数据堂—750小时印地语自然对话手机采集语音数据

750小时印地语自然对话手机采集语音数据约有1300名发音人参与录制，以自然方式进行交流，针对给定的数个话题自由发挥，涉及领域广泛，发音人语音自然流利，符合实际对话场景。750小时印地语自然对话手机采集语音数据由人工转写文本，准确率高。

This 750-hour Hindi natural conversational speech dataset was collected via mobile devices. Approximately 1,300 speakers participated in the recording sessions, where they communicated naturally and spoke freely on several assigned topics spanning diverse domains. The speakers' speech was natural and fluent, aligning with real-world conversational scenarios. All text transcripts of the dataset are manually produced with high accuracy.

搜集汇总

背景与挑战

背景概述

该数据集由数据堂提供，包含750小时通过手机采集的印地语自然对话语音，用于相关模型的测试任务。数据涉及约1300名说话者，覆盖广泛话题，音频格式为16kHz单声道WAV，并附有高准确度的人工转录文本。