Improving sequence-to-sequence speech recognition training with on-the-fly data augmentation

TL;DR

This paper explores on-the-fly data augmentation techniques to improve sequence-to-sequence speech recognition models, demonstrating state-of-the-art results without relying on extra text data.

Contribution

It introduces two novel data augmentation methods—time perturbation in the frequency domain and sub-sequence sampling—for S2S ASR models.

Findings

Achieved state-of-the-art performance on Switchboard and Fisher datasets.

Demonstrated effectiveness of data augmentation in reducing overfitting.

Models trained solely on speech data without additional text data.

Abstract

Sequence-to-Sequence (S2S) models recently started to show state-of-the-art performance for automatic speech recognition (ASR). With these large and deep models overfitting remains the largest problem, outweighing performance improvements that can be obtained from better architectures. One solution to the overfitting problem is increasing the amount of available training data and the variety exhibited by the training data with the help of data augmentation. In this paper we examine the influence of three data augmentation methods on the performance of two S2S model architectures. One of the data augmentation method comes from literature, while two other methods are our own development - a time perturbation in the frequency domain and sub-sequence sampling. Our experiments on Switchboard and Fisher data show state-of-the-art performance for S2S models that are trained solely on the speech training data and do not use additional text data.