For my Master's thesis project I created a music recommendation system prototype based on automatically extractable data from audio files. In order to do that, I studied the latest progress in scientific research related to Music information retrieval through open documentation published by ISMIR (International Society of Music Information Retrieval). Finally I conducted both quantitative and qualitative analysis on experimental and commercial music softwares available on the internet.
Most popular music recommendation systems (such as Spotify) do not offer any visual suggestion on music content. This means that users are totally unaware of music content until they listen to the tracks one by one. A time-consuming and often frustrating operation. To solve this problem, Musigraph uses a custom-designed visualization which describes up to 5 different music features. This allows even less-experienced users to visually recognize similarities and differences between tracks. Moreover, they can detect several musical features at a glance.
The main purpose of the design is to allow users to customize, as accurately as possible, his music listening and browsing experience. Compared to others traditional music recommendation system, such as last.fm and spotify, Musigraph offers:
- a wider exploration of the online music database;
- specific responses to different needs of the users, thanks to the extended set of search filters available;
- an active music listening and search experience. Through the dialogue with the user, Musigraph is, in fact, able to generate music learning processes.
To customize the recommendation one can modify search filters through specific selectors. When the search begins, all the selected parameters are combined to determine the outcome of the recommendation.
Then I presented three different scenarios describing possible applications of Musigraph in real-life situations. Using videos and step by step screenshots I explained how this software can actually meet the needs of very different users. In this video a first time user discovers new artists while customizing his recommendation.