为非洲语言打造的自监督语音模型，支持1226种语言，性能超越现有方案。
AfriHuBERT: A self-supervised speech representation model for African languages
- 在1万小时语音数据上持续预训练，扩展至1226种非洲语言
- 语音识别任务中词错误率降低2.1%，语言识别F1提升3.6%
- 特别适合资源极低的非洲语言语音应用，泛化能力更强
本文提出AfriHuBERT，基于mHuBERT-147（原覆盖16种非洲语言）进行持续预训练，利用超过10,000小时多源语音数据，将语言覆盖扩展至1,226种，惠及超过6亿非洲用户。在FLEURS基准上评估其在语音语言识别（SLID）与自动语音识别（ASR）两个关键任务上的表现：相较于mHuBERT-147，SLID的F1分数提升3.6%，ASR平均词错误率（WER）降低2.1%，性能可与更大规模的MMS和XEUS等自监督模型媲美。进一步分析表明，基于AfriHuBERT训练的ASR模型具备更强的跨语料泛化能力，在极端低资源场景下仍具竞争力。
原文摘要 · Abstract (English)
In this work, we present AfriHuBERT, an extension of mHuBERT-147, a compact self-supervised learning (SSL) model pretrained on 147 languages. While mHuBERT-147 covered 16 African languages, we expand this to 1,226 through continued pretraining on 10K+ hours of speech data from diverse sources, benefiting an African population of over 600M. We evaluate AfriHuBERT on two key speech tasks, Spoken Language Identification (SLID) and Automatic Speech Recognition (ASR), using the FLEURS benchmark. Our results show a +3.6% F1 score improvement for SLID and a -2.1% average Word Error Rate (WER) reduction for ASR over mHuBERT-147, and demonstrates competitiveness with larger SSL models such as MMS and XEUS. Further analysis shows that ASR models trained on AfriHuBERT exhibit improved cross-corpus generalization and are competitive in extremely low-resource ASR scenarios.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。