SegAug: CTC-Aligned Segmented Augmentation For Robust RNN-Transducer Based Speech Recognition

TL;DR

SegAug tackles deletion errors in RNN-T by decoupling acoustic sensitivity from internal language modeling through CTC-aligned segmentation and targeted on-the-fly augmentations. It uses four sub-augmentations—SegDrop, SegPerm, SegCrop, SegMix—applied to waveform data with CTC alignments to generate diverse, low-semantic audio-text pairs, improving robustness and reducing deletion error rates. Empirical results on LibriSpeech and Tedlium-v3 show meaningful WER gains, especially through deletion reductions, and SegAug generalizes to AED while exhibiting limited benefit for CTC. This work offers a scalable, low-overhead augmentation strategy to strengthen end-to-end ASR systems across varied data regimes.

Abstract

RNN-Transducer (RNN-T) is a widely adopted architecture in speech recognition, integrating acoustic and language modeling in an end-to-end framework. However, the RNN-T predictor tends to over-rely on consecutive word dependencies in training data, leading to high deletion error rates, particularly with less common or out-of-domain phrases. Existing solutions, such as regularization and data augmentation, often compromise other aspects of performance. We propose SegAug, an alignment-based augmentation technique that generates contextually varied audio-text pairs with low sentence-level semantics. This method encourages the model to focus more on acoustic features while diversifying the learned textual patterns of its internal language model, thereby reducing deletion errors and enhancing overall performance. Evaluations on the LibriSpeech and Tedlium-v3 datasets demonstrate a relative WER reduction of up to 12.5% on small-scale and 6.9% on large-scale settings. Notably, most of the improvement stems from reduced deletion errors, with relative reductions of 45.4% and 18.5%, respectively. These results highlight SegAug's effectiveness in improving RNN-T's robustness, offering a promising solution for enhancing speech recognition performance across diverse and challenging scenarios.

Figures (1)

• Figure 1: Illustration of SegAug transformation methods.