The Data Sources
To perform researches of this nature, scientists need loads of data, and these group of scholars knew that. They fetched data from three huge public sources; two of which weren’t initially created for this purpose. Around 90,000 songs were downloaded from Ultimate Guitar, a website where people upload their own musical transcriptions.
To associate some sort of emotions to the lyrics of the song, the scientists turned to labMT, which is a crowdfunded website which gives “emotional valence” ratings to words, which basically means the extent to which a word is good or bad. Lastly, data related to the origin of the songs was taken from Gracenote.