ABC_DJ: Music & Brand Values



BETTER SOUND 2018 | Category: Research & Development

HearDis! GmbH


The multi-national EU research and development project ABC_DJ provides groundbreaking software solutions for any given audio branding scenario. Focussing on the systematic identification of brand-fitting music, its advanced recommendation algorithm predicts the perceived brand-fit of music within particular target groups with an accuracy of 80.1%.

Project description

Turning Brand Values into Music.

ABC_DJ (Artist to Business to Business to Consumer Audio Branding System) is an international research and development project funded by the European Union’s Horizon 2020 research and innovation programme under grant agreement No 688122. Integrating both the perspectives of marketing executives and consumers, the scientific core objective of the project is constituted by the systematic translation of brand identities to the musical domain.

In an unprecedented and interdisciplinary approach the semantic gap between sender (brand) and receiver (consumer) of a musical message (i.e. the difference in respective semantic interpretations of perceived musical content) has been bridged. Capturing adjectives marketing experts state with regards to being essential for brand communications and describing brand identities, the results of two extensive initial studies were transferred to a final list of 36 terms (e.g. “bright”, “authentic”) creating the General Music Branding Inventory (GMBI). The GMBI was then utilised in two large-scale listening experiments, where 10,144 participants from the UK, Spain and Germany were asked to rate the fit of each item from the inventory to four (1st exp.) respectively six (2nd exp.) randomly chosen and 30-second-long music excerpts typically comprising 1st verse and chorus of a song.

In compliance with the novel ABC_DJ taxonomy, the music corpus employed was composed of 9 pieces associated to each of the 61 musical styles (e.g. “Disco”, “Indie-Rock”) resulting in 549 original tracks. All individual samples of the music corpus were analysed with regards to 392 hard (low-level, e.g. “BPM”, “frequency response”) and 108 soft (high-level, e.g. “style”, “timbre”) features making use of both music information retrieval and machine learning techniques. Furthermore socio-demographic data as well as the affiliation of participants to the SINUS-Milieus were gathered in the listening experiments allowing for the integration of social data in the system. Interconnecting the ground truth captured in the listening experiments with MIR and machine learning, the research team of ABC_DJ was able to develop a powerful recommendation algorithm capable of predicting the perceived brand-fit of any musical content within particular target groups with an accuracy of 80.1%.

On the development side of the project, which concentrates on creating tools improving the workflow and competitiveness of European creative agencies and European artists, this algorithm is implemented in the “Brand Filter/Prediction” software, the tool representing the heart of the modular ABC_DJ system. The cutting-edge and mostly browser based ABC_DJ applications range from a novel music library manager to specialized software for audio branding visualisations and unique in-store music solutions which can be employed individually or combined in any possible manner according to the needs of the user.