Data Science and Music

  YOU HAVE NOW ENTERED THE DATA DIMENSION. WE HOPE YOU ENJOY YOUR STAY.

Data Science and Music

The pandemic has without a doubt affected all of us in some way or form. But one question that I’ve had is whether it’s translated to the music we listen to as well. Did the pandemic affect the music that we listened to?

During the summer, I’ve been working with an MIT student on a data science project. The project in particular poses the question “Did the pandemic affect the types of music that we listened to?”. This question allows me to bridge together my love for music and data science.

For this project, I plan on using a code that looks at certain lyrics of the biggest songs, and based on that analyzes whether a song is more happy or sad. There are already algorithms that do something similar. Spotify is able to take a song’s audio features and analyzes those to recommend similar songs to a user. From a data science article that I’ve read, Spotify looks at a song’s acousticness, danceability, energy, instrumentalness, liveness, loudness, speechiness, tempo, and valence, and then uses those to create a playlist of positive songs or sad songs.

• Acousticness: A song’s acousticness indicates whether a song is more acoustic or not.
• Danceability: Danceability measures how suitable the song is for dancing.
• Energy: Energy measures intensity and activity, and more energetic tracks generally feel louder, faster, and noisier.
• Instrumentalness: Instrumentalness predicts if a song has no vocals, and “ooh aah” vocals and other adlibs generally count as instrumental.
• Speechiness: On the flipside, speechiness measures how often vocals are used in the track.
• Liveness: Liveness predicts if the song was recorded in a live setting or if it perceives an audience in the song.
• Loudness: Loudness, as the name suggests, measures how loud the song is.
• Valence: Valence is how fun or positive the song is
• Tempo: Finally, tempo estimates the BPM (beats per minute) of a track.

Admittedly, there are a few flaws: songs can have sad lyrics and still be upbeat, which can create major discrepancies. A song that scores high on tempo or energy may result in a high valence score, but it doesn’t take the lyrics into account. Thus, something else that has to be considered is the sentimentality of a song, which is the attitude (ex: happy or sad) towards the track.

This project is still in the works, but in the meantime, what do I think the overall effect was? Well, simply looking at the Hot 100 year-end list for 2020, there are varied results. Songs mainly were gaining success thanks to the app TikTok and some songs had references directly to the pandemic in them. Overall, I’d say that the songs were overall slightly sadder compared to normal. Songs like “Supalonely” by BENEE, “death bed” by Powfu, “If The World Was Ending” by JP Saxe and Julia Michaels, and “stuck with u” by Ariana Grande and Justin Bieber all have a more sad sentiment to them and became hits mainly due to the COVID-19 pandemic. That said, there are more than a few outliers to this. For example, “Sunday Best” by Surfaces was big during the pandemic, and it’s quite possibly the most positive song to be created since Justin Timberlake’s “Can’t Stop The Feeling”. Additionally, the biggest song of 2020, “Blinding Lights” by The Weeknd, is a very energetic song and isn’t very sad at all.

So ultimately, there really isn’t that much of a definite answer to this question yet. It’s as if while the world around us went crazy in 2020, the music scene did as well, and I can’t really say the year-end list for 2020 is definitively more happy than sad or vice versa. That said, this is a project that I’m excited about doing, and I’m eager to see if there is a strong correlation between the pandemic and the music itself.

YOU ARE NOW EXITING THE DATA DIMENSION. SEE YOU SOON!