First International Conference on Spoken Language Processing (ICSLP 90)

Kobe, Japan
November 18-22, 1990

A Japanese Text Dictation System Based on Phoneme Recognition Using a Modified LVQ2 Method

Shozo Makino, Akinori Ito, Mitsuru Endo, Ken'iti Kido

Research Center for Applied Information Sciences, Tohoku University, Sendai, Japan

A Japanese text dictation system has been developed based on phoneme recognition and a dependency grammar. The phoneme recognition is carried out using a modified LVQ2 method proposed. A linguistic processor is composed of a Bunsetsu-unit(Japanese Phrase) spotting processor and a syntactic processor with semantic constraints. In the Bunsetsu-unit spotting processor, using a syntax driven continuous DP matching algorithm the Bunsetsu-units are spotted from a recognized phoneme sequence and then a Bunsetsu lattice is generated. In the syntactic processor, the Bunsetsu lattice is parsed based on the dependency grammar. The dependency grammar is expressed as the correspondence between a FEATURE marker in a modifier-Bunsetsu and a SLOT-FILLER marker in a head-Bunsetsu. Recognition scores of the Bunsetsu-unit and phoneme were 73.2% and 86.1% for 113 sentences uttered by each of two male speakers.

Full Paper

Bibliographic reference.  Makino, Shozo / Ito, Akinori / Endo, Mitsuru / Kido, Ken'iti (1990): "A Japanese text dictation system based on phoneme recognition using a modified LVQ2 method", In ICSLP-1990, 241-244.