I extracted these data using Spotify API and R package “spotifyr”. I extracted the audio features of 2017 and 2018 Spotify Top 100 playlists, and my 2017 and 2018 Top 100 playlists generated by Spotify. Then I combined them into two sets of data: Top Songs in 2017 & 2018 and Boon’s Top Songs in 2017 & 2018. I did a simple analysis to compare my taste to the general Spotify users’ tastes. I used R and Tableau to generate the diagrams and do statistical analyses.
The audio features explanations can be found here: https://developer.spotify.com/documentation/web-api/reference/tracks/get-audio-features/
1. Checking if there is any significant difference between the audio features of my 2017 & 2018 Top 100 playlists, and the 2017 & 2018 Spotify Top 100 playlists.
I did t-tests to see the features that are significant differences between these two datasets. I found out that energy, instrumentalness, valence, and tempo values have similar means. My playlists have higher average acousticness, liveness, and duration. The 2 years' top 100 playlists have higher average danceability, loudness, and speechiness.
In general, my playlists are more acoustic, less danceable, lower decibels, have less speech, have more live versions of songs, and have longer song lengths.
2. Comparing the patterns of these two datasets.
A comparison of averages is not very useful and does not provide a lot of insights to understand my music taste. I did histograms of these datasets to see the spread of the individual values of these elements. I only included those I think are interesting.
a) Energy
I listen to a good range of songs with different energy values. I have some slow songs (values close to 0) and some really fast energetic songs. Most of the songs have an energy value of 0.6 or higher. There are 2 peaks in my histogram, compared to Top Songs’ histogram with a more normal shape.
b) Danceability
My histogram is more left-“skewed”, relative to the Top Songs’ histogram, and has more songs with low danceability, but the minimum value is still 0.23. Regardless of the energy of my slower songs, they are still somewhat danceable.
c) Valence
This diagram just shows that both playlists have a wide range of valence values. There is a mix of happy songs and sad songs in these playlists. In general, we like both positive and negative songs. My playlist however does have more songs with valence values higher than 0.75. I prefer more positive songs to keep my positive mental attitude.
d) Speechiness
I have more than 80 songs with speechiness value less than 0.05. My highest speechiness value is only 0.33, while the Top Songs playlists’ highest value is 0.53. The Top Songs playlists have wider spread and more high speechiness value.
3. Understanding the audio features of songs of different languages.
I wanted to see if songs of different languages in my playlists have certain audio features that stand out. I used average values in these analyses, and these graphs are generated using data from my playlists. So, they are inherently biased.
1. 58% of my playlists are English songs; Chinese songs and Korean songs are 40%, and the remaining are Japanese and Hispanic songs. N/A is just classical music.
Not too surprising. Most songs on Spotify are English.
2. I wanted to see if high danceability means high valence. But it seems like at least in my playlists, this is not the case. Hispanic songs on my playlists are the most danceable songs. But Korean songs are the most positive songs (high valence means cheerful, positive, euphoric). Chinese songs are the more mellow ones.
Honestly, almost all Chinese songs I listen to are sad love songs. And Korean songs are K-pop on my playlists, and they are mainly bubblegum, cheery kind. They are my cheer up songs and Twice's Cheer Up was my most played song in 2017.
3. I also wanted to see if danceability equates to energy of the song. Apparently, no. Japanese songs have the highest energy, but they are not very danceable.
An explanation for this: my Japanese songs are basically anime songs, and they are great for workouts. Imagine listening to Attack on Titan's opening when you run. That's my daily life.
4. Does high energy equate to high tempo? At least for the Japanese anime openings I listen to, yes. Chinese songs are pretty weird somehow: they have high tempo but the lowest energy. There could be some outliers.
5. Chinese songs have the longest duration with the lowest speechiness.
This means most of the time, there is just instrumental for Chinese songs. Pretty interesting. I have never noticed that. Korean songs surprisingly have the highest speechiness despite being the shortest in terms of duration. Do Kpop stars just sing throughout the entire song?
Conclusion:
My music taste is not that far off from the general Spotify users' tastes. I like more acousticness and lower loudness in my songs. Japanese songs are great for workouts. Hispanic songs on my playlists are highly danceable. Chinese songs are the saddest; Korean songs keep me happy.
Comentários