationships between a phrase’s syntax and semantics.