analiza aiqbench
This research introduces a novel boundary-free decoding method for speech recognition in agglutinative languages like Japanese and Chinese, overcoming limitations of traditional word-segmentation approaches. By utilizing character-level automata and depth-adaptive gating, the technique significantly boosts rare-word recall on frozen models without requiring retraining or additional inference passes. The release of an open benchmark further establishes a new standard for evaluating contextual biasing in low-resource scenarios.
Boundary-Free Contextual Biasing: Depth-Adaptive Gating and Reading-Space Matching for Unsegmented Languages
arXiv:2610.09467v1 Announce Type: cross Abstract: Contextual biasing supplies an ASR system with a list of expected words at inference time, but existing methods rely on word boundaries that Japanese and Chinese do not provide. We present a boundary-free biasing decoder for frozen public CTC models, built on a character-level Aho-Corasick automaton, with no training and no second pass. Two evidence-based mechanisms replace the boundary: a depth-adaptive gate that sets how hard to push from match depth, and reading-space matching for when the audio is right but the characters are wrong. On Aishell-1 NE's hard R1 subset we reach 66.5% recall, above the trained CLAS baseline (64%), transferring to WenetSpeech and to a second architecture without retuning. We release the first open Japanese contextual-biasing benchmark, where biasing lifts rare-word recall by 25 points at precision above 97%, and still by 19 and 22 points against 1,000-word lists.
Abstract:Contextual biasing supplies an ASR system with a list of expected words at inference time, but existing methods rely on word boundaries that Japanese and Chinese do not provide. We present a boundary-free biasing decoder for frozen public CTC models, built on a character-level Aho-Corasick automaton, with no training and no second pass. Two evidence-based mechanisms replace the boundary: a depth-adaptive gate that sets how hard to push from match depth, and reading-space matching for when the audio is right but the characters are wrong. On Aishell-1 NE's hard R1 subset we reach 66.5% recall, above the trained CLAS baseline (64%), transferring to WenetSpeech and to a second architecture without retuning. We release the first open Japanese contextual-biasing benchmark, where biasing lifts rare-word recall by 25 points at precision above 97%, and still by 19 and 22 points against 1,000-word lists.
Audio and Speech Processing (eess.AS); Computation and Language (cs.CL); Sound (cs.SD)
arXiv:2610.09467 [eess.AS]
(or arXiv:2610.09467v1 [eess.AS] for this version)
https://doi.org/10.48550/arXiv.2610.09467 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 243 cuvinte
Subiecte conexe
Challenge: Hand coding weights for efficient sequence memorisation
AI Alignment Forum · 23 iul. 2026
Article: Governing AI in the Cloud: A Practical Guide for Architects
InfoQ · 15 iun. 2026
10 Ways AI Helps You Build What Your Business Needs
AppDirect · 15 mai 2026
[AINews] not much happened today
Latent Space · 10 sept. 2026