I heard an interview on CBC Radio’s Spark this weekend; the subject… Rap Metrics™. That’s right, the statistical analysis of rap music… now there’s one for the books!
Liban Ali Yusuf, a chemical engineering student at the University of Waterloo, has explored “the idea of statistically analyzing music lyrics and gaining insight into music with numbers.”
He’d be following the discourse in sports (think Moneyball and SabreMetrics) and the fact that statistics are being used frequently to evaluate players and describe the game. Yet, he found no such objective measures existed in the world of music; people’s opinions were rarely backed up with facts.
His algorithms include measuring such concepts as rhyme density, line length variability and syllables per word – and something that looks at how narcissistic the artist is, the “I to you ratio”. He examines a variety of hip hop and rap artists through this lens. Unlike the batting averages of baseball players, these statistics are not used so much to rate artists, as to understand the different genres.
While I suspect the exercise is more a labour of love – and perhaps a fun way to get through his university thesis – there are real-world applications of his work. Yusuf cites two rappers, MF Doom and Notorious B.I.G., with very different styles. He examined the Rhyme Density™ [(Total number of syllables that are a part of a rhyme) ÷ (Total syllables)] in songs by these two artists; MF Doom was found to have one the highest rhyme densities of all the artists he studied. Notorious B.I.G. was far less “efficient” in his use of rhymes – “an efficient rapper uses all parts of a line in his rhymes”.
Notorious B.I.G. however has had far more commercial success than MF Doom. He suggests that record producers can look at measures like these to determine whether young up-and-coming artists were more similar in style to one of these artists or the other? Perhaps they’ll be more willing to throw more marketing money at the act that they feel has a greater chance of being more commercially viable.
These statistics could also be used (and may already be) to power recommendation engines. If you like MF Doom you are perhaps more likely to like Eminem (also has high Rhyme Density™) than Notorious B.I.G.
This is really just a very specialized form of text mining – customized to the idiosyncrasies of music, rather than the call centre or social media space.
And naturally it was only a matter of time before ‘analyzing rap lyrics’ turned into using analysis to create them; on his blog he refers to his Drag-Drop rhymer. Using the tool you can create ‘rhymes’ that leverage the work he has done. I wonder when we’ll see the first commercial success generated by an algorithm. For the sake of art, I hope it’s not for a long time!!
More on the topic:
The interview: http://www.cbc.ca/spark/2011/02/spark-138-february-20-23-2011/
More on Rap Metrics™:
Other examples of music lyric analysis: