Lyrics Project: How to Measure Cultural Change

Tagged with: , ,
Posted in Projects, Research, Tools Research
Lyrics Profile Image

Image via CiteLab

The Lyrics Project comes out of McGill’s CiteLab, a group formed by Stéfan Sinclair (Associate Professor, Digital Humanities), Andrew Piper (Associate Professor, German Studies), and Mark Algee-Hewitt (Mellon Postdoctoral Fellow until 2012, currently an Associate Director for Research with the Stanford Literary Lab).

The project looks at twentieth-century song titles and lyrics in order to measure vast cultural change by its lexical changes. Inspired by the journalistic trend claiming that culture is changing more slowly than it ever used to, and that the changes are more replacements—think Lady Gaga for Madonna—than real innovation (Vanity Fair), the Lyrics Project uses DH tools to try to measure the huge scope of this claim. What better than song titles and lyrics when you are looking for the storehouse of an era’s cultural language?

There are several approaches to quantifying the lexicon and its changes. One initial search of titles (using R to cluster data) reflects some really interesting cultural changes across the century: ‘blue’ and ‘blues’ rising in the 20s, ‘baby’ and ‘woman’ popular in the 60s and early 70s, and currently, ‘dance,’ ‘forever,’ ‘tonight.’

To measure the variation and ‘innovation’ in year by year titles, another search looks at unique words. The higher the proportion of unique words to repeating words (the type-token ratio), the more lexically innovative that year could be considered.

With each method of searching comes questions for further research: how different (or similar) are one year’s unique words to those of another? Some years may be equally ‘innovative,’ and therefore similar in the rate of change, but are the words themselves undergoing significant change? And how can we measure innovation—what about sampling and song covers?

Distance Tables created in R, which place near to one another the years that share words at a similar rate of recurrence, show that the title words became steadily less different over the first half of the century (excluding the post-war period) and the rate of change remained consistent from the late 60s on. From the insights provided by the table, it is not necessarily that culture is no longer changing, but that the time-frames of change are growing: the twenty year interval shows similar amounts of change as the 1-5 year interval.

Lyrics R graph

Image via CiteLab

For further reading, please visit the CiteLab blog: http://digihum.mcgill.ca/citelab/category/lyrics/