arXiv cs.AI· Xiaoyu Gu, Yu Xi, Chencheng Wang, Zonghan Zhou, Haoyu Li, Jing Peng, Kai Yu·· 12 小时前AI 评分24
VISH-ASR：Viterbi 引导的热词选择，用于语音 LLM 上下文 ASR
VISH-ASR: Viterbi-Guided Selection of Hotwords for Speech-LLM Contextual ASR
AI 导读
VISH 是一个免训练、即插即用的热词预选框架，直接在帧级 CTC 后验上验证候选词的声学支持，将预选建模为受约束的后验搜索。M-VISH 通过帧同步共享前缀 Viterbi 搜索与有界 top-K 剪枝实现大规模词表验证，Bi-MVISH 则利用反向证据增强鲁棒性。
正文 · 原文
Abstract:Prompt-based contextual ASR requires selecting a compact hotword subset from large dynamic inventories, yet existing pre-selection methods rely either on discrete ASR hypotheses that may discard acoustic evidence or on separately trained retrieval models. We propose Viterbi-guided Selection of Hotwords (VISH), a training-free, plug-and-play framework that directly verifies candidate-specific acoustic support over frame-level CTC posteriors. VISH formulates pre-selection as constrained posterior search; M-VISH makes large-list verification scalable through frame-synchronous shared-prefix Viterbi search with bounded top-K pruning, while Bi-MVISH exploits complementary backward evidence for robust verification. Across five benchmarks, M-VISH achieves up to 11.7% absolute improvement in Best-F1 over a strong learned-retrieval baseline and reduces candidate-search latency by 75.2% at 2,000 candidates relative to hypothesis-based filtering. These results demonstrate the effectiveness of posterior-level verification for accurate and scalable hotword pre-selection without additional model training.
| Subjects: | Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC) |
| Cite as: | arXiv:2610.27867 [eess.AS] |
| (or arXiv:2610.27867v1 [eess.AS] for this version) | |
| https://doi.org/10.48550/arXiv.2610.27867 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Xiaoyu Gu [view email]
[v1]
Tue, 20 Oct 2026 08:49:56 UTC (1,118 KB)
来源：arXiv cs.AI · arxiv.org