Catalan Accent Classification by Voice using Deep Learning

Speech characterization is a vital field in artificial intelligence, yet accent classification is often overlooked, particularly for the Catalan language. This project is centered on the classification of Catalan accents using the Catalan Common Voice dataset. We lay significant emphasis on our data...

ver descrição completa

Detalhes bibliográficos
Autor: Felip I Díaz, Bernat
Formato: tesis de maestría
Fecha de publicación:2023
País:España
Recursos:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/399370
Acesso em linha:https://hdl.handle.net/2117/399370
Access Level:acceso abierto
Palavra-chave:Machine learning
Artificial intelligence
Automatic speech recognition
artificial intelligence
deep learning
automatic speech recognition
accent classification
inteligencia artificial
aprendizaje profundo
reconocimiento automático del habla
clasificación de acento
Aprenentatge automàtic
Intel·ligència artificial
Reconeixement automàtic de la parla
Àrees temàtiques de la UPC::Enginyeria de la telecomunicació::Processament del senyal::Processament de la parla i del senyal acústic
Descrição
Resumo:Speech characterization is a vital field in artificial intelligence, yet accent classification is often overlooked, particularly for the Catalan language. This project is centered on the classification of Catalan accents using the Catalan Common Voice dataset. We lay significant emphasis on our data processing pipeline, striving to ensure the quality and accuracy of the dataset for both training and validation phases. A novel aspect of our approach is the application of Double Multi-Head Self-Attention Pooling, diverging from the traditional statistical pooling methods typically employed in this task. This methodology enables effective pooling and dimensionality reduction of the feature vector, thereby boosting the efficiency of our models. We conduct various experiments to explore the optimal utilization of available data and to fine-tune our model for improved results.