Evaluating Sentiment Similarity of Songs Based on Social Media Data

Music is related to the emotions and sentiments of human beings. However, for a long time, there are few emotional or sentimental factors involved in the research of music recommendation. In order to meet individual user's needs of music recommendation, and then to increase users' satisfaction on the performance of music platforms, this article seeks to address the two issues of how to distinguish the sentimental texts from massive user reviews, and how to find out the emotions/sentiments in music more accurately. To this end, we first construct a text sentiment classifier based on deep learning to extract the sentimental texts from online reviews. After obtaining the sentimental reviews of musics, an algorithm is then proposed for extracting features of each song, which can be used to identify the sentimental information of the song. Finally, the classifier constructed by these features may be used in retrieving songs having similar sentiment.