352小时全双工乌尔都语自然对话手机采集语音数据

全双工乌尔都语自然对话手机采集语音数据，基于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，由来自不同地域和文化背景的本土人录制，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。Full-duplex Urdu natural conversational speech data collected via mobile phones, recorded by simulating real-life scenarios centered on common topics. This dataset is annotated with multiple attributes including transcript text, sentence-level timestamps, speaker identity, gender, and others. Recorded by native speakers from diverse regional and cultural backgrounds, it features high accuracy and provides rich resources for speech recognition-related research and applications. Validated by multiple AI companies, this dataset has been proven to improve model performance when dealing with real-world diversity. We strictly abide by data protection laws and privacy regulations to safeguard user privacy and legitimate rights and interests throughout the entire process of data collection, storage and utilization, and all 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352小时通过手机采集的乌尔都语全双工自然对话语音，总容量为84GB。数据由来自不同地域和文化背景的本土人录制，并标注了文本内容、句时间戳、说话人身份及性别等多种属性，适用于语音识别、巴基斯坦语音助手等相关研究与应用。数据采集与处理严格遵循GDPR、CCPA和PIPL等数据保护法规。