One of my favorite parts of BunPro is, without a doubt, it’s “Dynamic Reviews” – the fact that you anticipate incorrect answers (or those technically correct that you’re not looking for) and which you comment on with these orange hints.
Broadly speaking, they fall into two categories, those that result in getting another try (typical example “more formal”), and those that are marked as incorrect, but with a hint what went wrong (“you need a な here”).
My crude estimation is that these hints tend to slightly differ from answer to answer, which suggests to me that some poor guy has to enter all these incorrect answers, and then enter an appropiate hint. I think that’s great. You can’t beat that.
But I think you could get some mileage out of an automated system for certain subsets of these hints.
Consider the following: A great many grammar points are created by using て-form and then adding something to the end. For example, this one, this one and this one.
It’s possible for a user to get an answer wrong, but still kind-of getting the grammar point right, by messing up the て-form. For example, in this question:
これを_____ (expecting これをみてもいい)
A user might enter:
これをみってもいい
This would be marked as incorrect, and the grammar point V[て]もいい 's SRS level would fall.
But I’d argue that while the answer was incorrect, it’s not because they did not understand the V[て]もいい-grammar point, but rather the V[て]-form!
These cases could be handled, by BunPro checking whether the end of the input string (that was incorrect) matches the end of the correct answer, and if it includes a sign that the て-form was attempted (て or で included in the first part) . If it does, then depending on the grammar point, you may want to tell the user a hint such as
"
Close! Incorrect V[て]-form!"
(with the little marking that it’s an automated hint)
Whether or not to then lower the V[て]-form’s SRS level instead of the V[て]もいい’s SRS level, I leave to you.
Of course, I focused very much on this example, but I think it generally could be done for many other things – verb-forms just seem to fall nicely into this. I do think, however, that you need to decide by hand whether a grammar point should use such a system or not.
Doing so would enhance BunPro’s greatest feature (in my opinion), by catching even more close incorrect replies, giving even more focused feedback.
And I think if implemented in the way described, it should not overwhelm you with millions of additional cases to check.
What do you think?