Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Confidence Estimation for Attention-based Sequence-to-sequence Models for Speech Recognition

About

For various speech-related tasks, confidence scores from a speech recogniser are a useful measure to assess the quality of transcriptions. In traditional hidden Markov model-based automatic speech recognition (ASR) systems, confidence scores can be reliably obtained from word posteriors in decoding lattices. However, for an ASR system with an auto-regressive decoder, such as an attention-based sequence-to-sequence model, computing word posteriors is difficult. An obvious alternative is to use the decoder softmax probability as the model confidence. In this paper, we first examine how some commonly used regularisation methods influence the softmax-based confidence scores and study the overconfident behaviour of end-to-end models. Then we propose a lightweight and effective approach named confidence estimation module (CEM) on top of an existing end-to-end ASR model. Experiments on LibriSpeech show that CEM can mitigate the overconfidence problem and can produce more reliable confidence scores with and without shallow fusion of a language model. Further analysis shows that CEM generalises well to speech from a moderately mismatched domain and can potentially improve downstream tasks such as semi-supervised learning.

Qiujia Li, David Qiu, Yu Zhang, Bo Li, Yanzhang He, Philip C. Woodland, Liangliang Cao, Trevor Strohman• 2020

Related benchmarks

TaskDatasetResultRank
Automatic Speech RecognitionSWITCHBOARD swbd
WER12.3
39
Automatic Speech RecognitionCHiME-4 simu (test)
WER9.8
31
Automated Speech RecognitionTED-LIUM V3
WER4.8
26
Automatic Speech RecognitionLRS2-BBC (test)
WER0.079
21
Automatic Speech RecognitionCHiME-4 (dev-real)
WER4.3
21
Automatic Speech RecognitionCHiME-4 (dev-simu)
WER (%)6.7
21
Speech RecognitionCHiME-4 real (test)
WER6.5
18
Automatic Speech RecognitionLibriSpeech-FreeSound (babble)
WER33.5
7
Automatic Speech RecognitionRATS radio
WER45.5
7
Automatic Speech RecognitionCommon Voice African
WER5.4
7
Showing 10 of 17 rows

Other info

Follow for update