Periodic Reporting for period 2 - ABC DJ (Artist-to-Business-to-Business-to-Consumer Audio Branding System) Reporting period: 2017-07-01 to 2018-12-31 Summary of the context and overall objectives of the project The EU H2020 project ABC_DJ (Artist-to-Business-to-Business-to-Consumer Audio Branding System) set out to provide new tools for European creative agencies which are active in the field of audio branding, i.e. the exploitation of the power of music in various commercial contexts. This actively included creators of music, (independent) labels as well as respective multipliers into the audio branding value chains. A major research aim of the project was to predict the perceived musical expression in the branding context via acoustic and musical features. To measure the perceived expression of music among different target groups, another major scientific challenge was the creation of a suitable list of attributes that is both able to describe the perceived expressive qualities of music from the consumer perspective and useful for describing intended brand identities in the context of audio branding. This was achieved to a very high satisfaction; the resulting General Music Branding Inventory (GMBI) is the first of its kind which is both scientifically founded and agreed upon by branding business experts. Thus, the research in the ABC_DJ project laid the groundwork for an important new application domain of future music recommendation systems. Several ICT tools were designed to enhance and facilitate the full gamut of audio branding processes. As the core entity of the ABC_DJ system architecture the Music Library Manager manages and stores music file metadata and standardises the communication methods and formats between the three other tools. These are namely: a) The Audio Branding Tool - an interactive browser application with the purpose of visualising criteria of a music selection to non-music-experts like marketing staff, and to make this music audible for communication processes, b) the Playlist Generator - enabling professional users to plan, manage and create sophisticated playlists, and c) The Cockpit Unit - enabling the branding agency personnel to control and manage in-store players remotely and from one single interface. Besides the development of novel tools and services, continuous monitoring and regular market trends research was carried out on creative and commercial business fields regarding existing tools, applications, products and services. The project further entailed an analysis of a range of legal aspects, of the landscape of European music creators and their respective multipliers in an audio branding and in-store music context and offers novel monetising schemes for stakeholders. Overall, ABC_DJ offers suitable means to include the high number of European creative micro-businesses and SMEs into a value chain. Here, many audio branding agencies shall be able to serve the diverse needs of many brands and counter the predominant 'one-sound-fits-all' situation. In doing so, the pool of music which is exploitable for branding services will be increased significantly. To guarantee the maximum impact of the project, final project results are disseminated and communicated to public, scientific, commercial, and artistic audiences. Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far Work of the coordinator TU Berlin (DE) focused on the development of a general vocabulary to describe the expressive content of music and to establish prediction models for the recommendation of brand fitting in-store music. Therefore, in the first step, international experts from the respective fields were invited to a music branding expert workshop to develop a comprehensive list of words being relevant to describe music in the branding context. After shortening this list in the course of an online survey with international marketing experts, two large-scale online listening experiments with over 10,000 participants from three countries (DE, ES, UK), age cohorts, and educational backgrounds, and both genders were carried out to generate a so-called ground truth about how people perceive the expressive content of music based on their socio-demographic background. The 500 music titles used in the experiment included 10 different major genres and 61 subgenres of the music library provided by the partner HearDis! (DE). Findings from the first online experiment suggest a 4-factor structure of our vocabulary (GMBI) describing the dimensions Arousal, Valence, Authenticity, and Timeliness. Results further revealed significant effects of country, education, and gender on people’s ratings. A second area of research, carried out by IRCAM (FR), concentrated on the development of new audio features to predict the perceived musical expression as well as to characterise the audio quality of music tracks and to improve music mixes. These features have either been derived from previously estimated higher-level concepts (such as structure, key or succession of chords) or by developing new signal processing algorithms. These regression models explain up to 80% in the variation of the musical expression as perceived by the participants of the experiments.Investigations regarding legal frameworks in the context of audio branding as well as entailed stakeholders, which are conducted by the partner Lovemonk (ES), have been accomplished. Among other things, all 91 EU-based Collective Management Societies have been included in a survey which for the first time gives a comprehensive overview on the various different methods and payment schemes in place to date.Key software tool developments within the project have been developed by the partners HearDis! and FINCONS (IT):- Music Library Manager ('iTunes for professionals')- Audio Branding Tool ('communicating the invisible')- Playlist Generator ('sophisticated music playlists along a dramaturgy')- Cockpit Unit ('remote control for in-store applications')All single developments as well as the overall system were evaluated by the market research partner INTEGRAL (AT) along 80 qualitative and quantitative KPIs, as well as with the help of sensor data collected in the store of the partner Piacenza (IT). Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far) On a scientific level, the project progressed beyond state of the art with the definition of a) a standardised methodology for audio branding processes and b) a vocabulary of human emotions and values, which constitutes a commercially applicable music description scheme for branding. In combination with MIR methods this allows for the creation of completely novel supporting tools on the implementation level. The final prediction model relies on novel random forest and multi-level regression analyses taking into account socio-demographics and the milieu membership of a listener. The final Playlist Generator incorporates an innovative algorithm capable of arranging audio tracks according to dramaturgic(!) rules set by a professional user like a music consultant. These rules refer to the appertaining metadata and enable desirable combinations of features or restrictions of combinations. The final Audio Branding Tool allows users to visualise various features of a music selection, like hard (low-level, e.g. BPM) and soft (high-level, e.g. instrumentation, semantic annotations according to the GMBI) features associated with a particular song. As a result professional users can generate music playlists with a brand fit of unprecedented accuracy, in a fraction of the time. Agencies can monitor their players remotely. And, as all tools were developed as browser applications, operating system in use no longer constitute an obstacle. Factor loading matrix for preGMBI Items and associated factors (study 3)
