Generative Annotation for ASR Named Entity Correction
This addresses catastrophic failures in downstream tasks for ASR systems by improving named entity accuracy, though it is incremental as it builds on existing NEC methods.
The paper tackles the problem of named entity correction in ASR transcripts, especially when word forms differ significantly, by proposing a generative method that uses speech sound features to retrieve and replace entities, resulting in significant improvements in entity accuracy as tested on open-source and self-constructed datasets.
End-to-end automatic speech recognition systems often fail to transcribe domain-specific named entities, causing catastrophic failures in downstream tasks. Numerous fast and lightweight named entity correction (NEC) models have been proposed in recent years. These models, mainly leveraging phonetic-level edit distance algorithms, have shown impressive performances. However, when the forms of the wrongly-transcribed words(s) and the ground-truth entity are significantly different, these methods often fail to locate the wrongly transcribed words in hypothesis, thus limiting their usage. We propose a novel NEC method that utilizes speech sound features to retrieve candidate entities. With speech sound features and candidate entities, we inovatively design a generative method to annotate entity errors in ASR transcripts and replace the text with correct entities. This method is effective in scenarios of word form difference. We test our method using open-source and self-constructed test sets. The results demonstrate that our NEC method can bring significant improvement to entity accuracy. The self-constructed training data and test set is publicly available at github.com/L6-NLP/Generative-Annotation-NEC.