arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2402.02302eess.AScs.CL
收藏

利用声学伪标记预测正迁移以改进低资源语音识别

Predicting positive transfer for improved low-resource speech recognition using acoustic pseudo-tokens

Stanford University(斯坦福大学);Athena Research Center(雅典娜研究中心);Wayne State University(韦恩州立大学);Atos zData

Nay San, Georgios Paraskevopoulos, Aryaman Arora, Xiluo He, Prabhjot Kaur, Oliver Adams, Dan Jurafsky

AI总结:

针对低资源语音识别,提出声学标记分布相似度(ATDS)度量,用于选择相似的高资源供体语言进行持续预训练,从而有效提升目标语言ASR性能。

摘要:

虽然像 wav2vec 2.0 XLSR-128 这样的大规模多语言语音模型可以直接针对自动语音识别(ASR)进行微调,但对于在预训练数据中代表性不足的语言,其下游性能仍然可能相对较差。对这些语言中 70-200 小时的未转录语音进行持续预训练会有所帮助——但对于没有那么多录音数据的语言呢?对于此类情况,我们表明,用来自相似的高资源“供体”语言的数据补充目标语言会有所帮助。例如,仅用 10 小时的低资源旁遮普语加上 60 小时的供体印地语进行持续预训练,几乎与用 70 小时旁遮普语进行持续预训练效果相当。相比之下,从相似度较低的供体(如孟加拉语)获取数据并不能改善 ASR 性能。为了指导供体语言的选择,我们提出了一种基于诱导声学单元序列分布的新颖相似度度量:声学标记分布相似度(ATDS)。在一组类型学上不同的目标语言(旁遮普语、加利西亚语、伊班语、茨瓦纳语)中,我们表明目标语言与其候选供体之间的 ATDS 能精确预测目标语言的 ASR 性能。

英文摘要:

While massively multilingual speech models like wav2vec 2.0 XLSR-128 can be directly fine-tuned for automatic speech recognition (ASR), downstream performance can still be relatively poor on languages that are under-represented in the pre-training data. Continued pre-training on 70-200 hours of untranscribed speech in these languages can help -- but what about languages without that much recorded data? For such cases, we show that supplementing the target language with data from a similar, higher-resource 'donor' language can help. For example, continued pre-training on only 10 hours of low-resource Punjabi supplemented with 60 hours of donor Hindi is almost as good as continued pretraining on 70 hours of Punjabi. By contrast, sourcing data from less similar donors like Bengali does not improve ASR performance. To inform donor language selection, we propose a novel similarity metric based on the sequence distribution of induced acoustic units: the Acoustic Token Distribution Similarity (ATDS). Across a set of typologically different target languages (Punjabi, Galician, Iban, Setswana), we show that the ATDS between the target language and its candidate donors precisely predicts target language ASR performance.

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。

↑