AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Star Temporal Classification: Sequence Modeling with Partially Labeled Data

Vineel Pratap, Awni Hannun, Gabriel Synnaeve, Ronan Collobert

NeurIPS 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Partial and multi-label learning · 0 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
Use STC instead of CTC on partially labeled dataTransformer · LibriSpeech p_drop=0.1test WER clean/other, greedy: 6.2/11.0 -> 4.1/9.0Improved
Use STC instead of CTC on partially labeled dataTransformer · LibriSpeech p_drop=0.4test WER clean/other, greedy: 48.2/56.9 -> 4.5/10.2Improved
Use STC instead of CTC on partially labeled dataTransformer · LibriSpeech p_drop=0.7test WER clean/other, greedy: 100/100 -> 6.9/15.1Improved
Use STC instead of CTC with p_drop varied by sample splitTransformer · LibriSpeech samples split in 3test WER clean/other, greedy: 58.6/63.9 -> 4.9/11.1Improved
Use STC instead of CTC with p_drop varied by word splitTransformer · LibriSpeech words split in 3test WER clean/other, greedy: 45.6/49.2 -> 5.8/13.0Improved
Add Transformer LM rescoring to STC decoding (ablation)STC · LibriSpeech p_drop=0.1test WER clean/other: 4.1/9.0 -> 3.3/6.9Improved
Add 5-gram LM beam search to STC decoding (ablation)STC · LibriSpeech p_drop=0.7test WER clean/other: 6.9/15.1 -> 7.0/14.8Mixed
Add pseudo labeling then CTC letter modelSTC · LibriSpeech p_drop=0.1, Transformer rescoringtest WER clean/other: 3.3/6.9 -> 2.7/5.2Improved
Add pseudo labeling then CTC letter modelSTC · LibriSpeech p_drop=0.7, Transformer rescoringtest WER clean/other: 5.1/11.1 -> 3.1/6.2Improved
Increase p_drop from 0.1 to 0.7STC · LibriSpeech greedytest WER clean/other: 4.1/9.0 -> 6.9/15.1Hurt
Use STC instead of CTCCNN · IAM p_drop=0.1test CER: 7.7 -> 7.2Improved
Use STC instead of CTCCNN · IAM p_drop=0.7test CER: 78.5 -> 26.7Improved

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