top of page

Music metadata - only the machines can save us

Highly recommended article in The Verge on the huge problems around metadata in our streaming-driven music world. As the headline says, metadata is the biggest little problem plaguing the music industry.

Metadata, the credits and identifying data around the main event, the actual file/song/recording/whatever, can be hard to get excited over in the abstract, but you don’t have to spend much time in a digital environment to run up against a metadata driven issue when you can’t find what you want or Alexa stubbornly refuses to play what you have in mind.

I remember having many conversations over the years with various music industry colleagues wondering why we couldn’t create a simple IMDB for music. IMDB’s power is that everyone from film fans who love curating collections to movie professionals who rely for future employment on clear and findable credits for their work is highly motivated to make the database as comprehensive as ever.

There are plenty of similar people in music of course but the music ecosystem is vastly more complex, with everything from multiple and competing rights holders and power bases of labels, music publishers, collecting societies, industry bodies and services, different combinations of collaborations and featured artists that complicate artist credits, songs written by sometimes very large teams over a period of time, subjectivities around genre and mood, and never mind a vast archive of human recordings created long before anyone had even dreamt of digital anything.

Taking all this and more means making creating a uniform, workable and scalable system a herculean task and one that no one has managed to properly crack yet.

Discogs is a very impressive database of record releases, but even a glance at a specific title can overwhelm with all the different formats and versions, editions and country releases, some with the same artwork and tracklistings, others quite different, sometimes with differences in catalogue numbers (metadata), sometimes not. You can get very lost very quickly.

Lots of talk around the growth of AI we’re seeing in commercial environments focuses on the risks machine learning has for eliminating jobs. This is a real issue, but too simplistic in my view as every technological change that is adopted widely yes reduces some areas of work but also opens up others. Since spreadsheets came along being an accounts clerk updating paper ledgers isn’t such a sought after skill any more. But the world of finance and accounts has been liberated by their existence enabling them to be much more creative and expansive at the same time (not always a good thing clearly) and it employs far more people today than ever.

In this case, creating a comprehensive and workable database of global music rights and metadata has literally proved beyond the wit of humans. Bring on the machines, the music business needs you!

bottom of page