Text Generation with Speech Synthesis for ASR Data Augmentation

TL;DR

This paper investigates using large-scale neural networks for text augmentation in ASR data synthesis, converting augmented texts to speech, and demonstrates significant WER improvements over traditional methods across multiple datasets.

Contribution

It introduces a neural text augmentation approach for ASR data synthesis and systematically compares it to traditional methods, showing superior performance.

Findings

Neural text augmentation achieves 9%-15% relative WER reduction.

Synthetic speech generated from augmented texts improves ASR accuracy.

Neural methods outperform traditional text augmentation techniques.

Abstract

Aiming at reducing the reliance on expensive human annotations, data synthesis for Automatic Speech Recognition (ASR) has remained an active area of research. While prior work mainly focuses on synthetic speech generation for ASR data augmentation, its combination with text generation methods is considerably less explored. In this work, we explore text augmentation for ASR using large-scale pre-trained neural networks, and systematically compare those to traditional text augmentation methods. The generated synthetic texts are then converted to synthetic speech using a text-to-speech (TTS) system and added to the ASR training data. In experiments conducted on three datasets, we find that neural models achieve 9%-15% relative WER improvement and outperform traditional methods. We conclude that text augmentation, particularly through modern neural approaches, is a viable tool for improving the accuracy of ASR systems.