6

Option to retry a kanji on bad input

Sometimes, I get a kanji wrong because i accidentally tap the screen somewhere before i even start drawing, etc. In cases like that, it would be good if I could:

 - mark as a flubbed input

 - retry without recording a success or failure, and without revealing the kanji (as this would tell me the answer)

Currently, the only option when this happens is to mark the answer as incorrect, as the fact that i can see the kanji and didn’t draw it means i can’t just assume i’d have gotten it right.

3 replies

K

I was thinking about this the other day and came up with what I thought might be sort of happy in-between:

when the user messes up a stroke show them only:

1. the correct strokes they've made so far

2. the incorrect stroke they drew (in the same way ringotan already does)

3. the correct version of the stroke they should have drawn

(but NOT the rest of the correct strokes)

then present the options to either

1. override the autograder ("my stroke was right") and resume entering strokes (if there are any - otherwise  advance to the results)

2. accept the mistake and advance to the results (unless this was the last stroke in which case maybe just jump to the next question?)

and then the results screen is the same (shows the whole kanji and colored based on whether the answer was right) as usual sans the option to override the grader and maybe with some sort of highlight (red?) for the strokes that were missing (including the one that was wrong).

overall, i don't think this harms the UX at all because regardless of whether you legitimately messed up or not* every screen serves a meaningful purpose and I'm hoping it's not a significant change to the implementation (the bar at the bottom just becomes a two-step process).

* Note: At least for me because I try to use the strict grading mode and because I'm often practicing on a moving bus this is a fairly common occurrence

Also, not really a huge reason by itself but since this approach would record exactly which strokes the autograder is wrong on I think that this data could potentially be used to improve it.

K

tl;dr; since mistakes occur on a stroke level (not a kanji level) don't spoil the whole kanji on a stroke error and offer the option to "override and continue" rather than to "retry"

S

I'd personally love an option (not by default, this would slow many people down) to have one "flub" do something like bring up a "nope" screen that requires a reasonably small button hit to dismiss, without showing the answer.  I do my practice standing up on the subway a lot, one-handed, and it's easy to bonk the screen badly, especially at the start of drawing, or to smear a line too long/short.

Good suggestion!

I don't want to not immediately show the kanji on an incorrect answer on the off-chance there was a slip-up, because that would dramatically worsen the user experience for a fairly rare occurrence.

I could however add "retry" as an option.  I think in order to not screw up the SRS, since you've already seen the answer and thus have it as a recently recalled memory, I'd have to ignore any SRS improvement (ie. if the last time it was shown was 2 days ago, it will be shown again in 2 days if you get it right).

What do you think of this?

E

That would be good I think, yes. It’s mostly just annoying to lose SRS progress in those cases

HB

I think it's a good option. I think your compromise is good to ignore SRS improvements is best. For what ever reason on normal leniency long horizontal/vertical strokes (that cross more than one quadrant) are judged very strict but not short ones (sometimes trying to fail a Kanji with a randkm stroke takes several tries). At least every other day i get a Kanji wrong (especially the vertical stroke in 木 as a component) because i make the stroke with my finger a little short). I'd rather have the option to retry as implementing a different algorithm for stroke matching may not be as feasible to implement.