Skip to main content
EuroPythonCode of ConductBuy tickets

Build your own Playlist Recommender System with Python using your GDPR Data

Room:
Liffey Hall 1
Start:
14:30 on 15 July 2022
Duration:
30 minutes

Abstract

In my talk, we explore our usage data requested according to GDPR and leverage it - together with Spotify’s Web API - to build a personalized playlist recommender system with Python.

In 2018, the General Data Protection Regulation (GDPR) became effective in the EU. It sometimes causes data scientists great headaches. But from a consumer and Pythonista point of view this can also be interesting data for exploration. It is very useful for building personalization technology, in particular recommender systems. And there are almost endless ways to use Python for it. So, let’s request and use our own data to build a playlist recommender system which infers our music taste from our streaming history and uses it to retrieve songs from our favorites in a new way. We will call it “Your Rediscover Past”, a personalized playlist based on your streaming history and saved songs.

TalkPyData: Deep Learning, NLP, CV

Description

Personalized Playlist Recommendations on Spotify are great – some of them let us discover new songs, some others help us to rediscover songs. However, rediscovery seems to be limited on the more recent past, i.e. going only a month backwards. This is a problem if you like to rediscover some of your favorite songs you might have listened to a longer while ago. Sometimes we add them to our "liked songs" where they likely fade away. However, you once explicitly declared those tracks as favorites. So, what is it that we can do about this missing piece in personalized playlist recommendations?

Well, the first thing we do is to request our personal usage data from Spotify according to GDPR. Second, we analyze and enrich it with track audio features offered by Spotify’s rich Web API. We derive the music taste profile of ourselves from 12 months of streaming history and use this taste profile to retrieve favorite songs we haven’t listened to for more than a year. In my talk, I present you the Python package I build for this purpose, possible extensions and enable you to create your own personalized playlist to rediscover your past!


The speaker

Marcel Kurovski

Senior Data Scientist and Innovation Lead at inovex Host of Recsperts - Recommender Systems Experts, the Podcast Show with industry and academia experts in Recommender Systems Building Recommenders and Personalization Solutions with Python for various industries since 5+ years Creator and Instructor of Python RecSys Training



← Back to schedule