返回全部动态
Whispering Wisely：面向ASR微调的智能子集选择
原标题：Whispering Wisely: Smart Subset Selection for ASR Fine-Tuning
AI 摘要
该论文提出一种基于嵌入的子集选择方法，用于降低Whisper等ASR模型的微调成本。研究者利用Whisper预训练编码器将音频样本映射到固定维度嵌入空间，并用贪心k-center算法挑选多样且具代表性的训练子集。在希伯来语和丹麦语上的实验表明，k-center选择在较小样本预算下常优于随机采样，且子集词汇多样性更高，为低资源语言ASR微调提供了实用策略。
以上摘要由 AI 生成，可能存在误差。事实请以原文为准。
正文节选
Whispering Wisely: Smart Subset Selection for ASR Fine-Tuning Abstract Automatic Speech Recognition (ASR) models such as OpenAI’s Whisper have substantially improved speech recognition quality but remain expensive to fine-tune, especially when large amounts of transcribed audio are available. This makes it challenging for many practitioners to adapt these models to new languages or domains. In this work, we introduce an embedding-based subset selection approach for selecting smaller training set
发布时间：2026-10-27 12:00
抓取时间：2026-10-27 12:00
来源机构：arXiv