First International Conference on Spoken Language Processing (ICSLP 90)

Kobe, Japan
November 18-22, 1990

A Large-Scale Japanese Speech Database

Yoshinori Sagisaka (1), Kazuya Takeda (2), M. Abel (1), Shigeru Katagiri (3), T. Umeda (4), H. Kuwabara (5)

(1) ATR Interpreting Telephony Res. Labs., Seika-chou, Kyoto, Japan
(2) KDD Kamifukuoka R&D Laboratories, Kamifukuoka-shi, Saitama, Japan
(3) ATR Auditory and Visual Perception Res. Labs., Kyoto, Japan
(4) NHK Science Technical Research Laboratories, Tokyo, Japan
(5) West Tokyo Science University, Uenohara, Yamanashi, Japan

This paper describes a large-scale Japanese speech database (JSDB) which is now under construction for the studies of speech recognition and synthesis. The database consists of (a) an isolated word JSDB, (b) an isolated sentence JSDB, (c) a mixed word and sentence JSDB and (d) a text JSDB. For multi-purpose use, five different transcriptions were made for almost all databases: phonetic symbols, acoustic events, allophonic variants, inseparable portions, and vowel centers. Moreover, in some of these databases, the grammatical information such as a part of speech, inflectional categories and phrase structure is given to the constituents in the corresponding sentences. These databases have been used for various research purposes in speech technology.

Full Paper

Bibliographic reference.  Sagisaka, Yoshinori / Takeda, Kazuya / Abel, M. / Katagiri, Shigeru / Umeda, T. / Kuwabara, H. (1990): "A large-scale Japanese speech database", In ICSLP-1990, 1089-1092.