Just now I answered “to value” as the meaning for the vocab 置く (because I confused the kanji with 値), and was marked correct–I guess because my answer was close enough to “to place”? Searching the forums, I know the devs have been talking for years about giving us the ability to “blacklist” wrong answers that are otherwise close enough to be marked right (this is an especially big issue with vocab that involves numbers).
Has anyone seen an update from the devs on this (post-mid-2017)? Barring that, are there any scripts/addons that enable this functionality? I know about the “no cigar” script that requires answers to be 100% correct, but it’s really nice not to have to retype answers that have genuine innocent typos.
I don’t use it so am not the best authority, but there’s also the double-check script:
It won’t let you blacklist particular answers, which is what you really want, but it might give you a little more flexibility than the no cigar script while still being more helpful than nothing (there’s a mistake delay option if your answer has typos, for example)?
Thanks–I think this would have helped in the instance I described above (although, ironically, it would not have helped if I had made a typo in “to value,” because then I would have thought the shake/delay was a result of the typo instead of my answer being completely wrong).
If any devs are reading, it also occurs to me that a helpful fix might be to calculate Levenshtein distances for each word individually, rather than the phrase as a whole, and mark the answer wrong if any of the words are too far apart? This seems like it would be particularly helpful with words that include numbers. For example, “four days” is treated as a correct answer for the vocab that means “five days” because they are long phrases with a Levenstein distance of 3. But if you just compared “four” to “five,” that same distance of 3 would result (as it should) in the answer being marked wrong.
This wouldn’t fix all of the problems that blacklist would, but it would help a lot.
I agree! I think this would also allow most number answers to work too, like “4 small animals” since 4 is only one character long and would have to be an exact match.
You might want to email them directly with this suggestion.
I think it’s a great idea too, although my one reservation is that it would be quite annoying to get things like verbs wrong just for mistyping ‘to’ (and it may perhaps be less obvious to users how the system works, leading to more frustration). Any error on a word that short is marked wrong, and it’s the part I least care about typing accurately
Right. It wouldn’t fix the problem with longer numerals (e.g. 1000 being accepted for 10000), but another good fix might be to require all numerals to match exactly.
Yeah, that would be annoying. Maybe they could “whitelist” the word “to,” or just set the threshold for two-letter words a little higher.
I think the current system is pretty opaque to most users–I didn’t understand it until today’s incident prompted me to dig into the forums to figure it out.
That’s one reason that I always use “to verb something” and “to be verbed”, if those are given as options. I’m tempted to add them as options to verbs where they’re not there. (Like, say, 続く and 続ける, which are both defined as “to continue”, with a note from Koichi in the meaning mnemonic saying “Eh, it’s impossible to tell these apart, you’re on your own for this”. Well no, one of them is “to continue something” and the other is “to be continued”. Actually the number of times Koichi’s mnemonics involve someone else doing the verb for intransitives makes me suspect Koichi doesn’t understand transitivity at all.)
Yeah, I’d so much appreciate this to be fixed. It’s such a simple thing to do, but that it can improve student’s retention of these items so much more.
