Listen Local

An open collaboration of music curators, open-source software developers, and the Digital Music Observatory to help decolonize the local music ecosystems

“Big data creates injustice.” – Cathy O’Neil, author of Weapons of Math Destruction

Listen Local is an open collaboration of music curators, open-source software developers, and the Digital Music Observatory to help decolonize the local music ecosystems.

In Europe, the music sector comprises hundreds of thousands of microenterprises, small enterprises, and freelancer networks that do not employ a data scientist, data engineer, statistician, or royalty economist.

This fragmented structure makes European music very vulnerable in the age of global streaming, when most music discovery and music sales take place on international platforms that have access to the data of hundreds of millions of songs and artists, hundreds of millions of fans, and can train algorithms to optimize their sales profit.

Table of Contents

Help for the local music ecosystem

We want to help the local music ecosystems: artists and their fans, local curators, radios, labels, publishers, collective management, cultural heritage organizations, and local government with open source software, open data, and open scientific results. Without a specialist data and IT team, they cannot enjoy the benefits of rights management and documentation automation, machine learning for curation, A&R, marketing, and tour planning.

Our Digital Music Observatory collaboration of music researchers, open source software developers, and industry practitioners offer the ability to gather data collectively, rely on open source software, and build know-how. With our help, we want to strengthen local music ecosystems’ competitiveness in the global streaming era.

We want to strengthen the competitiveness of local music ecosystems in the global streaming era. Our work is helping these stakeholders to work more efficiently and compete better in global markets:

  1. Local curators (for festivals, clubs, radio stations, streaming playlists, and music export offices) will be able to find relevant music according to location, to place their artists into the correct scene, venue, or playlist. We want to hear more music created in Vilnius in the stores and radios of Vilnius. We want to ensure that visitors to the city will take home new music discoveries.
  2. Local artists will be more visible on global streaming platforms and the increasingly global selection pool of radios or festivals. Twenty years ago, artists competed for the attention of their local gatekeepers and competed with about 50,000 other songs. On global platforms, they compete with about 100,000,000 songs. Reach is a big problem if your music is not visible. We are prototyping ways to check and improve your online reach.
  3. Music lovers will be able to find new artists they can personally meet and see on small stages in the neighborhood where they live, study or work, or go on vacation. The current for-profit and intransparent AI recommenders recommend music from only about 50 cities worldwide, and none of them are in Europe. We will provide recommendations from the city where you actually are.
  4. Musis businesses, i.e. small labels, publishers, talent managers, and their represented artists, very often do not get paid at all, or do not get fully paid on the complex, AI-driven autonomous systems of global platforms. These platforms require perfect, tedious, detailed, and frequently updated data about every recording in order to pay royalties correctly. The documentation cost is greater than the expected revenue for small labels or small country artists in years. We want to help with open source solutions that bring automation and scale to these processes.
  5. Music export offices, music information centers, and collective management organizations in small countries do not have the cost and talent base to take benefit from modern data science when it comes to working with data on the scale and cheap. We want to help them unleash the advantages of the semantic web, often called web 3.0. This way, we can radically reduce their IT and documentation costs and increase the quality of their data to perform their artist-supporting work better.
  6. National, regional and city governments (and in decentralized countries, regional and city governments) ensure that algorithms do not undermine their cultural and copyright policy or music education goals. We want to ensure dominant global music sales platforms, and AI-driven autonomous recommendation systems recommend local music to local audiences or relevant foreign audiences. We want to ensure that children and teenagers find music pertinent to the local government’s general music and cultural cohesion objectives.

Our findings in Slovakia

In our first feasibility phase, we found that most Slovak artists are only partially getting paid on global streaming platforms, and they get paid less than in many other EU countries.

We also found that 17% of the Slovak repertoire is not understandable for the Spotify recommender system. This music remains invisible and earns no royalties. We are confident that another part is poorly understood by the algorithms and often recommended to the wrong audience. Negative recommendation outcomes will build up an adverse listening history for these songs and they will eventually be buried by the algorithm.

Based on our findings, we hypothesize that in small countries it may be more costly to adequately describe the music for the global platforms than the expected revenue from releasing the recording. The situation is even worse for self-released artists or artists managed by small independent labels who do not have the IT and data skills to automate copyright and marketing data management.

Our Feasibility Study is written for policymakers, educators, music managers, artists, and collective management organizations. It is a non-technical introduction to promoting small country repertoires globally and avoiding the colonization of local music ecosystems by globally promoted music.

Plans in Lithuania

In Lithuania, we want to focus on the effect of documentation, mainly information that is not part of the copyright management metadata. Music recommender systems do not only rely on the listening histories of music fans and the rights management data provided by self-released artists and labels via distributors. They use data from various web sources to understand the music and the artist’s scene to place the recording in front of the right audience.

Whatever data was uploaded to the web by an artist, label, or distributor, it may have changed or contain an error. Different data may exist with a collective management agency, the record label, or global knowledge hubs like music information libraries, Wikidata or MusicBrainz. As algorithmic recommenders for playlists, radios, festivals, and venues use this data, it is essential to periodically review, correct, and improve the information available by the artist and the music itself.

The way to do this cost-effectively and error-free is using the semantic web or web 3.0 technologies. The semantic web contains data standards that make the information, independent of the Lithuanian or Slovak, or other languages, readable and importable by other databases, such as the global streaming platforms or festival promoters. The web 3.0 is organized around knowledge graphs, for example, knowledge hubs of music that radios, streaming platforms, booking agents, or festival promoters all use. Placing information on a knowledge graph is not for the faint heart and certainly falls beyond the skills expected from a small cultural institution or record label.

  1. We will ask artists and their representatives (labels, publishers, LATGA, and AGATA) to share some of their data with us so we can make a health check on their data in some parts of the web. We will do a limited number and scope of health checks supported by modern data science.
  2. We will ask their permission to correct data after their consent and measure to what extent this translate to positive outcomes (more recommendations going in their direction, higher visibility, increasing income.)
  3. We will ask music institutions that deal with the information about Lithuania’s cultural heritage (like the national library) or the heritage of music (like the Music Information Centre Lithuania) to allow us to sync their music with global knowledge hubs on the semantic web to see how much this increases the visibility and use of the works of the artists that they document.
  4. We would strive to document our project as a case study of trustworthy cultural data and AI policies (in the context of the new EU Data Governance Act and the AI Act, which will be in force in Lithuania within the timeframe of our project.) We would like to connect the experience of the Dutch AP Coalition PPP, international art+tech, cultural heritage, and music policy achievements with Lithuania’s cultural policies.

We aim to design collaborative, transparent, ethical services that Listen Local and its partners can carry out at low cost for the benefit of Lithuanian artists. Our experiments will conform to the rules of trustworthy AI and data management and can be seen as AI policy experiments. Because a microgrant from MusicAIRE supports our project, we do not intend to provide an automated service, only a prototype with manual support by a small and highly motivated team at MXF. We aim to create real-life, educational examples that help artists in real life and can be the basis of future, non-profit, affordable services in the country.

Daniel Antal
Daniel Antal
Data Scientist & Founder of the Digital Music Observatory

Founder of the Digital Music Observatory and co-founder of Reprex.