Follow Us:

In the deliverable ‘D5.4 – Music Platform pilot report v1, the FuturePulse team presents the results from the first pilot phase for the Background Music Platform (BMP) use case in the FuturePulse project. The first pilot phase is the smallest pilot of the three pilots being conducted, and consists mainly of internal testing of FuturePulse models for use case specific requirements.

In specific, the first pilot has been conducted by Soundtrack Your Brand through 4 iterations:

  1. A pre-pilot small scale test of the requirement for the Recognition level of a track (September - October 2018).
  2. A large online crowd experiment test of the requirement for the Recognition level of a track (November - December 2018).
  3. Internal testing and validation of the requirements for: i) Instrumental or vocals, major gender in track, ii) Fade in and fade out of a track and iii) Major or minor in a track (November 2018 - May 2019).
  4. Analysis of Soundtrack Your Brand clients real time usage of requirements for the Genre of a track, the Energy level in a track, the Original release year for a track and the Moods related to a track (January - May 2019). 
Collected data have been clustered and analysed, and the results of the first pilot phase are showing both strengths and future challenges regarding the FuturePulse solutions and functions for Music Platforms.

The first pilot phase for the Music Platform use case focused on testing the most innovative FuturePulse feature and requirement, –namely the Recognition level of a track, to such an extent that the model was verified through an online crowd experiment and the results published in a peer review conference (ISMIR 2019). As Daniel Johansson, Patrik Axelsson and Magnus Rydén from Soundtrack Your Brand (SYB) mention, ‘the model takes into account the cognitive perspective of the problem using exponential “forgetting curves” as the main mechanism for the estimation of collective memory decay and sigmoid “learning curves” for the estimation of the initial collective learning degree, with regards to songs.’

Considering this, the two online crowd experiments were used as tools for evaluating the requirement for the recognition level of a track during the first pilot. Out of the SYB catalogue of tracks that are streamed on the platform in Sweden, the 600 most recognised songs, as well as the 600 least recognised songs, were chosen based on the CERTH algorithm (T-REC) developed in the FuturePulse project. 50 tracks were then chosen randomly for each category, as representative for high recognition levels and low recognition levels.

An experimental group of 1041 online respondents in Sweden were introduced to the tracks and had the opportunity to answer how well they recognised songs. For the simplicity of the respondents, they were randomly divided into 10 groups with 10 randomly chosen tracks in each group (5 high level rec. and 5 low level rec.). We made use of 30 second samples of the tracks to which the respondents reacted whether he/she recognised the track.

The result from this pilot test where very promising. Some of the findings are very important though for the Music Platform use case, specifically when it comes to providing high level composites of variables for music choices on the streaming platform:

  • A track needs almost 7 weeks in the charts to reach higher levels of recognition and achieve a lower velocity down toward “oblivion”, and 25 weeks on the charts to reach such a level that it becomes highly recognizable for a longer period of time; 
  • The model itself seems to be highly competent in predicting how well an audience in a specific country is recognizing a certain track in a large catalogue. 
Last but not least, we must point out that overall the first pilot phase gave us new and valuable insights into how end users will exploit the FuturePulse functions in their day-to-day environment as business owners. It also gave us insights into which novel features to focus upon during the last phase of the project. 

Read the deliverable ‘D5.4 – Music Platform pilot report v1 here.

Photos by

Share This Story, Choose Your Platform!
European flag

Co-funded by the European Commission

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 761634. This website reflects the views only of the Consortium, and the Commission cannot be held responsible for any use which may be made of the information contained herein.