September 29, 2018
Taking “Data Blitz” as literally as possible
I have data!
Self-promotion
Inferring meaning from words
- Dictionaries
- Count how many words fall in each category
- Lots of pressure on the dictionary!
- Other stuff?
Distributed dictionary representation (DDR)
- Dehghani and colleagues (2017)
- Representing words as vectors
- Can help account for words missing from dictionary w/ similar meaning
My data
- Billboard charts since 1993
- Retrieved lyrics for around 70% overall
- More like 90% in recent years
- About 15,000 songs w/ lyrics
- Focusing on each chartās average
- Interest in music in general, genres comparatively
Pop over time
Rock over time
Country over time
Rap over time
Pop and politics
Rock and politics
Country and politics
Rap and politics
Politics vs.Ā morality
Politics vs.Ā morality
Genre comparisons
Genre comparisons
Genre comparisons
Genre comparisons
Genre comparisons
Many more options
I haveā¦
- Raw word counts (dictionary method)
- More genres, times
- If Billboard charted it, I can get it
- Audio features
- Key, tempo, duration, many others
- thanks to Spotify API