Analysis of Word Choice in Different Genres

We begin our analysis of word selection in song lyrics by dividing the lyrics according to song genres, counting the most frequently used words, and making word cloud pictures to demonstrate the mostly commonly used words in each genre.

To control for the popularity of the songs studied, we only choose the songs with high singer familiarity > 0.5. We use singer familiarity as a proxy for song popularity, so that the sample can better reflect the population of popular songs.

The result comes out quite mixed. First, as Graph 1 indicates, the 14 genres share a lot of common words. In fact, the top 10 most commonly used words are very similar across genres, with “love, know, time, like, go” to appear in all of the top lists. Eyeballing the word cloud pictures, one might even come to the impression that rock, punk, jazz are almost identical in terms of word frequency.

Graph 1: word clouds of the original lyrics, all 14 genres

One embedded issue with this naive approach is that, words like “go, oh, come” are so commonly used in any type of English writings that they themselves do not add any additional information to distinguish the content.

Summing up the frequency of of most used words, we find that by and large, the 100 most frequently used words in the 14 genres make up ~40% of total words in the lyrics, in each genre. Therefore we try examining what the word frequency would look like if we take out the top 100 words.

So we give the word clouds another try. We knock out the 100 most frequently used words in each genre, and as shown in Graph 3, the genres now look much more diverse.

Despite the obvious similarity of word choices across genres, however, the genre of rap songs stands out as an outlier. It is the only genre whose 10 most frequently used words do not include “love”. In fact, the word “love” does not even make into the top 20, whereas curse words such as “shit” or “fuck” both rank higher than “love”.

Therefore, our next step is to take a closer look at the rap genre, to study the trend of word choice over time.