Hi @Lucho_BG ,
great to see you around in the Forum! 
Differences in ways of speaking are actually being handled already in a nice way in many of our projects; this can be a certain accent which needs to be transcribed correctly by the Speech Recognition layer, or as you mention the usage of synonyms/slang etc. The latter is then a task for the NLU layer (Intent Recognition), and the hybrid approach described in the article can perform nicely here.
I see challenges more in the game design perspective, one thing is to understand everything the user says, another thing is to have a good answer/action/plot for it 
I think that’s why the first step is to analyze case by case (or game by game) what usage would make sense.
Thanks for your comment!
/Benjamin