LANGUAGE IDENTIFICATION FROM PROSODY WITHOUT EXPLICIT FEATURES

Fred Cummins; Felix Gers; Jürgen Schmidhuber

LANGUAGE IDENTIFICATION FROM PROSODY WITHOUT EXPLICIT FEATURES

Fred Cummins, Felix Gers, Jürgen Schmidhuber

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

Abstract

Most current language identification (LID) systems make little or no use of prosodic information, despite the importance of prosody in LID by humans. The greatest obstacle has been that of finding an appropriate feature set which captures linguistically relevant prosodic information. The only system to attempt LID entirely on the basis of prosodic variables uses a set of over 200 features which are selected and combined in a task-specific manner [12]. We apply a novel recurrent neural network model to the task of pairwise discrimination among languages. Network inputs are limited to delta-Fo and the first difference of the band limited amplitude envelope. Initial results are based on all pairwise combinations of English, German, Japanese, Mandarin and Spanish, with 90 speakers per language.

Original language	English (US)
Title of host publication	6th European Conference on Speech Communication and Technology, EUROSPEECH 1999
Publisher	The International Society for Computers and Their Applications (ISCA)
Pages	371-374
Number of pages	4
State	Published - Jan 1 1999
Externally published	Yes

Cite this

@inproceedings{c2a0d0cf5d2c449a8e69d9aaa39c6636,

title = "LANGUAGE IDENTIFICATION FROM PROSODY WITHOUT EXPLICIT FEATURES",

abstract = "Most current language identification (LID) systems make little or no use of prosodic information, despite the importance of prosody in LID by humans. The greatest obstacle has been that of finding an appropriate feature set which captures linguistically relevant prosodic information. The only system to attempt LID entirely on the basis of prosodic variables uses a set of over 200 features which are selected and combined in a task-specific manner [12]. We apply a novel recurrent neural network model to the task of pairwise discrimination among languages. Network inputs are limited to delta-Fo and the first difference of the band limited amplitude envelope. Initial results are based on all pairwise combinations of English, German, Japanese, Mandarin and Spanish, with 90 speakers per language.",

author = "Fred Cummins and Felix Gers and J{\"u}rgen Schmidhuber",

note = "Generated from Scopus record by KAUST IRTS on 2022-09-14",

year = "1999",

month = jan,

day = "1",

language = "English (US)",

pages = "371--374",

booktitle = "6th European Conference on Speech Communication and Technology, EUROSPEECH 1999",

publisher = "The International Society for Computers and Their Applications (ISCA)",

}

LANGUAGE IDENTIFICATION FROM PROSODY WITHOUT EXPLICIT FEATURES. / Cummins, Fred; Gers, Felix; Schmidhuber, Jürgen.
6th European Conference on Speech Communication and Technology, EUROSPEECH 1999. The International Society for Computers and Their Applications (ISCA), 1999. p. 371-374.

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

TY - GEN

T1 - LANGUAGE IDENTIFICATION FROM PROSODY WITHOUT EXPLICIT FEATURES

AU - Cummins, Fred

AU - Gers, Felix

AU - Schmidhuber, Jürgen

N1 - Generated from Scopus record by KAUST IRTS on 2022-09-14

PY - 1999/1/1

Y1 - 1999/1/1

N2 - Most current language identification (LID) systems make little or no use of prosodic information, despite the importance of prosody in LID by humans. The greatest obstacle has been that of finding an appropriate feature set which captures linguistically relevant prosodic information. The only system to attempt LID entirely on the basis of prosodic variables uses a set of over 200 features which are selected and combined in a task-specific manner [12]. We apply a novel recurrent neural network model to the task of pairwise discrimination among languages. Network inputs are limited to delta-Fo and the first difference of the band limited amplitude envelope. Initial results are based on all pairwise combinations of English, German, Japanese, Mandarin and Spanish, with 90 speakers per language.

AB - Most current language identification (LID) systems make little or no use of prosodic information, despite the importance of prosody in LID by humans. The greatest obstacle has been that of finding an appropriate feature set which captures linguistically relevant prosodic information. The only system to attempt LID entirely on the basis of prosodic variables uses a set of over 200 features which are selected and combined in a task-specific manner [12]. We apply a novel recurrent neural network model to the task of pairwise discrimination among languages. Network inputs are limited to delta-Fo and the first difference of the band limited amplitude envelope. Initial results are based on all pairwise combinations of English, German, Japanese, Mandarin and Spanish, with 90 speakers per language.

UR - http://www.scopus.com/inward/record.url?scp=0039222571&partnerID=8YFLogxK

M3 - Conference contribution

SP - 371

EP - 374

BT - 6th European Conference on Speech Communication and Technology, EUROSPEECH 1999

PB - The International Society for Computers and Their Applications (ISCA)

ER -

LANGUAGE IDENTIFICATION FROM PROSODY WITHOUT EXPLICIT FEATURES

Abstract

Other files and links

Fingerprint

Cite this