Less is More: Data Curation Matters in Scaling Speech Enhancement

TL;DR

This paper shows that in speech enhancement, carefully selecting high-quality training data can outperform using larger, noisier datasets, emphasizing the importance of data curation over sheer volume.

Contribution

It demonstrates that prioritizing high-quality data in large datasets can lead to better speech enhancement models than simply increasing data size.

Findings

Models trained on 700 hours of curated data outperform those trained on 2,500 hours of uncurated data.

Quality-focused data curation is more effective than data quantity in scaling speech enhancement.

Scaling with high-quality data yields better performance than using larger, noisier datasets.

Abstract

The vast majority of modern speech enhancement systems rely on data-driven neural network models. Conventionally, larger datasets are presumed to yield superior model performance, an observation empirically validated across numerous tasks in other domains. However, recent studies reveal diminishing returns when scaling speech enhancement data. We focus on a critical factor: prevalent quality issues in ``clean'' training labels within large-scale datasets. This work re-examines this phenomenon and demonstrates that, within large-scale training sets, prioritizing high-quality training data is more important than merely expanding the data volume. Experimental findings suggest that models trained on a carefully curated subset of 700 hours can outperform models trained on the 2,500-hour full dataset. This outcome highlights the crucial role of data curation in scaling speech enhancement systems effectively.