An Approach to Improve Robustness of NLP Systems against ASR Errors
arXiv:2103.13610
Abstract
Speech-enabled systems typically first convert audio to text through an automatic speech recognition (ASR) model and then feed the text to downstream natural language processing (NLP) modules. The errors of the ASR system can seriously downgrade the performance of the NLP modules. Therefore, it is essential to make them robust to the ASR errors. Previous work has shown it is effective to employ data augmentation methods to solve this problem by injecting ASR noise during the training process. In this paper, we utilize the prevalent pre-trained language model to generate training samples with ASR-plausible noise. Compare to the previous methods, our approach generates ASR noise that better fits the real-world error distribution. Experimental results on spoken language translation(SLT) and spoken language understanding (SLU) show that our approach effectively improves the system robustness against the ASR errors and achieves state-of-the-art results on both tasks.
arXiv comment9 pages, 3 figures
References in corpus (12)
- Distributed Representations of Words and Phrases and their Compositionality
- The Curious Case of Neural Text Degeneration
- Achieving Human Parity on Automatic Chinese to English News Translation
- Synthetic and Natural Noise Both Break Neural Machine Translation
- UniLMv2: Pseudo-Masked Language Models for Unified Language Model Pre-Training
- EDA: Easy Data Augmentation Techniques for Boosting Performance on Text Classification Tasks
- CoVoST 2 and Massively Multilingual Speech-to-Text Translation
- GPT-based Generation for Classical Chinese Poetry
- Improving Spoken Language Understanding By Exploiting ASR N-best Hypotheses
- Improving the Robustness of Speech Translation
- Data Augmentation for Training Dialog Models Robust to Speech Recognition Errors
- Robust Neural Machine Translation with Joint Textual and Phonetic Embedding