基于标签的歌曲相似性研究

基于音乐语义标签的音乐相似计算研究是音乐信息检索领域的另一个新的热点。该文提出一种基于标签挖掘的歌曲分类方法,以Last.fm音乐网站上的用户标签为特征进行歌曲相似性研究。文中将文本聚类中常用的潜在语义分析（LSA）方法和改进的K-means聚类方法相结合,应用于音乐语义标签的自动抽取;从音乐网站last.fm上抽取了6大类600首歌曲的8000多个用户标签作为音乐语义特征,并利用LSA进行歌曲向量的降维,形成了一个表示歌曲间相似关系的600×150维向量矩阵。最后利用K均值,根据音乐歌曲间的相似度进行歌曲分类,完成歌曲相似性比较。实验结果同没有LSA降维前及已有的HCC结果比较表明,使用文中提出的基于音乐标签的模型对歌曲进行分类,能得到较好的分类效果。

Tag-based Research of Music Similarity

Music similarity calculation based on semantic tags is a new hot spot in music information retrieval.In this paper,a method for music classification using tags on web mining has been presented,and make the user tags from last.fm as features to study the music similarity.We apply Latent Semantic Analysis (LSA) was appljed and K-means clustering method was improvedto automatic tag extraction.More than 8,000 tags of 600 rongs of 6 categories extracted from Last.fm are as music semantic feature.LSA is for dimension reduction and form a 600 × 150 matrix representing the similarities of songs.Finally,improved k-means get the clustering results according to the similarity between music.By the comparison of clustering results of the proposed method,clustering without LSA and Hierarchical Combined Clustering,our proposed method in this paper can get a better result in music classification.