数据堂— 197小时韩语手机采集语音数据

197小时韩语手机采集语音数据包括291名来自韩国、朝鲜的本土发音人，在安静无回音的环境的录音数据。录音内容广泛，包括经济，娱乐，新闻，口语，数字，字母等，每人约400句。197小时韩语手机采集语音数据的文本转写句错误率小于5%。匹配主流安卓、苹果系统手机

This 197-hour Korean speech dataset is collected via mobile phones, with 291 native speakers from South Korea and North Korea as participants. All recordings are made in quiet, echo-free environments. The recorded content covers a wide range of topics including economy, entertainment, news, colloquial speech, numbers, letters, etc., with each speaker providing approximately 400 sentences. The sentence-level transcription error rate of this dataset is less than 5%, and it is compatible with mainstream Android and Apple iOS mobile phones.

搜集汇总

背景与挑战

背景概述

该数据集是一个用于韩语语音识别模型测试的语音资源，包含197小时通过手机采集的韩语录音。数据由291名韩国和朝鲜的母语者在安静无回声环境中录制，内容覆盖经济、娱乐等多个领域，每个说话者贡献约400句话语，转录错误率低于5%，采用16kHz WAV格式，可商用。