A grapheme-based method for automatic alignment of speech and text data

A. Stan; P. Bell; S. King

doi:10.1109/SLT.2012.6424237

A grapheme-based method for automatic alignment of speech and text data

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

Abstract

This paper introduces a method for automatic alignment of speech data with unsynchronised, imperfect transcripts, for a domain where no initial acoustic models are available. Using grapheme-based acoustic models, word skip networks and orthographic speech transcripts, we are able to harvest 55% of the speech with a 93% utterance-level accuracy and 99% word accuracy for the produced transcriptions. The work is based on the assumption that there is a high degree of correspondence between the speech and text, and that a full transcription of all of the speech is not required. The method is language independent and the only prior knowledge and resources required are the speech and text transcripts, and a few minor user interventions.

Original language	English
Title of host publication	Spoken Language Technology Workshop (SLT), 2012 IEEE
Publisher	Institute of Electrical and Electronics Engineers
Pages	286-290
Number of pages	5
ISBN (Electronic)	978-1-4673-5125-6
ISBN (Print)	978-1-4673-5124-9
DOIs	https://doi.org/10.1109/SLT.2012.6424237
Publication status	Published - 2012

Keywords / Materials (for Non-textual outputs)

acoustic signal processing
natural language processing
speech synthesis
text analysis
word processing
automatic speech data alignment
automatic text data alignment
grapheme-based acoustic models
language independent method
orthographic speech transcripts
text transcription
unsynchronised-imperfect transcripts
utterance-level accuracy
word accuracy
word skip networks
Acoustics
Data models
Error analysis
Hidden Markov models
Speech
Speech recognition
Training
grapheme-based models
imperfect transcripts
speech alignment
word networks

Access to Document

10.1109/SLT.2012.6424237

Stan Bell King 2012 A Grapheme Based Method for Automatic Alignment of Speech and Text DataAccepted author manuscript, 1.05 MB

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6424237

Simple4All: Speech synthesis that improves through adaptive learning
King, S. (Principal Investigator) & Renals, S. (Co-investigator)
EU government bodies
1/11/11 → 31/10/14
Project: Research

File
HELP4MOOD:A computational distributed system to support the treatment of patients with major depression
Matheson, C. (Principal Investigator) & Wolters, M. (Co-Investigator (External))
EU government bodies
1/01/11 → 30/06/14
Project: Research

EACL 2014 keynote: Speech synthesis needs YOU!
Simon King (Speaker)
29 Apr 2014
Activity: Academic talk or presentation types › Invited talk

File

Cite this

@inproceedings{d3aba114b4a0404abdae8201ff435816,

title = "A grapheme-based method for automatic alignment of speech and text data",

abstract = "This paper introduces a method for automatic alignment of speech data with unsynchronised, imperfect transcripts, for a domain where no initial acoustic models are available. Using grapheme-based acoustic models, word skip networks and orthographic speech transcripts, we are able to harvest 55% of the speech with a 93% utterance-level accuracy and 99% word accuracy for the produced transcriptions. The work is based on the assumption that there is a high degree of correspondence between the speech and text, and that a full transcription of all of the speech is not required. The method is language independent and the only prior knowledge and resources required are the speech and text transcripts, and a few minor user interventions.",

keywords = "acoustic signal processing, natural language processing, speech synthesis, text analysis, word processing, automatic speech data alignment, automatic text data alignment, grapheme-based acoustic models, language independent method, orthographic speech transcripts, text transcription, unsynchronised-imperfect transcripts, utterance-level accuracy, word accuracy, word skip networks, Acoustics, Data models, Error analysis, Hidden Markov models, Speech, Speech recognition, Training, grapheme-based models, imperfect transcripts, speech alignment, word networks",

author = "A. Stan and P. Bell and S. King",

year = "2012",

doi = "10.1109/SLT.2012.6424237",

language = "English",

isbn = "978-1-4673-5124-9",

pages = "286--290",

booktitle = "Spoken Language Technology Workshop (SLT), 2012 IEEE",

publisher = "Institute of Electrical and Electronics Engineers",

address = "United States",

}

TY - GEN

T1 - A grapheme-based method for automatic alignment of speech and text data

AU - Stan, A.

AU - Bell, P.

AU - King, S.

PY - 2012

Y1 - 2012

N2 - This paper introduces a method for automatic alignment of speech data with unsynchronised, imperfect transcripts, for a domain where no initial acoustic models are available. Using grapheme-based acoustic models, word skip networks and orthographic speech transcripts, we are able to harvest 55% of the speech with a 93% utterance-level accuracy and 99% word accuracy for the produced transcriptions. The work is based on the assumption that there is a high degree of correspondence between the speech and text, and that a full transcription of all of the speech is not required. The method is language independent and the only prior knowledge and resources required are the speech and text transcripts, and a few minor user interventions.

AB - This paper introduces a method for automatic alignment of speech data with unsynchronised, imperfect transcripts, for a domain where no initial acoustic models are available. Using grapheme-based acoustic models, word skip networks and orthographic speech transcripts, we are able to harvest 55% of the speech with a 93% utterance-level accuracy and 99% word accuracy for the produced transcriptions. The work is based on the assumption that there is a high degree of correspondence between the speech and text, and that a full transcription of all of the speech is not required. The method is language independent and the only prior knowledge and resources required are the speech and text transcripts, and a few minor user interventions.

KW - acoustic signal processing

KW - natural language processing

KW - speech synthesis

KW - text analysis

KW - word processing

KW - automatic speech data alignment

KW - automatic text data alignment

KW - grapheme-based acoustic models

KW - language independent method

KW - orthographic speech transcripts

KW - text transcription

KW - unsynchronised-imperfect transcripts

KW - utterance-level accuracy

KW - word accuracy

KW - word skip networks

KW - Acoustics

KW - Data models

KW - Error analysis

KW - Hidden Markov models

KW - Speech

KW - Speech recognition

KW - Training

KW - grapheme-based models

KW - imperfect transcripts

KW - speech alignment

KW - word networks

U2 - 10.1109/SLT.2012.6424237

DO - 10.1109/SLT.2012.6424237

M3 - Conference contribution

SN - 978-1-4673-5124-9

SP - 286

EP - 290

BT - Spoken Language Technology Workshop (SLT), 2012 IEEE

PB - Institute of Electrical and Electronics Engineers

ER -

A grapheme-based method for automatic alignment of speech and text data

Abstract

Keywords / Materials (for Non-textual outputs)

Access to Document

Fingerprint

Projects

Simple4All: Speech synthesis that improves through adaptive learning

HELP4MOOD:A computational distributed system to support the treatment of patients with major depression

Activities

EACL 2014 keynote: Speech synthesis needs YOU!

Cite this