Metraipolis › archive › dc_c5d5ca0bf5cd
Learning New Words from Unlabeled Test Data in Automatic Speech Recognition — arXiv:2609.28877
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.28877
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T04:50:11.104Z
- SHA-256 of the archived text
14cd19064a7ea35abaf35f07ce5017dbb631bd93001635996f1e5e3d7b1505cb- Stable URL of this page
https://metraipolis.com/documento/dc_c5d5ca0bf5cd
How to cite
Learning New Words from Unlabeled Test Data in Automatic Speech Recognition — arXiv:2609.28877 — arXiv (en), https://arxiv.org/abs/2609.28877 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c5d5ca0bf5cd (Metraipolis archive dc_c5d5ca0bf5cd, sha256 14cd19064a7ea35abaf35f07ce5017dbb631bd93001635996f1e5e3d7b1505cb)
BibTeX
@misc{dcc5d5ca0bf5cd,
title = {Learning New Words from Unlabeled Test Data in Automatic Speech Recognition — arXiv:2609.28877},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.28877},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c5d5ca0bf5cd; sha256 14cd19064a7ea35abaf35f07ce5017dbb631bd93001635996f1e5e3d7b1505cb},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.28877. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
New words are invented every day. A human listener can learn a new word by hearing it clearly once and inferring its usage from sentence context. This paper proposes granting ASR a similar ability to learn the contextual representations and spellings of new words from unlabeled test data at test time. A frozen CTC acoustic model provides spellings, a frozen language model provides contextual evidence for out-of-vocabulary (OOV) word detection, and an adaptation module expands the vocabulary by learning the lexical token representations with distributions over CTC-generated candidates. The spelling model of each token is optimized by minimizing a Kullback-Leibler divergence (KLD) objective. We demonstrate that the CTC-weighted language model log likelihood ratio can be interpreted as the KLD between the unknown correct ASR and the unsupervised learned ASR, and that, using a Pinsker bound, the square root of KLD can be interpreted as an upper bound on the total variation distance between the true and estimated spelling of the unknown word. Experiments show relative OOV character-error-rate reductions of up to 14.97% on LibriSpeech and 6.67% on dysarthric Speech Accessibility Project data for recurring OOV words, relative to the corresponding rescoring system. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-25 in cs.CL). Categorie arXiv: Audio and Speech Processing (eess.AS) ; Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2609.28877 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-08T04:27:20.154Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-25?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).