Incorporation of Language Discriminative Information into Recurrent Neural Networks Models to LID Tasks

Christian Salamea; Ricardo Cordoba; Luis D’Haro; David Romero

doi:10.1007/978-3-030-46785-2_14

Incorporation of Language Discriminative Information into Recurrent Neural Networks Models to LID Tasks

Christian Salamea, Ricardo Cordoba, Luis D’Haro, David Romero

Research Group on Interaction, Robotics and Automatics (GIIRA)

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

1 Scopus citations

Abstract

Language Identification (LID) is an essential research topic in the Automatic Recognition Speech area. One of the most important characteristics relative to language is context information. In this article, considering a phonotactic approach where the phonetic units called “phone-grams” are used, in order to introduce such context information, a novel technique is proposed. Language discriminative information has been incorporated in the Recurrent Neural Network Language Models generation (RNNLMs) in the weights initialization stage to improve the Language Identification task. This technique has been evaluated using KALAKA-3 database that contains 108 h of audios of six languages to be recognized. The metric used in this work has been the Average Detection Cost metric C_avg. In relation to the phonetic units called “phone-grams” used in order to incorporate context information in the features used to train the RNNLM, it has been considered phone-grams of two elements “2phone-grams” and three elements “3phone-grams”, obtaining a relative improvement up to 17% and 15,44% respectively compared to the results obtaining using RNNLMs.

Original language	English
Title of host publication	Smart Technologies, Systems and Applications - 1st International Conference, SmartTech-IC 2019, Proceedings
Editors	Fabián R. Narváez, Diego F. Vallejo, Paulina A. Morillo, Julio R. Proaño
Publisher	Springer
Pages	165-175
Number of pages	11
ISBN (Print)	9783030467845
DOIs	https://doi.org/10.1007/978-3-030-46785-2_14
State	Published - 1 Jan 2020
Event	1st International Conference on Smart Technologies, Systems and Applications, SmartTech-IC 2019 - Quito, Ecuador Duration: 2 Dec 2019 → 4 Dec 2019

Publication series

Name	Communications in Computer and Information Science
Volume	1154 CCIS
ISSN (Print)	1865-0929
ISSN (Electronic)	1865-0937

Conference

Conference	1st International Conference on Smart Technologies, Systems and Applications, SmartTech-IC 2019
Country/Territory	Ecuador
City	Quito
Period	2/12/19 → 4/12/19

Bibliographical note

Publisher Copyright:
© Springer Nature Switzerland AG 2020.

Keywords

Automatic Recognition Speech
Language discriminative information
Language Identification
Recurrent Neural Networks

Access to Document

10.1007/978-3-030-46785-2_14

Cite this

Salamea, C., Cordoba, R., D’Haro, L., & Romero, D. (2020). Incorporation of Language Discriminative Information into Recurrent Neural Networks Models to LID Tasks. In F. R. Narváez, D. F. Vallejo, P. A. Morillo, & J. R. Proaño (Eds.), Smart Technologies, Systems and Applications - 1st International Conference, SmartTech-IC 2019, Proceedings (pp. 165-175). (Communications in Computer and Information Science; Vol. 1154 CCIS). Springer. https://doi.org/10.1007/978-3-030-46785-2_14

Salamea, Christian ; Cordoba, Ricardo ; D’Haro, Luis et al. / Incorporation of Language Discriminative Information into Recurrent Neural Networks Models to LID Tasks. Smart Technologies, Systems and Applications - 1st International Conference, SmartTech-IC 2019, Proceedings. editor / Fabián R. Narváez ; Diego F. Vallejo ; Paulina A. Morillo ; Julio R. Proaño. Springer, 2020. pp. 165-175 (Communications in Computer and Information Science).

@inproceedings{abbc8834f2b4477ea39a0569c5cd9ccd,

title = "Incorporation of Language Discriminative Information into Recurrent Neural Networks Models to LID Tasks",

abstract = "Language Identification (LID) is an essential research topic in the Automatic Recognition Speech area. One of the most important characteristics relative to language is context information. In this article, considering a phonotactic approach where the phonetic units called “phone-grams” are used, in order to introduce such context information, a novel technique is proposed. Language discriminative information has been incorporated in the Recurrent Neural Network Language Models generation (RNNLMs) in the weights initialization stage to improve the Language Identification task. This technique has been evaluated using KALAKA-3 database that contains 108 h of audios of six languages to be recognized. The metric used in this work has been the Average Detection Cost metric Cavg. In relation to the phonetic units called “phone-grams” used in order to incorporate context information in the features used to train the RNNLM, it has been considered phone-grams of two elements “2phone-grams” and three elements “3phone-grams”, obtaining a relative improvement up to 17% and 15,44% respectively compared to the results obtaining using RNNLMs.",

keywords = "Automatic Recognition Speech, Language discriminative information, Language Identification, Recurrent Neural Networks",

author = "Christian Salamea and Ricardo Cordoba and Luis D{\textquoteright}Haro and David Romero",

note = "Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2020.; 1st International Conference on Smart Technologies, Systems and Applications, SmartTech-IC 2019 ; Conference date: 02-12-2019 Through 04-12-2019",

year = "2020",

month = jan,

day = "1",

doi = "10.1007/978-3-030-46785-2_14",

language = "English",

isbn = "9783030467845",

series = "Communications in Computer and Information Science",

publisher = "Springer",

pages = "165--175",

editor = "Narv{\'a}ez, {Fabi{\'a}n R.} and Vallejo, {Diego F.} and Morillo, {Paulina A.} and Proa{\~n}o, {Julio R.}",

booktitle = "Smart Technologies, Systems and Applications - 1st International Conference, SmartTech-IC 2019, Proceedings",

