GENERATION OF VCV/CVC BALANCED WORD SETS FOR SPEECH DATA BASE.

S. Hayamizu*, K. Tanaka, S. Yokoyama, K. Ohta

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

15 Citations (Scopus)

Abstract

This paper describes a method how to generate word sets for the study of phonetic features in various contexts and shows statistical properties of the generated word sets. Two word sets are generated so as to contain as many kinds of VCVs and CVCs in as few words as possible. In addition, these sets are phonetically balanced by considering the entropy for VCVs and CVCs. The words are selected from 42,859 words on a Japanese dictionary phonetically described. The dictionary contains 44,000 different entry words when the homophones are counted as one word.

Original languageEnglish
Pages (from-to)803-834
Number of pages32
JournalDenshi Gijutsu Sogo Kenkyusho Iho/Bulletin of the Electrotechnical Laboratory
Volume49
Issue number10
Publication statusPublished - 1985
Externally publishedYes

ASJC Scopus subject areas

  • Condensed Matter Physics
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'GENERATION OF VCV/CVC BALANCED WORD SETS FOR SPEECH DATA BASE.'. Together they form a unique fingerprint.

Cite this