A spoken language understanding approach which combines the parsing robustness with the interpretation deepness

摘要

In this paper we present a spoken language understanding approach which combines the parsing robustness with the interpretation deepness. The processing is done following four main steps: at the first a pre-processing module normalizes spoken utterances from lexical extragrammaticalities, repetitions and self-corrections. The second step consists of parsing the input utterances. Parsing is done on Semantic Tree Association Grammar STAG. We propose this formalism as an extended version of traditional semantic grammar. The key features of this formalism are the parsing deepness and the efficiency for processing the spoken language syntactic phenomena. Then the post-processing step corrects the input from false starts and incompleteness. Finally, the parsed utterances are associated with a pure semantic interpretation by the IF module.