My two cents : this is a job to do in conjunction with structured wiktionary, who will be able to deal with lexical entities.

We however have some properties here and there to deal with such languages issues to deal with this inside Wikidata, female form of occupation name for example, but the logic to deal with those datas is coded in the clients like the infoboxes.

Coding this inside Wikidata would still require a step that is far from reach imho : code a per language language grammatical model that would select some lexical forms considering the context ... Not that easy to do.