边缘设备上的质量感知自校正语音翻译
Quality-Aware Self-Correcting Speech Translation on an Edge Device
- University of Surrey（萨里大学）
- Institute for People-Centred AI（以人为本人工智能研究所）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
提出一种在 Jetson Nano 上运行的离线语音到语音翻译流水线，利用质量估计门控触发二次校正，发现 QE 作为门控有效但作为排序器不佳，移除它可释放内存且不损质量，并发布系统支持六语言对实时翻译。
AI中文摘要：
我们提出了一种完全离线的语音到语音翻译流水线，该流水线在 Jetson Nano（4 GB）上运行，无需重新训练即可纠正自身较弱的翻译。一个 Whisper-tiny 自动语音识别（ASR）模块将语音转换为文本，并馈送给 Opus-MT 翻译器；多语言 BERT 余弦相似度作为质量估计（QE）门控，当置信度低于预定义阈值 $\ au$ 时，触发二次通道校正。我们比较了三种校正方法：QE 重排序（M1）、最小贝叶斯风险解码（M2）和约束波束搜索（M3）。在 1,012 个 FLORES-200 句子（英语-西班牙语）上，M2 在 $\ au=0.90$ 时，与贪婪解码相比，在 BLEU（+0.67，p<0.001）、ChrF（+0.51，p<0.001）和 COMET（+0.0020，N=3，p=0.002）上产生了统计显著的改进；M1 未产生显著增益，而 M3 显著差于基线（p>0.99）。我们的核心发现是，QE 作为门控有效，但作为排序器效果不佳：从候选选择中移除 QE 模型（M1$\ o$M2）不会损害质量，并从关键路径中释放了 680 MB 内存。利用改编自后期编辑工作量文献的增益与编辑比率，我们进一步表明，较小的候选池（N=3）能产生更精准的校正，具有更好的语义充分性，而较大的候选池（N=10）能最大化词汇奖励。我们发布了该系统，并展示了跨六个语言对的实时翻译。
英文摘要：
We present a fully offline speech-to-speech translation pipeline that runs on a Jetson Nano (4 GB) and corrects its own weak translations without retraining. A Whisper-tiny ASR feeds an Opus-MT translator; multilingual BERT cosine similarity acts as a Quality Estimation (QE) gate, triggering a secondary-pass correction when confidence falls below a pre-defined threshold $τ$. We compare three correction methods: QE reranking (M1), Minimum Bayes-Risk decoding (M2), and constrained beam search (M3). On 1,012 FLORES-200 sentences (English-Spanish), M2 at $τ=0.90$ produces statistically significant improvements over greedy decoding on BLEU (+0.67, p<0.001), ChrF (+0.51, p<0.001), and COMET (+0.0020 at N=3, p=0.002); M1 yields no significant gains, and M3 is significantly worse than baseline (p>0.99). Our central finding is that QE functions effectively as a gate but poorly as a ranker: removing the QE model from candidate selection (M1$\to$M2) does not hurt quality and frees 680 MB from the critical path. Using a gain-to-edit ratio adapted from the post-editing-effort literature, we further show that smaller candidate pools (N=3) yield more surgical corrections with better semantic adequacy, while larger pools (N=10) maximise lexical reward. We release the system and demonstrate live translation across six language pairs.
