AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

FastCorrect: Fast Error Correction with Edit Alignment for Automatic Speech Recognition

Yichong Leng, Xu Tan, Linchen Zhu, Jin Xu, Renqian Luo, Linquan Liu, Tao Qin, Xiangyang Li, Edward Lin, Tie-Yan Liu

NeurIPS 2021 · Architectures and efficiency › RNNs and normalization · approach: Non-autoregressive translation · 29 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
remove edit alignment (use soft length-difference predictor) (ablation)FastCorrect · AISHELL-1 testWER (%): 4.16 -> 4.67Hurt
remove edit alignment (ablation)FastCorrect · Internal dataset testWER (%): 10.27 -> 12.27Hurt
remove pre-training on pseudo correction data (ablation)FastCorrect · AISHELL-1 testWER (%): 4.16 -> 4.83Hurt
remove pre-training (ablation)FastCorrect · Internal dataset testWER (%): 10.27 -> 10.33Hurt
remove pre-training (ablation)AR model · AISHELL-1 testWER (%): 4.08 -> 16.01Hurt
remove pre-training (ablation)AR model · Internal dataset testWER (%): 10.22 -> 10.26No clear effect
remove fine-tuning (pre-training only) (ablation)FastCorrect · AISHELL-1 testWER (%): 4.16 -> 5.19Hurt
remove fine-tuning (ablation)FastCorrect · Internal dataset testWER (%): 10.27 -> 11.74Hurt
use edit-alignment NAR FastCorrect instead of LevTcorrection model · Internal dataset testWER (%): 11.26 -> 10.27Improved
use FastCorrect instead of FELIX greedy alignmentcorrection model · AISHELL-1 testWER (%): 4.63 -> 4.16Improved
use FastCorrect NAR instead of AR with deep encoder shallow decoder (10-2)correction model · AISHELL-1 testGPU latency (ms/sent): 84.0 -> 21.2Improved
use NAR FastCorrect instead of AR modelcorrection model · AISHELL-1 testWER (%): 4.08 -> 4.16Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy