Typical ML thinking, it seems. The author should ask the question: What does my "model" actually assign here? Instead, this very core of the work is ignored an

Typical ML thinking, it seems. The author should ask the question: What does my "model" actually assign here? Instead, this very core of the work is ignored and the focus rests completely on the numbers put out of a black box... If the model simply assigns a closer similarity to more modern words (i.e., it would evaluate older words as "weird"), we would expect exactly this outcome, no?