With the rapid development of network, traditional music charts do not work well with the popularity of musical recordings. The search charts of music do pretty well with that. However, there are many new questions with search charts need to be concerned and solved. In this paper, we choose the special keywords on music search charts to study. The keywords of music search charts are special thanks to the online user behaviors. Based on the special keywords, we proposed a novel music categorization method. The criterion of our method is the correlation between the special keyword and the common one of the same song. We separate the music into three types with the method. Our study will be utilitarian in future because it is based on user behaviors.