Du måste aktivera javascript för att sverigesradio.se ska fungera korrekt och för att kunna lyssna på ljud. Har du problem med vår sajt så finns hjälp på http://kundo.se/org/sverigesradio/

Meet our audio fingerprinting tracklist robot: Algorytm

Publicerat torsdag 20 april kl 13.37
Algorytm-driven tracklist for Musikguiden i P3
2 av 2
Algorytm-driven tracklist for Musikguiden i P3

– It's allowing our radio hosts to focus on making great content and offering great tracklists to our music hungry listeners at the same time.

The Swedish Radio is one of Swedens largest music distributors and we have hundreds of thousands of music listeners that tune in to their favorite radio shows every day.

We're always committed to providing great digital services to our users and we're continuously adding new features and adressing issues in all of our digital platforms every day. But every once in a while, we decide to nice it up even more by letting the development teams be creative and work on something out of the ordinary.

Meet Algorytm, our audio fingerprinting tracklist robot. 

"Who is Algorytm, and why do you spell it like that?", you ask?

It is the result of a week long hackathon involving many of our cross functional development teams and it is responsible for the high quality tracklists we have today.

It's 100% autonomous and greatly reduces workload on our radio hosts.
The reason it's spelled "Algorytm" and not "Algorhythm" is because we like Swedish wordplay. In fact, this is the third system that gets a humorous name like this.

So how did it all start?

Well, for a while we had been feeling that the music tracklists that we provide our listeners had been somewhat inconsistent and varying in quality. Not for a second did we accuse our radio hosts of poor work, instead we suspected that the reason was the troublesome manual endeavor of track reporting and how error-prone it was. And while we have some of the best radio hosts in the world, the thing they're mostly committed to is not (surprisingly) drudging time consuming track reporting - it's creating great shows.

Algorytm makes use of a software-as-a-service provider called ACRCloud that uses digital audio fingerprinting, a technology that can analyze an audio stream to identify music known to the service. The service can then provide ISRC's (so far only American ISRC's), Spotify & YouTube id's and more for each identified track.

After setting the fingerprinting service up to listen to all of the web channels that we wanted to monitor, we started to get track hits that were pushed to a secure Elasticsearch cluster. The hits are then preprocessed by a microservice before it ships them off to Isidor, our content management system, where it is placed in the correct tracklist.
Since our channels broadcast 24/7 and multiple shows can be broadcasting in different channels at the same time, there was some tedious synchronizing logic that needed to be built here.

The result is reliable track listing without any manual effort.
Not only do we include the track artist and track name, but thanks to this particular audio fingerprinting service provider, we can also collect album cover art and create clickable Spotify and YouTube links, allowing the user to further indulge in their favorite music.

We believe the end result is a success.

We needed a few iterations to get it right, but thanks to good feedback from our listeners and from within the company, we feel like we did and now the radio hosts are thanking us.

Thanks to Algorytm our radio hosts can now focus on making great content instead of drudging manual reporting and we can offer these amazing tracklists to our music hungry listeners at the same time.

Grunden i vår journalistik är trovärdighet och opartiskhet. Sveriges Radio är oberoende i förhållande till politiska, religiösa, ekonomiska, offentliga och privata särintressen.
Har du frågor eller förslag gällande våra webbtjänster?

Kontakta gärna Sveriges Radios supportforum där vi besvarar dina frågor vardagar kl. 9-17.

Du hittar dina sparade avsnitt i menyn under "Min Lista".