Melon Playlist Dataset

Melon Playlist Dataset是由Kakao公司从韩国流行的音乐平台Melon收集的一个公开数据集，包含649,091首歌曲的mel-spectrogram和148,826个由30,652个不同标签标注的播放列表。该数据集反映了韩国的音乐消费情况，主要包含韩国流行音乐和西方音乐。数据集的创建过程涉及从Melon用户中选择并验证高质量的播放列表，这些用户被称为Melon DJs。此外，数据集还包括歌曲和播放列表的元数据，如标签、标题和艺术家信息。Melon Playlist Dataset适用于音乐信息检索任务，特别是自动标签和自动播放列表续播，旨在解决音乐推荐系统中的冷启动问题，并支持深度学习方法的研究。

Melon Playlist Dataset is a public dataset collected by Kakao from South Korea's popular music streaming platform Melon. It contains mel-spectrograms for 649,091 songs and 148,826 playlists annotated with 30,652 unique tags. This dataset mirrors the music consumption patterns in South Korea, predominantly featuring K-pop and Western music. The construction of this dataset involves selecting and validating high-quality playlists from Melon users known as Melon DJs. Furthermore, the dataset includes metadata for both songs and playlists, such as tags, titles, and artist information. The Melon Playlist Dataset is applicable to music information retrieval tasks, particularly automatic tagging and automatic playlist continuation, aiming to address the cold-start problem in music recommendation systems and support research on deep learning methodologies.