数据堂—303小时中英混读手机采集语音数据

该数据由1113名中国本土人员参与录制，口音覆盖七大方言区。录音文本均为中英混合句子，涵盖通用场景及人机交互场景，内容丰富，转写精准。可用于改善语音识别系统对中英混读语音的识别效果

This dataset was recorded with the participation of 1113 native Chinese speakers, whose accents cover all seven major Chinese dialect regions. All recorded utterances are Chinese-English code-switched sentences, covering both general daily scenarios and human-computer interaction scenarios, featuring rich content and highly accurate transcriptions. It can be utilized to improve the speech recognition performance of automatic speech recognition systems on Chinese-English mixed speech.

搜集汇总

背景与挑战

背景概述

该数据集是一个包含303小时中英混读语音数据的集合，由1113名中国本土人员录制，覆盖七大方言区，录音内容为中英混合句子，适用于改善语音识别系统对中英混读语音的识别效果。数据格式为16kHz的wav文件，标注准确率达97%，版权归数据堂所有，为商用数据。