With so much choice available online, music fans the world over often face a dilemma in terms of choosing what music to listen to for a given mood. The issue existed when one had a limited collection of “vinyl” to choose from but was perhaps not quite as acute given the potential for anyone of us with an internet connection to be able to listen to almost any available piece of music recorded music with the simple tap of an icon.
Music recommendations systems have been around for many years, almost as long as music downloads and streaming have been available to Internet users. Some work better than others, often simply recommending other artists in a genre. Research in the International Journal of Reasoning-based Intelligent Systems has taken a new approach and use multi-label tags associated with different songs to personalise your playlist based on the emotional content of the songs available.
The system has been developed by Yuan Luo of the Academy of Music and Dance at Hunan City University in Yiyang and Qiuji Chen of the Wenzhou Yue Theatre in Wen Zhou, China. It uses principal component analysis to analyze the emotional aspects of music. By reducing the dimension of music features, the system can process large amounts of data more efficiently. This analysis is combined with a method called cosine similarity, which calculates the similarity between songs based on their emotional content. The system also applies multiple labels to better define the character of each song.
To create personalized recommendations, the system calculates your interest in dozens of different emotional labels. This allows it to recommend music that aligns with your mood and preferences. In testing, the system was found to have a high accuracy rate of 98.3%. This means that the recommended songs were in line with the user’s preferences almost all the time. Additionally, the system is very fast, able to recommend 500 pieces of music in just under twenty seconds.
Luo, Y. and Chen, Q. (2023) ‘A method for personalised music recommendation based on emotional multi-label’, Int. J. Reasoning-based Intelligent Systems, Vol. 15, No. 2, pp.97–104.