arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2210.06324cs.CLcs.LGcs.SDeess.AS
收藏

SQuId:衡量多种语言的语音自然度

SQuId: Measuring Speech Naturalness in Many Languages

Google(谷歌)

Thibault Sellam, Ankur Bapna, Joshua Camp, Diana Mackinnon, Ankur P. Parikh, Jason Riesa

AI总结:

本文提出多语言语音自然度预测模型 SQuId,利用百万级评分覆盖 65 个语言地区,证明多语言联合训练与跨地区迁移显著优于单语基线。

摘要:

文本转语音研究在很大程度上依赖人工评估,这会产生高昂成本并拖慢开发过程。在高度多语言的应用中,该问题尤其突出,因为招募和调查评判者可能需要数周时间。我们提出 SQuId(Speech Quality Identification,语音质量识别),这是一个多语言自然度预测模型,基于超过一百万条评分进行训练,并在 65 个语言地区进行测试,是迄今为止同类工作中规模最大的一次。核心洞见是:在许多语言地区上训练一个模型,会持续优于单语言地区基线。我们介绍了任务和模型,并表明其性能比基于 w2v-BERT 和 VoiceMOS 的强基线高出 50.0%。随后,我们证明了微调过程中跨语言地区迁移的有效性,并强调其对零样本语言地区(即没有微调数据的语言地区)的作用。通过一系列分析,我们强调了声音伪影等非语言因素在跨语言地区迁移中的作用。最后,我们通过若干消融实验展示了模型规模、预训练多样性和语言再平衡等设计决策的影响。

英文摘要:

Much of text-to-speech research relies on human evaluation, which incurs heavy costs and slows down the development process. The problem is particularly acute in heavily multilingual applications, where recruiting and polling judges can take weeks. We introduce SQuId (Speech Quality Identification), a multilingual naturalness prediction model trained on over a million ratings and tested in 65 locales-the largest effort of this type to date. The main insight is that training one model on many locales consistently outperforms mono-locale baselines. We present our task, the model, and show that it outperforms a competitive baseline based on w2v-BERT and VoiceMOS by 50.0%. We then demonstrate the effectiveness of cross-locale transfer during fine-tuning and highlight its effect on zero-shot locales, i.e., locales for which there is no fine-tuning data. Through a series of analyses, we highlight the role of non-linguistic effects such as sound artifacts in cross-locale transfer. Finally, we present the effect of our design decision, e.g., model size, pre-training diversity, and language rebalancing with several ablation experiments.

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。

↑