michsethowusu/ghana-speech

这是一个多语言音频数据集，包含六种低资源语言：Akuapem_Twi、Anyin、Asante_Twi、Avatime、Kasem和Konkomba，这些语言主要来自加纳等地区。数据集提供了音频文件及其对应的文本转录，音频采样率为16000Hz。每个样本包含id、语言代码、文本内容、持续时间、源文件路径和音频数据。训练分割总计有约484.82小时的音频时长，覆盖从9,956到143,383个不等的段数。该数据集旨在支持语音识别、语言建模和其他NLP研究，特别关注资源稀缺语言的开发。This is a multilingual audio dataset comprising six low-resource languages: Akuapem_Twi, Anyin, Asante_Twi, Avatime, Kasem, and Konkomba, primarily from regions such as Ghana. The dataset includes audio files paired with corresponding text transcriptions, with an audio sampling rate of 16000 Hz. Each sample features fields such as id, language code, text content, duration, source file path, and audio data. The training split totals approximately 484.82 hours of audio duration, with segment counts ranging from 9,956 to 143,383. This dataset is designed to support speech recognition, language modeling, and other NLP research, with a focus on under-resourced languages.