Japanese word prediction and large vocabulary speech recognition based on stochastic language model

• ZHOU Min

  Toyohashi University of Technology, Department of Information & Computer Sciences

• NAKAGAWA Seiichi

  Toyohashi University of Technology, Department of Information & Computer Sciences

Bibliographic Information

Other Title

• 確率モデルによる後続単語予測と大語彙日本語連続音声認識

Abstract

A study on comparing different types of stochastic language models based on the ATR dialog database is given. The comparison of a part of speech/word-unit perplexity and prediction ability for the following words based on hidden Markov model, stochastic context-free grammar, bigram, trigram, and bigram-HMM is evaluated. By using the bigram model, we performed the continuous speech recognition of a 4699 word vocabulary size on the speaker adaptation models.

Journal

• IPSJ SIG Notes

  IPSJ SIG Notes 6 1-8, 1995-05-25

  Information Processing Society of Japan (IPSJ)

Keywords

• stochastic model
• HMM
• context-free grammar
• bigram
• trigram
• entropy
• perplexity
• word prediction

Details

• CRID: 1572824502051751040
• NII Article ID: 110002916955
• NII Book ID: AN10442647
• ISSN: 09196072
• Text Lang: ja
• Data Source:
  • CiNii Articles