AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Revisiting the Entropy Semiring for Neural Speech Recognition

Oscar Chang, Dongseong Hwang, Olivier Siohan

ICLR 2023 · Label-efficient and robust learning › Knowledge distillation · approach: Distillation for machine translation · 1 citations (OpenAlex)

Read the paper

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
Add alignment entropy regularizationCTC LSTM, 22M · Librispeech Test-CleanWER: 7.7 ± 0.3 -> 7.2 ± 0.3Improved
Add alignment entropy regularizationCTC Conformer, 9M · Librispeech Test-CleanWER: 4.1 ± 0.2 -> 4.1 ± 0.2No clear effect
Add alignment entropy regularizationRNN-T LSTM, 25M · Librispeech Test-CleanWER: 7.4 ± 0.3 -> 7.2 ± 0.3Improved
Add alignment entropy regularizationRNN-T Conformer, 10M · Librispeech Test-CleanWER: 2.8 ± 0.2 -> 2.7 ± 0.2Improved
Add alignment entropy regularizationCTC LSTM, 22M · Librispeech Test-OtherWER: 21.8 ± 0.6 -> 20.8 ± 0.6Improved
Add alignment entropy regularizationCTC Conformer, 9M · Librispeech Test-OtherWER: 10.2 ± 0.4 -> 9.9 ± 0.4Improved
Add alignment entropy regularizationRNN-T LSTM, 25M · Librispeech Test-OtherWER: 24.0 ± 0.8 -> 23.1 ± 0.7Improved
Add alignment entropy regularizationRNN-T Conformer, 10M · Librispeech Test-OtherWER: 6.7 ± 0.3 -> 6.5 ± 0.3Improved
Add token-space state-wise distillation0.6B causal Conformer RNN-T · Librispeech Test-CleanWER: 2.7 ± 0.2 -> 2.1 ± 0.2Improved
Add token-space state-wise distillation0.6B causal Conformer RNN-T · Relative latencylatency (ms): 0.0 -> +2.7Hurt
Remove token-space state-wise distillation, keeping alignment distillation (ablation)Semiring distillation, 0.6B causal Conformer RNN-T · Relative latencylatency (ms): -5.2 -> -65.7Improved
Remove token-space state-wise distillation, keeping alignment distillation (ablation)Semiring distillation, 0.6B causal Conformer RNN-T · Librispeech Test-OtherWER: 5.2 ± 0.3 -> 5.8 ± 0.3Hurt

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