Affect related Vocab when Kanji is wrong


#1

Hello,

I have realized while doing my reviews that if I get a Kanji wrong, I get to see the answer, and then when related Vocab comes up, because I have just seen the Kanji’s reading, I have a good chance of getting the Vocab right. This invalidates the SRS progress. Now, I could deliberately fail the Vocab, but seems to me that this shouldn’t even be necessary.

So here’s an idea:
If Kanji reading is wrong, reset the countdown for related Vocab review (could do this for only Vocab which use the same reading as learnt with the Kanji, or for all related Vocab.

Example:

  1. Kanji is due, On’yomi is expected response.
  2. User inputs wrong answer and fails Kanji.
  3. All Vocab which incorporates the Kanji has the countdown reset (I know it’s not actually implemented as a countdown to the next review, but you know what I mean…). Alternatively, Vocab which uses the on’yomi reading of the Kanji has their counter reset (as on’yomi is what the user failed in 2.).

Right now, it feels wrong to get vocab right when the Kanji is failed.

Thoughts?

(Yes, there are drawbacks to this, I’m sure, but perhaps as a feature which can be turned on or off if the opinion is split, then.)


#2

Not every related vocab is going to come up soon after a kanji is reviewed. Where would you propose to draw the line on resetting the related vocab?


#3

Imagine that the next review date is implemented as countdown. two options spring to mind:

a) If a Vocab has a remaining time till next review of 4 days, reset it to whatever the number of days it would have required if it had just leveled up to its current level.

b) If a Vocab is, for example SRS 7, it would need 1 month to move up to 8 (that’s how the SRS is currently set up). Imagine that it’s been SRS 7 for three weeks. It therefore needs another week to be up for a review to go up to SRS 8. Treat the three weeks it accumulated at SRS 7 as “credit”. The system could auto-fail the item (so level it down from SRS 7 to SRS 5 as if the review had just happened) and then move the item up to SRS 6 because of the three weeks of credit that were accumulated at SRS 7 (by using two of those weeks), and then add the last week of credit towards the review date to SRS 7. I can imagine that my explanation is confusing, but this may be a more accurate way of dealing with this as it deals more in time than SRS levels which are just proxies.


#4

I think this is a very interesting point. However for me personally, vocabulary reading and individual kanji reading is not always related in this way. Personally I have kanji where if I see the individual kanji on it’s own outside of ANY context, it is hard for me to come up with the reading while when inside a word with another kanji I know it immediately. This has however not happened in the context of WaniKani specifically so we’ll see how things go for me during later levels.

I think setting a rule/algorithm for this issue would end up in a lot of faulty assessments. I would do this manually myself if I realised that “wow I would not have known this if it wasn’t for just looking that up”.


#5

This starts to get complicated the more you consider it though.

What if you get a vocab item wrong that uses the same reading as the kanji - and the kanji is also in your current review queue. Should the kanji be downgraded since you might get it right after having seen that vocab word?

I like the idea, but I get the feeling that there are further implementation details that will come to light and make it difficult to handle.


#6

There are definitely design considerations to evaluate, no doubt.
The particular concern you bring up would be solved by partially ordering reviews such that vocab cannot appear before its Kanji has. And to answer your question about whether the Kanji should be downgraded in the case you describe, I would go with “yes” (this also means no ordering of reviews is necessary).


#7

It’s outside of the scope of WK’s mechanics, but there is merit in the idea of separating reviews-for-advancement from reviews-for-practice. It gives you a lot more flexibility in implementing interleaving, and somewhat lessens the importance of passing or failing a review. You can address the learner’s weak spots while generally continuing to move forward in other areas.


#8

Apologies, but would you mind expounding on that, please? I fear I do not completely understand…


#9

Ahh, sorry :slight_smile:

On WK, the reviews are the only form of built-in study. All other studying is left to the user’s initiative, such as self-quizzing, anki decks, reading, etc.

If they had some form of practice built in, you could be given an increase in exposure to related vocab words to help reinforce a kanji that you got wrong, without affecting the progress of those vocab items.

And with built-in practice, it could coordinate interleaved study better by pairing up related items in different ways, rather than it being so heavily dependent on fluke timing of a review schedule.


#10

If you’re getting a word right due to seeing the kanji for it earlier, then practicing that kanji should also reinforce any weaknesses for that vocab. If it doesn’t, then the SRS system should take care of any words that you got a free pass on. Unless you got that free pass on the word during the burning phase, which would be quite a coincidence.

So I don’t see this being a super huge problem.


#11

I don’t think it makes sense. There are multiple ways of reading a kanji, and you can get the vocab right and get the kanji wrong. It doesn’t mean you didn’t learn the vocab.
To me they are two completely different things. It’s like saying we can’t learn any word using the letter A just because one time we spelt A as E.


#12

If you want to talk about improving WKs system, consider this:
sometimes I confuse kanji - especially when their meanings are closely related.
so if I type in the wrong translation for one kanji, not only should this kanji fall back in the SRS levels, but the kanji I confused it with should also be reset (because I associated the meaning wrongly). This could even be extended to the readings, but that could be impractical. (I don’t want every しょう kanji be reset when I type in shou for a kanji with a different reading)


#13

I have to say I would be super against this - unless it were entirely optional, of course.

There are many cases where I know a vocab word just fine, but get stuck on either the meaning or reading of one of its constituent kanji when said kanji is reviewed independently. I would not accept to being set back on a vocab I knew because I got one of the kanji wrong. Particularly if they didn’t even have the same reading, which is often the case.

I do understand the desire to feel like you’re not “cheating the system”, but I also don’t think getting draconian with penalties for being incorrect is great either. The SRS will eventually catch up and correct any undeserved progress. Maybe a 'I shouldn’t have really gotten that…" will get you through Apprentice or Guru, but if you didn’t truly know it, you’re eventually gonna get knocked back when it comes up for Master, Enlightened, or Burned.

So I think that, over enough time, the SRS is already smoothing out a lot of the little quirks that might allow someone to progress slightly ahead of their actual knowledge.


#14

Pretty sure this is referring to things like… the kanji 石 (which takes the kunyomi reading いし) comes up, and you get the reading wrong. Then, the next day, the vocab 石 comes up, and you get it right. Is that really an accurate representation of your knowledge of the vocab? Had the vocab come up first, you probably would have gotten it wrong. All that happened was you had a one day SRS gap, but the system treated it like you remembered it since the previous correct response, maybe months ago.


#15

I think the reason some of the scenarios outlined here don’t worry me from a SRS standpoint is that, practically speaking, if we’re studying kanji here we’re probably also coming across kanji “in the wild” as well. Maybe that’s just casual exposure via hobbies, maybe it’s via Anki or any other learning methods used alongside WK, etc.

Getting a review correct in whole or in part because of a related review you recently got wrong and learned from isn’t, in my eyes, really any different from forgetting a reading, coming across it outside of WK, being prompted by that encounter to remember it, and then getting it right on your next WK review, when maybe you wouldn’t have otherwise.

So while I do like the SRS concept and by and large want to let it do its thing, the “purity” of my SRS experience isn’t paramount to me. Literally anything that helps me remember a word seems valid. And if I’m truly not retaining it, I’m going to get it wrong in a review eventually. It’d take an improbable series of coincidences to luck my way through a kanji/vocab all the way through Burned. (And I could just unburn if it did happen.)