But when someone starts using language less than fluently, it starts to fumble. The funny thing is: there are typical error sets within one or another second/foreign language learner population: common errors to speakers with this or that first language. Language teachers abroad learn to parse the common (and less common but still semi-standard) errors for the group among whom they’re immersed, when one dominant group exists. It strikes me this is something people do, but software probably doesn’t do… yet.