Hacker News new | ask | show | jobs
by not2b 1590 days ago
The second and third graphs aren't zero-based, which exaggerates the effect.

Also, the author may be over-interpreting the result: it's odd that very different fields are showing roughly the same change in cosine difference, and it is a measurement over individual words. I think a deeper analysis is needed to figure out in more detail what is going on.