数据堂—535小时德国人说英语手机采集语音数据

535小时德国人说英语手机采集语音数据是由1162名德国人参与录制，口音正宗。录音文本由语言专家参与设计，涵盖通用、交互、车载、家居等多类别，内容丰富。文本经过人工校对，准确率高。匹配主流安卓、苹果系统手机。本套德国人说英语手机采集语音数据可用于语音识别、声纹识别模型训练、机器翻译语料库的建设或算法研究。

This 535-hour English speech dataset was collected via mobile phones, with recordings made by 1,162 German participants who speak authentic German-accented English. The corresponding transcribed texts were designed in collaboration with linguists, covering a wide range of categories including general scenarios, interactive scenarios, in-vehicle scenarios, smart home scenarios and more, resulting in rich and comprehensive content. All transcribed texts have undergone manual proofreading to guarantee high accuracy. It is compatible with mainstream Android and Apple iOS mobile phones. This dataset can be utilized for training speech recognition and speaker verification models, constructing machine translation corpora, or conducting algorithm research.

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535小时由1,162名母语为德语的参与者通过手机采集的英语语音数据，专用于测试英语语音识别模型。音频为16kHz、16位、单声道WAV格式，内容涵盖通用话题、交互场景等多个类别，所有文本均经人工校对，准确率不低于98%，适用于语音识别模型训练和算法研究等商业用途。