Project Euphonia
收藏资源简介:
Project Euphonia是由Google Research发起的一个大型数据集,旨在改善对失语症患者语音的自动识别。该数据集包含了来自约2000名不同病因的失语症患者的超过120万条语音记录。数据集的创建过程包括了多样化的语音样本收集、人工校对转录和音频质量标签的添加,以及超过40种语音特征标签的标注。该数据集的应用领域主要集中在提高失语症患者语音识别技术的准确性和包容性,旨在解决现有语音识别系统对失语症患者支持不足的问题。
Project Euphonia is a large-scale dataset initiated by Google Research, aimed at advancing automatic speech recognition (ASR) for individuals with aphasia. This dataset contains over 1.2 million speech recordings from approximately 2,000 people with aphasia across various etiologies. The development process of the dataset includes diversified speech sample collection, manual transcription and proofreading, addition of audio quality tags, and annotation of more than 40 speech feature tags. Its core applications focus on enhancing the accuracy and inclusivity of speech recognition technologies for aphasia patients, aiming to resolve the issue that existing speech recognition systems provide insufficient support for this group.

- 1Learnings from curating a trustworthy, well-annotated, and useful dataset of disordered English speechGoogle Research · 2024年



