数据堂—505小时客家话自然对话手机采集语音数据

客家话自然对话手机采集语音数据，基于30余个常见主题来模拟录制。此数据集标注了文本内容、句时间戳、说话人身份、性别等多种属性，由近900名来自广东地区的客家话使用者录制，准确性高，为语音识别相关研究及应用提供了丰富的资源，经多家AI公司验证：有助于模型面对真实世界的多样性时能够表现出色。我们严格遵循数据保护法规和隐私规定，确保数据采集、存储和使用的过程中维护用户的隐私和合法权益，所有数据均遵循GDPR,CCPA,PIPL。

Hakka natural conversational speech data collected via mobile phones. This dataset was recorded through simulated natural conversations covering over 30 common topics, and annotated with multiple attributes including textual transcripts, sentence-level timestamps, speaker identities and genders. Recorded by nearly 900 Hakka speakers from Guangdong Province, this high-accuracy dataset provides abundant resources for speech recognition-related research and applications. Verified by multiple AI companies, this dataset helps models deliver outstanding performance when dealing with real-world diversity. We strictly abide by data protection regulations and privacy rules, ensuring that user privacy and legitimate rights and interests are well-protected throughout the entire process of data collection, storage and utilization. All collected data complies with GDPR, CCPA and PIPL.

搜集汇总

背景与挑战

背景概述

该数据集包含505小时的客家话自然对话语音数据，由894名来自广东地区的客家话使用者通过手机采集，覆盖30多个常见主题，并标注了文本内容、时间戳、说话人身份和性别等信息。数据适用于语音识别和声纹识别研究，字符准确率达98%，并严格遵守数据隐私法规。