Skip to content

Publications

Detail of publication

Citation

Vaněk J. and Psutka J. : Anti-Models: An Alternative Way to Discriminative Training . Text Speech nad Dialoque - TSD 2014, Text Speech nad Dialoque - TSD 2014, p. 449-456, Springer, 2014.

Download PDF

PDF

Abstract

Traditional discriminative training methods modify Hidden Markov Model (HMM) parameters obtained via a Maximum Likelihood (ML) criterion based estimator. In this paper, anti-models are introduced instead. The anti-models are used in tandem with ML models to incorporate a discriminative information from training data set and modify the HMM output likelihood in a discriminative way. Traditional discriminative training methods are prone to over-fitting and require an extra stabilization. Also, convergence is not ensured and usually "a proper" number of iterations is done. In the proposed anti-models concept, two parts, positive model and anti-model, are trained via ML criterion. Therefore, the convergence and the stability are ensured.

Detail of publication

Title: Anti-Models: An Alternative Way to Discriminative Training
Author: Vaněk J. ; Psutka J.
Language: English
Year: 2014
Type of publication: Papers in proceedings of reviewed conferences
Book title: Text Speech nad Dialoque - TSD 2014
Title of journal or book: Text Speech nad Dialoque - TSD 2014
Page: 449 - 456
DOI: 10.1007/978-3-319-10816-2_54
ISBN: 978-3-319-10815-5
ISSN: 0302-9743
Publisher: Springer
/ 2014-11-12 12:22:14 /

Keywords

ASR, HMM, Acoustic Modeling, Discriminative Training, Anti-Models, MMI, MCE, MPE

BibTeX

@INPROCEEDINGS{VanekJ_2014_Anti-ModelsAn,
 author = {Van\v{e}k J. and Psutka J.},
 title = {Anti-Models: An Alternative Way to Discriminative Training },
 year = {2014},
 publisher = {Springer},
 journal = {Text Speech nad Dialoque - TSD 2014},
 pages = {449-456},
 booktitle = {Text Speech nad Dialoque - TSD 2014},
 ISBN = {978-3-319-10815-5},
 ISSN = {0302-9743},
 doi = {10.1007/978-3-319-10816-2_54},
 url = {http://www.kky.zcu.cz/en/publications/VanekJ_2014_Anti-ModelsAn},
}