对比正则化用于口音鲁棒的ASR
Contrastive Regularization for Accent-Robust ASR
AI总结：
提出使用监督对比学习作为轻量级口音不变辅助目标，在CTC微调中正则化编码器表示，无需架构修改或显式口音监督，在L2-ARCTIC基准上实现高达25-29%的未见口音词错误率降低。
摘要：
基于自监督声学预训练和CTC微调的ASR系统在母语语音上表现强劲，但对口音变化仍然敏感。我们研究监督对比学习（SupCon）作为CTC微调的轻量级、口音不变辅助目标。一个话语级对比损失正则化编码器表示，无需架构修改或显式口音监督。在L2-ARCTIC基准上的实验表明，多个预训练编码器均实现一致的WER降低，在未见口音评估下相对降低高达25-29%。使用转录内余弦离散度分析表明，SupCon在口音变化下促进更紧凑和稳定的表示几何结构。总体而言，SupCon提供了一种有效且模型无关的正则化策略，用于提高口音鲁棒性。
英文摘要：
ASR systems based on self-supervised acoustic pretraining and CTC fine-tuning achieve strong performance on native speech but remain sensitive to accent variability. We investigate supervised contrastive learning (SupCon) as a lightweight, accent-invariant auxiliary objective for CTC fine-tuning. An utterance-level contrastive loss regularizes encoder representations without architectural modification or explicit accent supervision. Experiments on the L2-ARCTIC benchmark show consistent WER reductions across multiple pretrained encoders, with up to 25 -- 29\% relative reduction under unseen-accent evaluation. Analysis using within-transcript cosine dispersion indicates that SupCon promotes more compact and stable representation geometry under accent variability. Overall, SupCon provides an effective and model-agnostic regularization strategy for improving accent robustness.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