}

Salamea, C, Cordoba, R, D’Haro, L & Romero, D 2020, Incorporation of Language Discriminative Information into Recurrent Neural Networks Models to LID Tasks. in FR Narváez, DF Vallejo, PA Morillo & JR Proaño (eds), Smart Technologies, Systems and Applications - 1st International Conference, SmartTech-IC 2019, Proceedings. Communications in Computer and Information Science, vol. 1154 CCIS, Springer, pp. 165-175, 1st International Conference on Smart Technologies, Systems and Applications, SmartTech-IC 2019, Quito, Ecuador, 2/12/19. https://doi.org/10.1007/978-3-030-46785-2_14

Incorporation of Language Discriminative Information into Recurrent Neural Networks Models to LID Tasks. / Salamea, Christian; Cordoba, Ricardo; D’Haro, Luis et al.
Smart Technologies, Systems and Applications - 1st International Conference, SmartTech-IC 2019, Proceedings. ed. / Fabián R. Narváez; Diego F. Vallejo; Paulina A. Morillo; Julio R. Proaño. Springer, 2020. p. 165-175 (Communications in Computer and Information Science; Vol. 1154 CCIS).

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

TY - GEN

T1 - Incorporation of Language Discriminative Information into Recurrent Neural Networks Models to LID Tasks

AU - Salamea, Christian

AU - Cordoba, Ricardo

AU - D’Haro, Luis

AU - Romero, David

N1 - Publisher Copyright: © Springer Nature Switzerland AG 2020.

PY - 2020/1/1

Y1 - 2020/1/1

N2 - Language Identification (LID) is an essential research topic in the Automatic Recognition Speech area. One of the most important characteristics relative to language is context information. In this article, considering a phonotactic approach where the phonetic units called “phone-grams” are used, in order to introduce such context information, a novel technique is proposed. Language discriminative information has been incorporated in the Recurrent Neural Network Language Models generation (RNNLMs) in the weights initialization stage to improve the Language Identification task. This technique has been evaluated using KALAKA-3 database that contains 108 h of audios of six languages to be recognized. The metric used in this work has been the Average Detection Cost metric Cavg. In relation to the phonetic units called “phone-grams” used in order to incorporate context information in the features used to train the RNNLM, it has been considered phone-grams of two elements “2phone-grams” and three elements “3phone-grams”, obtaining a relative improvement up to 17% and 15,44% respectively compared to the results obtaining using RNNLMs.

AB - Language Identification (LID) is an essential research topic in the Automatic Recognition Speech area. One of the most important characteristics relative to language is context information. In this article, considering a phonotactic approach where the phonetic units called “phone-grams” are used, in order to introduce such context information, a novel technique is proposed. Language discriminative information has been incorporated in the Recurrent Neural Network Language Models generation (RNNLMs) in the weights initialization stage to improve the Language Identification task. This technique has been evaluated using KALAKA-3 database that contains 108 h of audios of six languages to be recognized. The metric used in this work has been the Average Detection Cost metric Cavg. In relation to the phonetic units called “phone-grams” used in order to incorporate context information in the features used to train the RNNLM, it has been considered phone-grams of two elements “2phone-grams” and three elements “3phone-grams”, obtaining a relative improvement up to 17% and 15,44% respectively compared to the results obtaining using RNNLMs.

KW - Automatic Recognition Speech

KW - Language discriminative information

KW - Language Identification

KW - Recurrent Neural Networks

UR - http://www.scopus.com/inward/record.url?scp=85084830220&partnerID=8YFLogxK

UR - https://www.mendeley.com/catalogue/1ce03c5a-c7a5-3282-aae3-e41b309daf69/

U2 - 10.1007/978-3-030-46785-2_14

DO - 10.1007/978-3-030-46785-2_14

M3 - Conference contribution

AN - SCOPUS:85084830220

SN - 9783030467845

T3 - Communications in Computer and Information Science

SP - 165

EP - 175

BT - Smart Technologies, Systems and Applications - 1st International Conference, SmartTech-IC 2019, Proceedings

A2 - Narváez, Fabián R.

A2 - Vallejo, Diego F.

A2 - Morillo, Paulina A.

A2 - Proaño, Julio R.

PB - Springer

T2 - 1st International Conference on Smart Technologies, Systems and Applications, SmartTech-IC 2019

Y2 - 2 December 2019 through 4 December 2019

ER -

Salamea C, Cordoba R, D’Haro L, Romero D. Incorporation of Language Discriminative Information into Recurrent Neural Networks Models to LID Tasks. In Narváez FR, Vallejo DF, Morillo PA, Proaño JR, editors, Smart Technologies, Systems and Applications - 1st International Conference, SmartTech-IC 2019, Proceedings. Springer. 2020. p. 165-175. (Communications in Computer and Information Science). doi: 10.1007/978-3-030-46785-2_14

Incorporation of Language Discriminative Information into Recurrent Neural Networks Models to LID Tasks

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this