Online music users are to benefit from a pioneering new way of discovering lesser known artists and songs developed by researchers at Robert Gordon University (RGU).
Researchers from RGU’s Smart Data Technologies Centre and the School of Computing Science and Digital Media have introduced a new approach to online music recommenders based on a more comprehensive form of social tagging that learns from both human tagging and audio content.
The new approach was recently introduced at an international conference in an award winning paper which highlights how this pioneering method will allow people to discover new and lesser known music that suits their tastes.
Professor Susan Craw, one of the researchers and co-author of the paper which was presented at the 23rd International Conference on Case-Based Reasoning in Frankfurt, explains: “Millions of people use online music services every day and recommender systems are essential to browse these music collections.
“Users are looking for high quality recommendations, but also want to discover tracks and artists that they do not already know, new releases and the more niche music found in the ‘long tail’ of online music.
“While traditional tag-based recommenders benefit from useful semantic information for recommendation including genres, topics and opinions, they are inherently biased towards popular tracks and are not effective in less mainstream online music.
“By learning to augment existing tagging, our recommender generates quality recommendations while helping users to discover music.”
The new recommender exploits the combined knowledge from audio and tagging to extend a track’s social tagging by adding tag-based knowledge from similar music tracks.
Professor Craw continues: “We conducted an online study with real users as well as a larger experiment using Last.fm user data. Both show that our new recommender system provides better quality recommendations than when only social tags are used, and increases the discovery of new, unknown and niche music.
“This approach of using information from similar songs may offer opportunities to improve other online services such as image browsing, movie recommendation and online shopping.”
The new approach has been developed by Professor Susan Craw, RGU research fellow Dr Stewart Massie, and Dr Ben Horsburgh who previously had been a Computer Science undergraduate, PhD student and research fellow at RGU, and is now a senior data scientist at Tesco PLC.
All three are authors of the paper entitled ‘Music Recommendation: Audio Neighbourhoods to Discover Music in the Long Tail’ which was awarded the ‘Best Paper’ accolade at the conference.