The modern ability to stream music using services such as Spotify, Pandora, and Apple music has revolutionized how it is consumed. There is now more music accessible at our fingertips than ever before in history. As someone who has played an instrument their whole life and with some music theory knowledge, I always strive to better understand what I’m listening to. This motivated me to dive into finding out how music is quantified, specifically on Spotify, and what makes up people’s musical taste. I will be approaching music tastes from 3 angles: Spotify, music critics, and my own analysis and knowledge. To do so I am using a dataset that includes the top 100 most popular songs from 2018, off the playlist made by Spotify. This represents what Spotify and the public consider to be the best of the year. I will be comparing this to two separate top 10 lists of the ’10 worst songs of 2018’ to get the critics perspective. While some of the songs were shared between both lists, of the 16 I use, interestingly 6 of them are also in the top 100 playlist. In order to attempt to understand this anomaly, I will be comparing songs based off the variables Spotify uses to quantify their music (I get into the specifics below) and analyzing the clustering and outliers I see when graphing them. I also hope to bring some of my music theory knowledge, as well as a more contextual approach into understanding why popular music can be simultaneously very successful, while also being considered poor quality.
Amsterdam, Noah, "Analyzing popular music using Spotify’s Machine Learning Audio Features" (2019). IPHS 300: Artificial Intelligence for the Humanities: Text, Image, and Sound. Paper 14.