First International Conference on Spoken Language Processing (ICSLP 90)

Kobe, Japan
November 18-22, 1990

A Large Vocabulary Continuous Speech Recognition System With High Prediction Capability

Minoru Shigenaga (1), Yoshihiro Sekiguchi (2), Toshihiko Hanagata (2), Takehiro Yamaguchi (2), Ryouta Masuda (2)

(1) School of Computer and Cognitive Sciences, Chukyou University, Toyta, Japan
(2) Faculty of Engineering, Yamanashi University, Kofu, Japan

A large vocabulary (with 1018 words and 1379 kinds of inflectional endings) continuous speech recognition system with high predictability applicable to any task and aiming to have unsupervised speaker adaptation capability is described. Phoneme identification bases on various features. Speaker adaptation is done using reliablely identified phonemes. Using prosodic information, phrase boundaries are detected. The syntactic analyzer uses a syntactic state transition network and outputs their syntactic interpretations. The semantic analyser deals with meaning of each words, relationship between words, and extended case structures of predicates. Recognizing noun phrases first, predicates are predicted, and vice versa.

Full Paper

Bibliographic reference.  Shigenaga, Minoru / Sekiguchi, Yoshihiro / Hanagata, Toshihiko / Yamaguchi, Takehiro / Masuda, Ryouta (1990): "A large vocabulary continuous speech recognition system with high prediction capability", In ICSLP-1990, 721-724.