A popular approach to streaming speech translation is to employ a single offline model with a \textit{wait-$k$} policy to support different latency requirements, which is simpler than training multiple online models with different latency constraints. However, there is a mismatch problem in using a model trained with complete utterances for streaming inference with partial input. We demonstrate that speech representations extracted at the end of a streaming input are significantly different from those extracted from a complete utterance. To address this issue, we propose a new approach called Future-Aware Streaming Translation (FAST) that adapts an offline ST model for streaming input. FAST includes a Future-Aware Inference (FAI) strategy that incorporates future context through a trainable masked embedding, and a Future-Aware Distillation (FAD) framework that transfers future context from an approximation of full speech to streaming input. Our experiments on the MuST-C EnDe, EnEs, and EnFr benchmarks show that FAST achieves better trade-offs between translation quality and latency than strong baselines. Extensive analyses suggest that our methods effectively alleviate the aforementioned mismatch problem between offline training and online inference.

翻译：流式语音翻译的常用方法是采用单一离线模型配合\textit{等待-$k$}策略以支持不同延迟需求，这比训练多个具有不同延迟约束的在线模型更为简洁。然而，使用基于完整语音训练序的模型进行部分输入的流式推理存在失配问题。我们证明流式输入结束时提取的语音表征与完整语音提取的表征存在显著差异。为解决此问题，我们提出一种名为未来感知流式翻译（FAST）的新方法，该方法将离线语音翻译模型适配至流式输入场景。FAST包含两项关键技术：通过可训练掩码嵌入融合未来上下文的未来感知推理（FAI）策略，以及将近似全语音的未来上下文迁移至流式输入的蒸馏框架（FAD）。在MuST-C EnDe、EnEs和EnFr基准测试上的实验表明，FAST在翻译质量与延迟之间取得了优于强基线方法的权衡。大量分析证明，我们的方法有效缓解了离线训练与在线推理之间的上述失配问题。