[yorwba]

> Yes, precisely that's what I would expect from an NLP system

> I'm still interested in a possible technique which could lead to this type of bias without it being explicit (or requiring google to have an extremely naive approach).

I don't understand. If it's what you expect, then what's left to explain?