# Towards Improved Speech Recognition through Optimized Synthetic Data Generation

## TL;DR

This paper proposes a method to enhance speech recognition by generating optimized synthetic audio data from text-only sources, aiming to match real data performance and addressing confidentiality constraints.

## Contribution

It introduces an optimized synthetic data generation process using finetuning, filtering, and evaluation to improve ASR training with synthetic speech.

## Key findings

- Synthetic data quality significantly impacts ASR performance.
- Optimized synthetic data can approach real data accuracy in speech recognition.
- Method is validated on Quebec French spontaneous speech datasets.

## Abstract

Supervised training of speech recognition models requires access to transcribed audio data, which often is not possible due to confidentiality issues. Our approach to this problem is to generate synthetic audio from a text-only corpus using a state-of-the-art text-to-speech model with voice cloning capabilities. Our goal is to achieve automatic speech recognition (ASR) performance comparable to models trained on real data. We explore ways to optimize synthetic data generation through finetuning, filtering and evaluation, and its use for training an end-to-end encoder-decoder ASR model. Experiments were conducted using two datasets of spontaneous, conversational speech in Qu\'ebec French. We show that improving data generation leads to large improvements in the final ASR system trained on synthetic data.