Salvato in:
Dettagli Bibliografici
Autori principali: Jain, Yash, Chan, David, Dheram, Pranav, Khare, Aparna, Shonibare, Olabanji, Ravichandran, Venkatesh, Ghosh, Shalini
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2403.19822
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913289571663872
author Jain, Yash
Chan, David
Dheram, Pranav
Khare, Aparna
Shonibare, Olabanji
Ravichandran, Venkatesh
Ghosh, Shalini
author_facet Jain, Yash
Chan, David
Dheram, Pranav
Khare, Aparna
Shonibare, Olabanji
Ravichandran, Venkatesh
Ghosh, Shalini
contents Recent advances in machine learning have demonstrated that multi-modal pre-training can improve automatic speech recognition (ASR) performance compared to randomly initialized models, even when models are fine-tuned on uni-modal tasks. Existing multi-modal pre-training methods for the ASR task have primarily focused on single-stage pre-training where a single unsupervised task is used for pre-training followed by fine-tuning on the downstream task. In this work, we introduce a novel method combining multi-modal and multi-task unsupervised pre-training with a translation-based supervised mid-training approach. We empirically demonstrate that such a multi-stage approach leads to relative word error rate (WER) improvements of up to 38.45% over baselines on both Librispeech and SUPERB. Additionally, we share several important findings for choosing pre-training methods and datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2403_19822
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Multi-Stage Multi-Modal Pre-Training for Automatic Speech Recognition
Jain, Yash
Chan, David
Dheram, Pranav
Khare, Aparna
Shonibare, Olabanji
Ravichandran, Venkatesh
Ghosh, Shalini
Computation and Language
Artificial Intelligence
Recent advances in machine learning have demonstrated that multi-modal pre-training can improve automatic speech recognition (ASR) performance compared to randomly initialized models, even when models are fine-tuned on uni-modal tasks. Existing multi-modal pre-training methods for the ASR task have primarily focused on single-stage pre-training where a single unsupervised task is used for pre-training followed by fine-tuning on the downstream task. In this work, we introduce a novel method combining multi-modal and multi-task unsupervised pre-training with a translation-based supervised mid-training approach. We empirically demonstrate that such a multi-stage approach leads to relative word error rate (WER) improvements of up to 38.45% over baselines on both Librispeech and SUPERB. Additionally, we share several important findings for choosing pre-training methods and datasets.
title Multi-Stage Multi-Modal Pre-Training for Automatic Speech Recognition
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2403.19822