IMPROVED SEARCH STFUTEGU FOR LARGE VOCABULARY CONTINUOUS MANDA4RIN SPEECH RECOGNITION
Tai-Hsuan Ho, Kae-Cherng Yang, Kuo-Hsun Huang, Lin-shan Lee · 1998
'l'his paper presents a new search strategy for large vocabulary continuous Mandarin speech recognition considering the special structure of Chinese language. This strategy is composed of a forward and a backward passes, between which a high-quality syllable lattice is generated to bridge the syllable-level and lord-level decoding processes. In the forward pass, considering the small number of syllables in Chinese language, a framesJnchronous stack decoder is used to inUegrate the high-order syllable N-Gram language model, so as to generate a very accurate and compact syllable lattice In the backward pass, considering the special monosyllabic wording structure in Chinese language, the search space for the word-level decoding is expanded dynamically from the syllable lattice, and the best word sequence is extracted based on the knowledge provided by the word pronunciation lexicon and the word N-Gram language model. In the preliminary experiments, it was found that, with this strategy, the character error rate can be reduced by more than 20% as compared with a previous system using syllablealigned lattice approach on a speaker-adaptive continuous speech recognition task.