CLEAR

CLEAR数据集是由谢布鲁克大学电气与计算机工程系的研究团队开发的，用于声学问题回答（AQA）任务。该数据集包含50,000个声学场景，每个场景由10种乐器的声音组成，总计200万个问题和答案。数据集的创建过程涉及使用Good-Sounds数据库中的音乐录音，通过功能程序生成问题和答案，以探索声音属性之间的关系。CLEAR数据集旨在推动声学场景的AI和推理研究，特别适用于检测机器异常、危险事件识别及工业和社会机器人领域。

The CLEAR dataset was developed by a research team from the Department of Electrical and Computer Engineering at the University of Sherbrooke for the Acoustic Question Answering (AQA) task. It contains 50,000 acoustic scenes, each composed of sounds from 10 musical instruments, with a total of 2 million question-answer pairs. The dataset creation process uses musical recordings from the Good-Sounds database and generates questions and answers via functional programs to explore the relationships between sound attributes. The CLEAR dataset aims to advance AI and reasoning research for acoustic scenes, and is particularly applicable to machine anomaly detection, hazardous event recognition, and the fields of industrial and social robotics.