Exploring Capabilities of Monolingual Audio Transformers using Large Datasets in Automatic Speech Recognition of Czech

TL;DR

This paper demonstrates that large-scale pretraining of Czech monolingual audio transformers significantly improves automatic speech recognition, achieving competitive results and strong zero-shot capabilities on multiple datasets.

Contribution

It introduces a large-scale pretraining and fine-tuning framework for Czech monolingual audio transformers, showing their robustness and zero-shot learning ability in ASR tasks.

Findings

Monolingual Wav2Vec 2.0 models outperform traditional systems.

Models effectively leverage large unlabeled and labeled datasets.

Strong zero-shot performance on unseen datasets.

Abstract

In this paper, we present our progress in pretraining Czech monolingual audio transformers from a large dataset containing more than 80 thousand hours of unlabeled speech, and subsequently fine-tuning the model on automatic speech recognition tasks using a combination of in-domain data and almost 6 thousand hours of out-of-domain transcribed speech. We are presenting a large palette of experiments with various fine-tuning setups evaluated on two public datasets (CommonVoice and VoxPopuli) and one extremely challenging dataset from the MALACH project. Our results show that monolingual Wav2Vec 2.0 models are robust ASR systems, which can take advantage of large labeled and unlabeled datasets and successfully compete with state-of-the-art LVCSR systems. Moreover, Wav2Vec models proved to be good zero-shot learners when no training data are available for the target ASR task.