Incubator boundaries of algorithm vs bilingual input.
Hi I'm curious as to the boundaries of the duolingo algorithms vs moderator input.
If I translate a French sentence "Elle suis une fille", duolingo tells me I got suis wrong and WHY! I can't imagine that if moderators are merely empirically inputting sentences, the system can know why an answer is correct.
This particular one may be easy because once the word is identified, and the wrong one is used from a group, we can match it, but that is a programmatic rule. There are more complicated rules that can be coded.. maybe it will tell me that I should be saying "c'est une fille" given some context...
Can moderators add programmatic rules to the incubators? Or is the modus operandi to merely provide more binary correct/incorrect solutions and leave the explanations for success/failure to the discussions pages.
The same kind of questions relate to the coding of the synthesized voice, although if it is merely being outsourced, I imagine that such software is supposed to be able to read standard sentences fairly well (although it should be optimisable when it has got it wrong).
I.e. Will the incubators always be a poor relative of the official languages supported? Or is it structured in such a way that such optimizations can be done on top of the work done by the bilinguals?