数据堂—500小时法语自然对话手机采集语音数据

500小时法语自然对话手机采集语音数据约有700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，领域广泛，语音自然流利，符合实际对话场景。人工转写文本，准确率高。

This 500-hour French natural conversational speech dataset was collected via mobile phones, with approximately 700 speakers participating in the recording. The participants communicated naturally and spoke freely on a variety of given topics spanning extensive domains. The collected speech is natural and fluent, aligning with real-world conversational scenarios. The accompanying manual transcriptions feature high accuracy.

搜集汇总

背景与挑战

背景概述

该数据集包含500小时的法语自然对话语音，由约700名说话者通过手机采集，用于测试法语自然对话语音识别模型。录音格式为16kHz、16位、单声道的WAV，转录准确率不低于95%，内容涵盖多个话题的流畅对话。