Thesis Sistema de reconocimiento de voz a texto de ingredientes y recetas de cocina: audio sense
No Thumbnail Available
Date
2024-05-07
Journal Title
Journal ISSN
Volume Title
Program
Ingeniería en Informática
Departament
Campus
Sede Concepción
Abstract
El presente trabajo aborda la falta de un sistema de reconocimiento de voz
especializado en el contexto culinario, enfocado en ingredientes y recetas. Mediante
técnicas de deep learning, específicamente redes neuronales, se desarrolla una solución
para transcribir instrucciones de voz a texto preciso. Se investigan modelos de vanguardia
como Wav2Vec 2.0 y Transformer. Luego se diseña y entrena una arquitectura neuronal
sobre un conjunto de datos en español. Finalmente, se evalúa con métricas estándar. Los
resultados sentarán bases para una incorporación efectiva en la app FoodieSpal, mejorando
la experiencia gastronómica de los usuarios.
This paper addresses the lack of a speech recognition system specialized in the culinary context, focused on ingredients and recipes. Using deep learning techniques, specifically neural networks, a solution is developed to transcribe voice instructions into accurate text. State-of-the-art models such as Wav2Vec 2.0 and Transformer are investigated. Then, a neural architecture is designed and trained on a Spanish data set. Finally, it is evaluated with standard metrics. The results will lay the groundwork for an effective incorporation into the FoodieSpal app, improving the users' dining experience.
This paper addresses the lack of a speech recognition system specialized in the culinary context, focused on ingredients and recipes. Using deep learning techniques, specifically neural networks, a solution is developed to transcribe voice instructions into accurate text. State-of-the-art models such as Wav2Vec 2.0 and Transformer are investigated. Then, a neural architecture is designed and trained on a Spanish data set. Finally, it is evaluated with standard metrics. The results will lay the groundwork for an effective incorporation into the FoodieSpal app, improving the users' dining experience.
Description
Keywords
Reconocimiento de voz, Deep learning, Redes neuronales, Ajuste fino, Ingredientes
