GLOBE
收藏资源简介:
GLOBE数据集是由新南威尔士大学计算机科学与工程学院创建的高质量英语语料库,包含来自23,519名全球不同口音的说话者,总计535小时的语音数据。该数据集通过严格的过滤和增强过程,从原始的Common Voice数据集中提炼而来,旨在解决当前零样本说话者自适应文本到语音(TTS)系统在适应口音方面的局限性。GLOBE数据集不仅提供了详细的说话者元数据,包括口音、年龄和性别,还广泛覆盖了164种全球口音,显著提高了零样本TTS模型对不同口音的泛化能力。该数据集的应用领域主要集中在提升TTS系统的口音适应性和语音合成的自然度,解决现有TTS系统在处理多样口音时的性能问题。
The GLOBE dataset is a high-quality English corpus created by the School of Computer Science and Engineering at the University of New South Wales. It contains 535 hours of speech data from 23,519 speakers with diverse global accents. Derived from the original Common Voice dataset through rigorous filtering and augmentation processes, this dataset aims to address the limitations of current zero-shot speaker adaptive Text-to-Speech (TTS) systems in accent adaptation. The GLOBE dataset not only provides detailed speaker metadata including accent, age and gender, but also covers 164 distinct global accents, significantly enhancing the generalization capability of zero-shot TTS models across various accents. Its main application areas focus on improving the accent adaptability of TTS systems and the naturalness of speech synthesis, as well as resolving the performance issues of existing TTS systems when dealing with diverse accents.

- 1GLOBE: A High-quality English Corpus with Global Accents for Zero-shot Speaker Adaptive Text-to-Speech新南威尔士大学计算机科学与工程学院 · 2024年



