发表机构
University of New South Wales; University of Melbourne; University of Sydney（新南威尔士大学; 墨尔本大学; 悉尼大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
研究极低资源环境下语言相似度对ASR跨语言迁移的作用，提出结合声学与语言相似度的框架对高资源源语言排名，实验表明相似语言优于基线，阿萨姆语和印地语效果好，还分析了不同相似度对迁移的影响。
AI 中文摘要
本文研究在极低资源环境下，语言相似度如何改善自动语音识别（ASR）的跨语言迁移。澳大利亚原住民语言瓦尔皮里语的转录语音数据非常有限，因此迁移学习至关重要。我们提出一个框架，将预训练语音模型的声学相似度与基于类型学、音素表、语法和句法特征的语言相似度相结合，对高资源源语言进行排名，并评估它们对瓦尔皮里语ASR迁移的有效性。使用Whisper的实验表明，声学和类型学上相似的语言优于单语和多语基线。阿萨姆语和印地语在单词和字符错误率上有显著降低。相关性分析进一步表明，声学相似度是微调性能的最强预测指标，而音素表和类型学相似度更能解释零样本迁移。
英文摘要
This paper investigates how language similarity can improve cross-lingual transfer for automatic speech recognition (ASR) in extremely low-resource settings. Warlpiri, an Australian Aboriginal language, has very limited transcribed speech data, making transfer learning essential. We propose a framework combining acoustic similarity from pre-trained speech models with linguistic similarity based on typology, phoneme inventories, grammatical, and syntactic features to rank high-resource source languages and evaluate their effectiveness for ASR transfer to Warlpiri. Experiments with Whisper show that acoustically and typologically similar languages outperform monolingual and multilingual baselines. Assamese and Hindi achieve substantial reductions in word and character error rates. Correlation analysis further indicates that acoustic similarity is the strongest predictor of fine-tuning performance, while phoneme inventory and typological similarity better explain zero-shot transfer.
CommentsAccepted by Interspeech 2026
