Integrated speech and morphological processing in a connectionist continuous speech understanding for Korean

摘要

A new tightly coupled speech and natural language integration model ispresented for a TDNN-based continuous possibly large vocabulary speechrecognition system for Korean. Unlike popular n-best techniques developed forintegrating mainly HMM-based speech recognition and natural language processingin a {\em word level}, which is obviously inadequate for morphologicallycomplex agglutinative languages, our model constructs a spoken language systembased on a {\em morpheme-level} speech and language integration. With thisintegration scheme, the spoken Korean processing engine (SKOPE) is designed andimplemented using a TDNN-based diphone recognition module integrated with aViterbi-based lexical decoding and symbolic phonological/morphologicalco-analysis. Our experiment results show that the speaker-dependent continuous{\em eojeol} (Korean word) recognition and integrated morphological analysiscan be achieved with over 80.6% success rate directly from speech inputs forthe middle-level vocabularies.