Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

Using Accent Information to Correctly Select Japanese Phrases Made of Strings of Syllables

Tetsuo Araki (1), Satoru Ikehara (2), Hideto Yokokawa (1)

(1) Faculty of Engineering, Fukui University, Fukui, Japan
(2) NTT Communication Science Laboratories, Kanagawa, Japan

This paper proposes a method to determine the most suitable string of syllable candidates using 2nd-order Markov model of accent information added to syllable characters, assuming that the correct positions of accents can be obtained from speech by acoustic processing. From the experiment which uses the statistical data for 5 issues of a daily Japanese newspaper, the accuracy rate that the first candidate of syllable strings are correct was shown to be improved by 6.4 %.

Full Paper

Bibliographic reference.  Araki, Tetsuo / Ikehara, Satoru / Yokokawa, Hideto (1994): "Using accent information to correctly select Japanese phrases made of strings of syllables", In ICSLP-1994, 2155-2158.