Perhaps if entire corpus of 1920s literature and obscure licensing rules are trained into the model it might.

I can certainly try though.