microsoft/NOTSOFAR
收藏Hugging Face2025-01-23 更新2025-04-08 收录
下载链接：
https://hf-mirror.com/datasets/microsoft/NOTSOFAR
下载链接
链接失效反馈官方服务：
资源简介：
NOTSOFAR-1挑战的数据集包含两部分：录制的会议数据集和模拟的训练数据集。录制的会议数据集由315个平均6分钟长的会议组成，涵盖30个会议室、4-8名与会者和35名独特的讲话者，反映了各种真实世界的声学条件和对话动态。模拟的训练数据集是一个1000小时的模拟数据集，包含15000个真实的声学传递函数，用于增强模型在真实世界中的泛化能力。
The NOTSOFAR-1 Challenge dataset consists of two parts: the recorded meeting dataset and the simulated training dataset. The recorded meeting dataset includes 315 meetings averaging 6 minutes each, recorded across 30 conference rooms with 4-8 attendees and featuring a total of 35 unique speakers, reflecting a wide range of real-world acoustic conditions and conversational dynamics. The simulated training dataset is a 1000-hour simulated dataset synthesized with enhanced authenticity for real-world generalization, incorporating 15,000 real acoustic transfer functions.
提供机构：
microsoft
