数据堂—357小时韩语手机采集语音数据

357小时韩语手机采集语音数据由999名韩国人参与录制；录音环境安静，录音内容丰富；所有文本均由专业标注人员人工转写，句准确率95%；357小时韩语手机采集语音数据可应用于语音识别、机器翻译、声纹识别

A 357-hour Korean speech dataset collected via mobile phones, recorded by 999 Korean participants. The recordings were conducted in quiet environments with diverse and rich content. All transcriptions were manually completed by professional annotators, with a sentence-level accuracy of 95%. This dataset can be applied to speech recognition, machine translation and speaker recognition.

搜集汇总

背景与挑战

背景概述

该数据集包含357小时的韩语手机采集语音数据，由999名韩国参与者在安静室内录制，内容为常用句子，经专业标注人员转录后句准确率达95%。数据格式为16kHz单声道WAV，适用于语音识别、机器翻译及声纹识别等任务，版权归数据堂所有，支持商业使用。