[论文] TurboBias 2.0: Streaming Context-Biasing for Production-Efficient ASR ...

论文概要

研究领域: NLP
作者: Vladimir Bataev, Lilit Grigoryan, Andrei Andrusenko, Nikolay Karpov, Vitaly Lavrukhin, Boris Ginsburg
发布时间: 2026-08-21
arXiv: 2608.21343

中文摘要

语境化对于生产级自动语音识别（ASR）系统至关重要，其中用户提供的短语必须在严格延迟约束下被准确识别。虽然许多语境偏置方法改进了识别准确性，但它们往往未解决现代生产ASR系统的实际需求：流式推理、高效批处理解码、用户特定语境列表和低运行时开销。我们提出了TurboBias 2.0，一个面向生产的Transducer基ASR系统的高效短语增强框架。该框架通过不区分大小写的增强图和每流批处理解码扩展了GPU加速的TurboBias，允许批次中的每个话语使用独立的语境偏置配置。这实现了多同时用户的个性化语境偏置，无需共享或混合他们的语境列表。所提出的框架支持离线推理和流式推理，可与贪婪搜索和束搜索解码一起使用。实验表明，TurboBias 2.0在保持低延迟和高吞吐量的同时改进了语境短语识别。

原文摘要

Contextualization is essential for production automatic speech recognition (ASR) systems, where user-provided phrases must be recognized accurately under strict latency constraints. Although many context-biasing methods improve recognition accuracy, they often do not address the practical requirements of modern production ASR systems: streaming inference, efficient batched decoding, user-specific context lists, and low runtime overhead. We propose TurboBias 2.0, a production-oriented framework for efficient phrase boosting in Transducer-based ASR systems. The framework extends GPU-accelerated TurboBias with a case-insensitive boosting graph and per-stream batched decoding, allowing each utterance in a batch to use an independent context-biasing configuration. This enables personalized cont...

自动采集于 2026-08-25

#论文 #arXiv #NLP #小凯