A regularized discriminative training method of acoustic models derived by minimum relative entropy discrimination

Yotaro Kubo, Shinji Watanabe, Atsushi Nakamura, Tetsunori Kobayashi

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Citations (Scopus)

Abstract

We present a realization method of the principle of minimum relative entropy discrimination (MRED) in order to derive a regularized discriminative training method. MRED is advantageous since it provides a Bayesian interpretations of the conventional discriminative training methods and regularization techniques. In order to realize MRED for speech recognition, we proposed an approximation method of MRED that strictly preserves the constraints used in MRED. Further, in order to practically perform MRED, an optimization method based on convex optimization and its solver based on the cutting plane algorithm are also proposed. The proposed methods were evaluated on continuous phoneme recognition tasks. We confirmed that the MRED-based training system outperformed conventional discriminative training methods in the experiments.

Original languageEnglish
Title of host publicationProceedings of the 11th Annual Conference of the International Speech Communication Association, INTERSPEECH 2010
PublisherInternational Speech Communication Association
Pages2954-2957
Number of pages4
Publication statusPublished - 2010

Publication series

NameProceedings of the 11th Annual Conference of the International Speech Communication Association, INTERSPEECH 2010

Keywords

  • Discriminative training
  • Optimization
  • Speech recognition

ASJC Scopus subject areas

  • Language and Linguistics
  • Speech and Hearing

Fingerprint Dive into the research topics of 'A regularized discriminative training method of acoustic models derived by minimum relative entropy discrimination'. Together they form a unique fingerprint.

Cite this