用缓存语言模型概率提升语音识别，无需训练
Cached LLM Probability Retrieval for Speech Recognition
- 离线用大模型预计算上下文-词的概率，识别时直接查表
- 在39个测试场景中，28个优于单次推理结果
- 适合想快速改进现有语音识别系统的人
大型语言模型（LLM）通过提供语言先验增强自动语音识别（ASR），但直接重评分成本高昂，需对每个N-best候选进行评估。本文提出“缓存的LLM概率检索”：离线用本地教师模型为相关上下文-目标词对查询下一词概率，并在识别时通过缓存查找、回退策略和可选打分处理显著错误。该方法无需训练，可无缝集成到现有识别器中，不修改声学模型。多模型评估显示，在39个设置中有28个优于1-pass ASR，且非最优错误更低。上下文长度分析表明，收益在上下文长度为8时达到峰值，证明该方法是有效且轻量的ASR适配方案，相比生成式纠错（GER）或知识蒸馏（KD）所需的繁重训练更具优势。
原文摘要 · Abstract (English)
Large language models (LLMs) enhance automatic speech recognition (ASR) by providing linguistic priors; however, their direct rescoring is costly because it requires evaluating every N-best hypothesis. This paper introduces "cached LLM probability retrieval," which involves querying a local teacher LLM offline to obtain next-token probabilities for ASR-relevant context-target pairs. These probabilities are then utilized during recognition via cache lookups, backoff strategies, and optional scoring for significant misses. The method is training-free and can integrate with existing recognizers without requiring modifications to acoustic models. Evaluations across various ASR models reveal that cached retrieval outperforms 1-pass ASR in 28 of 39 settings and achieves lower non-oracle errors. Context length analysis indicates that benefits peak at a context length of 8, suggesting that cached probability retrieval is an effective and lightweight ASR adaptation method, in contrast to the heavy training required for Generative Error Correction (GER) or knowledge distillation (KD).
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。