There's going to be diminishing returns in splitting the languages where you get less information related to the region / concept just because you're avoiding mixing languages. The language was not the only aspect: "cultural background, and in-depth regional knowledge". There's going to be lots of information shared in south/North languages just because of the geographically close (relatively anyway) distance.
I mean you wouldn't want to split a model into 3 separate ones, where one contains Austrian, another Slovakian, and another Hungarian, since there's going to be lots of cultural overlap.
I mean you wouldn't want to split a model into 3 separate ones, where one contains Austrian, another Slovakian, and another Hungarian, since there's going to be lots of cultural overlap.