skesiraju/Shrutilipi

Shrutilipi数据集是一个包含8种印度语言的广播音频转录的全版本数据集，这些语言包括孟加拉语、印地语、卡纳达语、马拉雅拉姆语、马哈拉施特拉语、奥里亚语、泰米尔语和泰卢固语。该数据集由AI4Bharat团队自动整理而成，用于为资源匮乏的语言提供自动语音识别（ASR）的训练数据。数据集包含自动生成的转录文本和对应的质量评分，用户可以根据质量评分来筛选数据。

The Shrutilipi dataset is a full version of broadcast audio transcriptions in eight Indian languages, including Bengali, Hindi, Kannada, Malayalam, Marathi, Odia, Tamil, and Telugu. Curated automatically by the AI4Bharat team, this dataset is intended to provide training data for automatic speech recognition (ASR) for low-resource languages. It includes automatically generated transcription texts along with quality scores, allowing users to filter the data based on the quality requirements.