MetaScore
收藏资源简介:
MetaScore是由加州大学圣地亚哥分校的研究团队创建的一个大型音乐数据集,包含963,000条音乐乐谱及其丰富的元数据。数据集内容包括音乐的流派、作曲家、复杂度、调号、时间签名以及用户交互统计等。数据集的创建过程包括从MuseScore论坛收集数据、使用机器学习算法填补缺失的流派标签,并利用大型语言模型生成自然语言描述。MetaScore主要应用于可控音乐生成领域,旨在通过文本或标签生成符合特定描述的音乐作品。
MetaScore is a large-scale music dataset developed by a research team at the University of California, San Diego. It contains 963,000 musical scores along with rich metadata, including music genre, composer, complexity, key signature, time signature, user interaction statistics and other relevant information. The dataset construction process involves collecting data from the MuseScore forum, imputing missing genre labels via machine learning algorithms, and generating natural language descriptions using large language models (LLMs). MetaScore is primarily applied in the field of controllable music generation, with the goal of generating musical works that meet specific descriptions through text or labels.




