In recent years, deep learning technique has received intense attention owing to its great success in image recognition. A tendency of adaption of deep learning in various information processing fields has formed, including music information retrieval (MIR). In this paper, we conduct a comprehensive study on music audio classification with improved convolutional neural networks (CNNs). To the best of our knowledge, this the first work to apply Densely Connected Convolutional Networks (DenseNet) to music audio tagging, which has been demonstrated to perform better than Residual neural network (ResNet). Additionally, two specific data augmentation approaches of time overlapping and pitch shifting have been proposed to address the deficiency of labelled data in the MIR. Moreover, an ensemble learning of stacking is employed based on SVM. We believe that the proposed combination of strong representation of DenseNet and data augmentation can be adapted to other audio processing tasks.

No abstract

With the booming development and popularity of mobile applications, different verticals accumulate abundant data of user information and social behavior, which are spontaneous, genuine and diversified. However, each platform describes users' portrait in only certain aspect, and it is difficult to combine those internet footprints together. In our research, we proposed a modeling approach to model users' online behavior across different social media platforms. Users' data of same users shared by NetEase Music and Sina Weibo was collected for crossplatform modeling of correlations between music preference and other users' characteristics. Based on music genre and mood tags, users are clustered into five typical genre groups and four typical mood groups by analyzing their collected song lists. Moreover, based on collected Weibo data of same users, correlation between music preference (e.g. genre, mood) and Big Five personalities and basic information (e.g. gender, region, self-description tag) have been comprehensively studied, forming full-scale user profiles with finer grain. The results indicate that people's music preference can be reflected by their social activities in many ways. For example, people living in mountainous areas generally have favor for Folk music, while people living in rich regions tend to like Pop music. Meaningly, dog lovers prefer Sad music compared to cat lovers. Further discussion has been suggested in our paper. Promisingly, our proposed cross-platform modeling approach could be extended to other verticals, providing an online automatic way for profiling users more precisely and comprehensively.

No abstract