Remember the time where we needed to download music on our computers and transfer those songs to our mp3, not being able to listen to more than 20 songs? Times have changed since then, and almost everyone now uses the app Spotify. Hundreds of millions of people around the world use Spotify to listen to their music. With over 50 million songs and podcast episodes, it is not surprisingly beating the mp3 (;
But Spotify is doing more than just giving people access to podcasts and artists and their albums; Spotify is using technology to give their users an exceptional personal experience.
For example, they brought in Discover weekly, where every monday players receive a new playlist with 50 tracks. This playlist is based on songs they like and recently listened to, but haven’t heard before. How does Spotify do this? They use a form of machine learning. One of the used techniques is Collaborative Filtering, where an algorithm compares the songs you’ve listened to with other user-created playlists with similar songs. Another technique that uses a similar algorithm, but in a different way is Natural Language Processing (NLP). NLP is the ability of an algorithm to search through the web to find music related articles and blog posts. This way, the algorithm can match songs based on the way they are being discussed on the internet and new songs can be added to the discover weekly list. Additionally, they use Convolutional Neural Networks (CNN) to make sure also less-popular songs are considered for the playlist. With CNN Spotify matches songs based on their attributes (e.g. beats per minute, loudness).
The company also just released a new feature on the 9th of September called ‘Enhance’. With this feature Spotify adds recommended tracks to your own playlist based on the already existing tracks using similar algorithms as described above.
All these techniques make it easier for us to discover new artists and tracks more than ever. Algorithms track what we like, then give us what they think we like. This also made me think how hard it has become to discover new types of music that differs from what we already know. I still have a record player in my room and go to record stores every once in a while to step out of this filer bubble and wander into fresh territory. How about you?
References
E. (2021, January 15). On Netflix and Spotify, algorithms hold the power. But there’s a way to get it back. Experience Magazine. https://expmag.com/2019/11/endless-loops-of-like-the-future-of-algorithmic-entertainment/
How Spotify Uses Artificial Intelligence, Big Data, and Machine Learning. (2021). Data Science Central. https://www.datasciencecentral.com/profiles/blogs/6448529:BlogPost:1041799
Tambekar, A. (2020, May 11). How Spotify Uses Machine Learning Models to Recommend You The Music You Like. GreatLearning Blog: Free Resources What Matters to Shape Your Career! https://www.mygreatlearning.com/blog/3-machine-learning-models-spotify-uses-to-recommend-music-youll-like/#:%7E:text=Convolutional%20Neural%20Networks&text=Each%20song%20is%20converted%20into%20a%20raw%20audio%20file%20as%20a%20waveform.&text=With%20these%20key%20machine%20learning,would%20have%20never%20found%20otherwise.