FastCorrect: Fast Error Correction with Edit Alignment for Automatic Speech Recognition
Dec 6, 2021
Speakers
Yichong Leng
Speaker · 0 followers
Linchen Zhu
Speaker · 0 followers
Renqian Luo
Speaker · 0 followers
Linquan Liu
Speaker · 0 followers
Xiang-Yang Li
Speaker · 0 followers
Edward Lin
Speaker · 0 followers
Tie-Yan Liu
Speaker · 1 follower
About
Error correction techniques have been used to refine the output sentences from automatic speech recognition (ASR) models and achieve a lower word error rate (WER) than original ASR outputs. Previous works usually use a sequence-to-sequence model to correct an ASR output sentence autoregressively, which causes large latency and cannot be deployed in online ASR services. A straightforward solution to reduce latency, inspired by non-autoregressive (NAR) neural machine translation, is to use an NAR sequence generation model for ASR error correction, which, however, comes at the cost of significantly increased ASR error rate. In this paper, observing distinctive error patterns and correction operations (i.e., insertion, deletion, and substitution) in ASR, we propose FastCorrect, a novel NAR error correction model based on edit alignment. In training, FastCorrect aligns each source token from an ASR output sentence to the target tokens from the corresponding ground-truth sentence based on the edit distance between the source and target sentences, and extracts the number of target tokens corresponding to each source token during edition/correction, which is then used to train a length predictor and to adjust the source tokens to match the length of the target sentence for parallel generation. In inference, the token number predicted by the length predictor is used to adjust the source tokens for target sequence generation. Experiments on the public AISHELL-1 dataset and an internal industrial-scale ASR dataset show the effectiveness of FastCorrect for ASR error correction: 1) it speeds up the inference by 6-9 times and maintains the accuracy (8-14Error correction techniques have been used to refine the output sentences from automatic speech recognition (ASR) models and achieve a lower word error rate (WER) than original ASR outputs. Previous works usually use a sequence-to-sequence model to correct an ASR output sentence autoregressively, which causes large latency and cannot be deployed in online ASR services. A straightforward solution to reduce latency, inspired by non-autoregressive (NAR) neural machine translation, is to use an NAR…
Organizer
NeurIPS 2021
Account · 1.9k followers
About NeurIPS 2021
Neural Information Processing Systems (NeurIPS) is a multi-track machine learning and computational neuroscience conference that includes invited talks, demonstrations, symposia and oral and poster presentations of refereed papers. Following the conference, there are workshops which provide a less formal setting.
Like the format? Trust SlidesLive to capture your next event!
Professional recording and live streaming, delivered globally.
Sharing
Recommended Videos
Presentations on similar topic, category or speaker
Scaling Up Exact Neural Network Compression by ReLU Stability
Watch later
Thiago Serra, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Closing remarks
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Regularized Frank-Wolfe for Dense CRFs: Generalizing Mean Field and Beyond
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Scalable Variational Approaches for Bayesian Causal Discovery
Watch later
Chris Cundy, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
NLP with Synthetic Text
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Control in Dialogue: When does it work?
Watch later
Total of 1 viewers voted for saving the presentation to eternal vault which is 0.1%