This Use Case is part of nextMedia.Hamburg
Prototyping Lab: Music Prediction Machine
N-Joy is a hit radio program of the NDR with a younger target group that mainly wants to listen to music. Streaming services such as Spotify and Amazon Music have created a new challenge for radio programs. In order to remain successful in this competitive situation, the selection of new songs must be even more accurate. The music editorial team selects the new titles from 100-200 new publications that the listeners might like. However, it wasn't until weeks later that the radio station knew through market research which song was well received. An AI should help (the music desk) to select new songs that the listener will most likely like.
The Music Prediction Machine is supposed to examine the music titles for similarities and assess how well a new song is likely to be received by the listeners. The accuracy of the AI can be checked in the test phase with the help of survey results and 500 songs that have been rated in advance by the editorial team. N-Joy can thus become the first German operator of music prediction software for radio stations.
The developed prototype consists of three components: a music analysis tool, the AI and a user interface. Song files are read in the analysis tool and characteristics such as speed, key, but also more complex properties such as "danceability" are examined. In addition, the text of the song is analyzed and the proportion of positive and negative words is determined. A total of 33 characteristics can be analyzed per song. The AI is a simple neural network that learns the relationship between the characteristics and the callout rating of a song and can thus make predictions for new songs. To ensure user-friendliness for editors, a simple graphical user interface was developed for Linux systems.
The concept and the prototype were presented internally and will be tested with several customers after further preliminary work.