1,652小时粤语手机采集语音数据

中国粤语语音数据_朗读（手机），基于给定的脚本朗读并模拟录制，由4888名来自广东省发音人，在安静的室内环境下的录音数据。录音内容广泛，覆盖50 万句常用口语语句，平均重复次数1.5次，平均句长12.5字，包括微博高频词、日常用语等。文本经过人工校对，准确率高，匹配主流安卓、苹果系统手机，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。

Chinese Cantonese Read-aloud Speech Dataset (Mobile Phone): This dataset comprises audio recordings of speakers reading aloud from given scripts under simulated mobile recording scenarios, collected from 4888 speakers across Guangdong Province in quiet indoor environments. The recordings span a diverse range of content, including 500,000 commonly used spoken utterances, with an average repetition count of 1.5 per utterance and an average sentence length of 12.5 Chinese characters. The corpus covers high-frequency terms from microblogs, daily conversational phrases and other common spoken content. The accompanying texts have undergone manual proofreading with high accuracy, and the dataset is compatible with mainstream Android and iOS mobile devices, providing a valuable resource for research and applications in the field of speech recognition. It has been validated by multiple AI companies that this dataset enables models to achieve excellent performance when dealing with real-world diversity. We strictly adhere to global data protection regulations and privacy norms, ensuring that user privacy and legitimate rights and interests are fully protected throughout the entire lifecycle of data collection, storage and utilization. All data within this dataset comply with GDPR, CCPA and PIPL.

搜集汇总

背景与挑战

背景概述

该数据集包含1,652小时粤语手机采集语音，由4,888名广东发言人在安静室内录制，覆盖50万条常用口语句子，文本经人工校对且准确率高，适用于语音识别研究。数据格式为16kHz WAV，遵循相关隐私法规，版权归DataTang所有。