What's In A Hit? A Computational Analysis of Topics and Features in Popular Song Lyrics

Abstract

What relationship do lyrics have with a song's commercial success? How is this dependent upon the time period in which the song was written? This project addresses these questions through more generally examining how the structure and semantics of popular song lyrics have evolved over time. The study as a whole forms a broad, statistical investigation into the nature of language in popular song lyrics, and its insights focus primarily on visualization and regression. A dataset of 13,960 songs' lyrics was scraped from the web for the purposes of this work, and three separate analytical approaches to the data are taken, each providing di erent insights and utility. The study nds no sure re predictors of commercial success within song lyrics alone, but a number of temporal linguistic phenomena are revealed.