This article analyzes reviews of listed songs on QQ music platform. According to the popular index of QQ music peak list, we compare the top five popular songs with those unpopular ones. Our work embodies text and sentiment analysis for evaluating the content and activity of the reviews, social network analysis for evaluating structural characteristics of the networks, and finally the K-means clustering method for getting features from the classified nodes for further comparison. The conclusion is that people prefer to comment on popular songs; The popularity of songs is not related to whether people prefer to use positive text for song comments; Movies, TV shows, variety shows, and other video media can trigger commentators' discussions about a song; The implicit network of popular songs' comments has scale-free, small world characteristics; The implicit network of popular songs' comments has more opinion leaders and information disseminators compared with general songs', also the ordinary commentators in popular songs network are more diverse.