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

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

Resultado de la investigación: Capítulo del libro/informe/acta de congresoContribución de conferenciarevisión exhaustiva

1 Cita (Scopus)

Resumen

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.

Idioma originalInglés
Título de la publicación alojadaSmart Technologies, Systems and Applications - 1st International Conference, SmartTech-IC 2019, Proceedings
EditoresFabián R. Narváez, Diego F. Vallejo, Paulina A. Morillo, Julio R. Proaño
EditorialSpringer
Páginas165-175
Número de páginas11
ISBN (versión impresa)9783030467845
DOI
EstadoPublicada - 1 ene. 2020
Evento1st International Conference on Smart Technologies, Systems and Applications, SmartTech-IC 2019 - Quito, Ecuador
Duración: 2 dic. 20194 dic. 2019

Serie de la publicación

NombreCommunications in Computer and Information Science
Volumen1154 CCIS
ISSN (versión impresa)1865-0929
ISSN (versión digital)1865-0937

Conferencia

Conferencia1st International Conference on Smart Technologies, Systems and Applications, SmartTech-IC 2019
País/TerritorioEcuador
CiudadQuito
Período2/12/194/12/19

Nota bibliográfica

Publisher Copyright:
© Springer Nature Switzerland AG 2020.

Huella

Profundice en los temas de investigación de 'Incorporation of Language Discriminative Information into Recurrent Neural Networks Models to LID Tasks'. En conjunto forman una huella única.

Citar esto