Articulatory Feature Classifiers Trained on 2000 hours of Telephone Speech

J. Frankel; M. Magimai-Doss; S. King; K. Livescu; Ã. Ãetin

Articulatory Feature Classifiers Trained on 2000 hours of Telephone Speech

J. Frankel, M. Magimai-Doss, S. King, K. Livescu, Ã. Ãetin

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

This paper is intended to advertise the public availability of the articulatory feature (AF) classification multi-layer perceptrons (MLPs) which were used in the Johns Hopkins 2006 summer workshop. We describe the design choices, data preparation, AF label generation, and the training of MLPs for feature classification on close to 2000 hours of telephone speech. In addition, we present some analysis of the MLPs in terms of classification accuracy and confusions along with a brief summary of the results obtained during the workshop using the MLPs. We invite interested parties to make use of these MLPs.

Original language	English
Title of host publication	Interspeech 2007
Subtitle of host publication	8th Annual Conference of the International Speech Communication Association
Pages	2485-2488
Number of pages	4
Publication status	Published - 1 Aug 2007

Access to Document

frankel_AF-MLP
© Frankel, J., Magimai-Doss, M., King, S., Livescu, K., & Ãetin, Ã. (2007). Articulatory Feature Classifiers Trained on 2000 hours of Telephone Speech. In Interspeech 2007: 8th Annual Conference of the International Speech Communication Association. (pp. 2485-2488)
Accepted author manuscript, 98.5 KB

Efficient model-level integration of novel methods into HMM speech recognition
King, S.
UK central government bodies/local authorities, health and hospital authorities
1/07/06 → 29/02/12
Project: Research
Streamed models for automatic speech recognition (EPSRC Advanced Research Fellowship)
King, S.
EPSRC
1/01/05 → 31/12/09
Project: Research
Espresso-II: SWITCHING LINEAR DYNAMIC SPEECH MODELS
King, S.
EPSRC
1/07/03 → 30/06/06
Project: Research

Cite this

@inproceedings{c37a15892f874d9a97ef9e93bf300c64,

title = "Articulatory Feature Classifiers Trained on 2000 hours of Telephone Speech",

abstract = "This paper is intended to advertise the public availability of the articulatory feature (AF) classification multi-layer perceptrons (MLPs) which were used in the Johns Hopkins 2006 summer workshop. We describe the design choices, data preparation, AF label generation, and the training of MLPs for feature classification on close to 2000 hours of telephone speech. In addition, we present some analysis of the MLPs in terms of classification accuracy and confusions along with a brief summary of the results obtained during the workshop using the MLPs. We invite interested parties to make use of these MLPs.",

author = "J. Frankel and M. Magimai-Doss and S. King and K. Livescu and {\~A}. {\~A}etin",

year = "2007",

month = aug,

day = "1",

language = "English",

isbn = "1990-9772",

pages = "2485--2488",

booktitle = "Interspeech 2007",

}

TY - GEN

T1 - Articulatory Feature Classifiers Trained on 2000 hours of Telephone Speech

AU - Frankel, J.

AU - Magimai-Doss, M.

AU - King, S.

AU - Livescu, K.

AU - Ãetin, Ã.

PY - 2007/8/1

Y1 - 2007/8/1

N2 - This paper is intended to advertise the public availability of the articulatory feature (AF) classification multi-layer perceptrons (MLPs) which were used in the Johns Hopkins 2006 summer workshop. We describe the design choices, data preparation, AF label generation, and the training of MLPs for feature classification on close to 2000 hours of telephone speech. In addition, we present some analysis of the MLPs in terms of classification accuracy and confusions along with a brief summary of the results obtained during the workshop using the MLPs. We invite interested parties to make use of these MLPs.

AB - This paper is intended to advertise the public availability of the articulatory feature (AF) classification multi-layer perceptrons (MLPs) which were used in the Johns Hopkins 2006 summer workshop. We describe the design choices, data preparation, AF label generation, and the training of MLPs for feature classification on close to 2000 hours of telephone speech. In addition, we present some analysis of the MLPs in terms of classification accuracy and confusions along with a brief summary of the results obtained during the workshop using the MLPs. We invite interested parties to make use of these MLPs.

M3 - Conference contribution

SN - 1990-9772

SP - 2485

EP - 2488

BT - Interspeech 2007

ER -

Articulatory Feature Classifiers Trained on 2000 hours of Telephone Speech

Abstract

Access to Document

Fingerprint

Projects

Efficient model-level integration of novel methods into HMM speech recognition

Streamed models for automatic speech recognition (EPSRC Advanced Research Fellowship)

Espresso-II: SWITCHING LINEAR DYNAMIC SPEECH MODELS

Cite this