Failing Forward: Improving Generative Error Correction for ASR with Synthetic Data and Retrieval Augmentation
This addresses a key limitation in ASR post-processing for improving accuracy across diverse domains and languages, though it is incremental as it builds on existing GEC methods.
The paper tackles the problem of generative error correction (GEC) for ASR systems struggling to generalize to unseen errors, especially in out-of-domain scenarios and with named entities, by proposing DARAG, which uses synthetic data and retrieval augmentation, resulting in 8-30% relative WER improvements in-domain and 10-33% out-of-domain.
Generative Error Correction (GEC) has emerged as a powerful post-processing method to enhance the performance of Automatic Speech Recognition (ASR) systems. However, we show that GEC models struggle to generalize beyond the specific types of errors encountered during training, limiting their ability to correct new, unseen errors at test time, particularly in out-of-domain (OOD) scenarios. This phenomenon amplifies with named entities (NEs), where, in addition to insufficient contextual information or knowledge about the NEs, novel NEs keep emerging. To address these issues, we propose DARAG (Data- and Retrieval-Augmented Generative Error Correction), a novel approach designed to improve GEC for ASR in in-domain (ID) and OOD scenarios. We augment the GEC training dataset with synthetic data generated by prompting LLMs and text-to-speech models, thereby simulating additional errors from which the model can learn. For OOD scenarios, we simulate test-time errors from new domains similarly and in an unsupervised fashion. Additionally, to better handle named entities, we introduce retrieval-augmented correction by augmenting the input with entities retrieved from a database. Our approach is simple, scalable, and both domain- and language-agnostic. We experiment on multiple datasets and settings, showing that DARAG outperforms all our baselines, achieving 8\% -- 30\% relative WER improvements in ID and 10\% -- 33\% improvements in OOD settings.