Continuously Learning New Words in Automatic Speech Recognition

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Huber, Christian, Waibel, Alexander
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866915127274504192
author Huber, Christian
Waibel, Alexander
author_facet Huber, Christian
Waibel, Alexander
contents Despite recent advances, Automatic Speech Recognition (ASR) systems are still far from perfect. Typical errors include acronyms, named entities, and domain-specific special words for which little or no labeled data is available. To address the problem of recognizing these words, we propose a self-supervised continual learning approach: Given the audio of a lecture talk with the corresponding slides, we bias the model towards decoding new words from the slides by using a memory-enhanced ASR model from the literature. Then, we perform inference on the talk, collecting utterances that contain detected new words into an adaptation data set. Continual learning is then performed by training adaptation weights added to the model on this data set. The whole procedure is iterated for many talks. We show that with this approach, we obtain increasing performance on the new words when they occur more frequently (more than 80% recall) while preserving the general performance of the model.
format Preprint
id arxiv_https___arxiv_org_abs_2401_04482
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Continuously Learning New Words in Automatic Speech Recognition
Huber, Christian
Waibel, Alexander
Computation and Language
Machine Learning
Despite recent advances, Automatic Speech Recognition (ASR) systems are still far from perfect. Typical errors include acronyms, named entities, and domain-specific special words for which little or no labeled data is available. To address the problem of recognizing these words, we propose a self-supervised continual learning approach: Given the audio of a lecture talk with the corresponding slides, we bias the model towards decoding new words from the slides by using a memory-enhanced ASR model from the literature. Then, we perform inference on the talk, collecting utterances that contain detected new words into an adaptation data set. Continual learning is then performed by training adaptation weights added to the model on this data set. The whole procedure is iterated for many talks. We show that with this approach, we obtain increasing performance on the new words when they occur more frequently (more than 80% recall) while preserving the general performance of the model.
title Continuously Learning New Words in Automatic Speech Recognition
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2401.04482