~/projects/general-conference/compared-with-byu/topics
Did the two pulpits drift together?
One thirty-topic model fitted to 7,288 conference talks and BYU speeches since 1950, so both can be read on the same scale. The distance between their topic mixes went from 0.2589 in the 1970s to 0.3628 in the 2020s: the two pulpits drew apart. BYU leans toward “education · academic · learning” and “don’t · i’m · didn’t”; conference toward “conference · express · thank” and “holy ghost · ghost · revelation”.
Method
The same pipeline as each hub's topic study (masked TF-IDF, thirty-component NMF), fitted once on both corpora from 1950 with campus words (BYU, university, students, faculty) removed so that the model cannot separate the corpora by setting alone. Each decade's mix is the mean normalised topic weight over that decade's talks in each corpus (twenty or more); the distance is the Jensen–Shannon distance between the two mixes, base two, so 0 is identical and 1 is disjoint.