What if you could decide accuracy?
So I know some people want to be pretty much professionals at a language, while others just want to be casual and know the jist of it, and some in between. I think it would be a really cool addition to see a difficulty setting. The easiest difficulty could allow for small mistakes in spelling and capitalization, where as the hardest you need to be exact, no exceptions. I would love to see this added as it could really push some people to be precise at their sentences, or others who get annoyed when they accidentally miss one letter and it count's the entire sentence as wrong (yes I know this doesn't happen all the time.) Sorry if this has already been recommended.
Duo realy only teaches up to a basic level so anyone who has finished their tree and needs more advanced levels need to find other resources.
To implement your idea would effectively be creating two extra courses per language and would tie up a massive amount of resources to implement.
Besides, it sort of has been implemented already.
I've noticed that Duo is quite tolerant to word order in English and Russian courses, allowing for imperfect (or exotic) though grammatically acceptable variants. Capitalization is pretty much ignored in all the three courses I'm doing, and missing an umlaut in German is accepted with a "memo" asking to be more attentive next time. And finally about a couple of typos per sentence are allowed (unless you accidentally write another meaningful word with your typo, which would understandably be a mistake), allowing for more than comfortable elbow room.
As for the expert level, it is true that Duo was never meant and is not suited for that. However, discussions about the sentences sometimes go into quite subtle details of contexts, uncommon grammar features, word applications, and tints and nuances of meaning they bring into the sentence. There's a lot to pick from there for those who are interested.
Create extra courses? No, it's creating a slider button with different options for error checking. Given that it's clear from the courses I'm doing that they already treat error checking very differently, the capacity seems like it's already there somehow. It really would be nice to see it corralled and presented in an organized fashion. The most strict error checking I've encountered was in the Ukrainian reverse tree. That's definitively not where I would have picked it to be, given free choice.
I assume the OP's request largely boils down to a request to turn some of that accommodation off. The word order, however, is all dealt with by human contributors. I couldn't see that being affected by any such system.
Incidentally, some issues in Russian punctuation are so hairy even the kinds of highly-linguistically-interested speakers hanging around helping English-speakers learn Russian can't get them straight, not to mention using "incorrect" punctuation being the actual de facto standard since somebody decided not to actually include the correct punctuation marks directly on keyboards...
could you share what the error checking stringency in the reverse UK tree entailed and in which language?
Doing the English from Ukrainian tree it seemed that at least half the time a single typo (generally misplacing one letter with another, which happened a lot, owing to my unfamiliarity with the Ukrainian keyboard layout) would be enough for the whole answer to be counted wrong. I checked a number to see if these typos had, by chance, generated another valid word. That undoubtedly happened sometimes, but not that often.
On the other hand, I've definitely seen cases in other courses where it's possible to be one letter off, thereby typing a different valid word that's most certainly taught in the course, and it still not be counted wrong.
thanks for explaining. to me it would just mean a different stringency level of brute-force string matching either between languages or between different time periods (i imagine there were dozens of A/B tests for that). there is only so far that approach can go, as you say yourself. features like weird word orders, awkward structures, non-idiomatic grammar, marginal interpretations, and loose synonyms will stay out of reach. add to it the string substitution layer ("20"->"twenty"), and even a single character deviation allows one not to learn squat while still collecting XPs.
all in all, i fear without major changes in the db structure (additional flags for correctness) and without the manual labor by the volunteers this stringency setting would be just another gimmick.
It would definitely be useful if you could make it more strict. In German I find it a little odd that there's no setting to make it mark e.g. o instead of ö or u instead of ü incorrect. They also mark missing capital letters as incorrect.
It could be handy for some people to get into the habit of marking more accurately.
Some people have said in the comments that Duo doesn't cover advanced material, so accuracy is irrelevant, but accuracy can still be important to certain beginners, and there are also definitely so very advanced language learners on here just to brush up on their skills
They also mark missing capital letters as incorrect.
My impression is that this isn't the case, and that's one of the reasons the Klingon course is held up, as it happens. Am I wrong?
Definitely would love to see this feature as someone who is learning a language where the difference between a and á is crucial, I'd rather be rejected than simply advised I missed a fáda
But I'm wondering, does DL really ignore these little typos? Sure, it lets you pass the exercise, but I wouldn't be surprised that it would be factored into the strength of an individual word. What I mean is that typing 'celebre' instead of 'célèbre' in the French course might not be considered an error grave enough to let you redo the exercise, but maybe it does let the word strength decay quicker, thus letting skills decay too.
I'm not sure if the algorithm works this way though, can anyone confirm this?