LargeScaleASR
收藏资源简介:
LargeScaleASR数据集是一个包含25,000小时转录和异构英语语音识别数据的数据集,适用于研究和商业用途。数据集由6个子集组成,分别是large、medium、small、clean、dev和test,每个子集包含不同小时数的转录语音数据。数据集主要用于自动语音识别任务,特别是鲁棒语音识别和噪声语音识别。数据集的创建涉及多个现有数据集的混合,包括VoxPopuli、LibriHeavy、Librispeech、YODAS、People's Speech和CommonVoice 18.0。数据集中的文本和音频都经过了标准化处理,以确保数据的一致性和质量。
The LargeScaleASR dataset is a collection of 25,000 hours of transcribed and heterogeneous English speech recognition data, suitable for both research and commercial use cases. The dataset comprises six subsets, namely large, medium, small, clean, dev, and test, with each subset containing transcribed speech data of varying durations. This dataset is primarily utilized for automatic speech recognition (ASR) tasks, especially robust speech recognition and noisy speech recognition. The creation of the dataset involves mixing multiple existing datasets, including VoxPopuli, LibriHeavy, LibriSpeech, YODAS, People's Speech, and CommonVoice 18.0. Both the textual and audio data in the dataset have been standardized to ensure data consistency and quality.




