I just got 旅行 as りょうこう instead of the correct version りょこう… so yeah it’ll keep happening but much less once you understand 2 things:
The WK team uses different people/mnemonics for the long and short version (with long I mean the one with an extra う ).
For example Jourm = じょう whereas Jo = じょ and I have clear in my mind the difference between the big hagrid looking Jourm and the delicate Jo. So as long as you get to the reading through the kanji mnemonic the probability of making a mistake should be low.
[The only exception I’ve found is on the current level where the mnemonic for す is Sue (in the previous levels Sue was associated with すう), but I think they just overlooked it i.e. it was a mistake.]
You can hear the difference between the short and long versions so, when you listen to Kenichi and Kyoko’ s audios, pay extra attention to the length of the vowel. Long ones are clearly longer than the rest of the word vowels (unless these are long as well). Replay until you can clearly hear the difference.
Also good luck learning all the ー in katakana only words, last time I tried I gave up…
I do but… it just ends up going back to Apprentice anyways, so I think it averages out for me. I have lots of items that drop back down… but I’m glad to have a week away from them because I’m either ready to advance them to Guru 2 or I’m ready to study them again and I’ve had the opportunity to devote my attention to other items in the meantime.
They’re usually pretty good at this, but I’ve already noticed a few exceptions. For example 女 and 同じ both use the mnemonic “owner” despite one being onna and one being ona.
I hadn’t been precise enough. I meant that the WK team has tried to make different mnemonics for the kanji reading. Three of those four examples are vocabulary … and well, let’s just say that, honestly, a lot of those are really random. I don’t think it matters too much because, on average, people use the vocabulary mnemonics much less. I mean, a lot of people here live in Japan or have one way or another a lot of exposure to vocabulary.
Random Example
I’ve never been in Japan, but おなじ was a word I already knew. When I found 同じ , having already learned that 同 means “same”, I immediately read it as おなじ (the okurigana help…). For a lot of words it’s like that…a lot of people skip many vocabulary mnemonics.
That being said, I totally agree that the system could be improved with more well-structured and coherent mnemonics.
P.S. I didn’t bother to look it up earlier, but 守 is the one I was pointing to earlier with a weird mnemonic Should it be reported? I don’t know, but it sure is confusing. The mnemonics are:
数 sue —> すう
崇 sue ----> すう
巣 suit ----> す
守 sue -----> す [Why oh Lord? Why Kanisama? Was a mnemonic with “suit” not the logical thing to do? I can remember it now, but I know I’ll get it wrong when it’ll come back as a burn review]
@I_I I don’t think there’s a tool that uses the WK API to make the calculations you did, so if/when you make it public please ping me! (A separation between Kanji and vocabulary would be cool. Statistics about Kanji from Appr 2 to Guru 1 are always skewed since we want to level up as soon as possible)
Update: The initial version of the tool is now live! It’s very barebones - there’s no caching or error handling or fancy features, but it should be good enough for now. Speaking of which, if anyone who’s been using WK for a long time is willing to share their data, please send me your API key. It would be helpful to be able to test future versions of the tool with much larger amounts of data, since I’ve only been on WK for a month and a half.
Anyway, here’s my new stats. With the extra week and a half of data since my original post, the pattern is even clearer than before.
My Radicals 1-4 and Kanji 1-4 can’t be trusted, since I override those first stages to level up. Vocabulary is the one that tells more in my case. “Apprentice 1” (i.e. the passage between Appr 1 and Appr 2) has the lowest passing rate. And I spend more time than the average person on the lessons…Apparently I’m not the only one though. (The same trend is present in the Kanji part, due to the first ~6 levels where I wasn’t overriding yet)
Everything else is more or less the same for me. I don’t have burned items yet, but I assume it will be as bad, if not worse than the “Apprentice 1”.
@l_l
Suggestion: Since the passing rate refers to items crossing from one SRS interval to the other wouldn’t it be clearer to write both SRS stages? For example:
“Appr 3 → Appr 4” instead of just “Apprentice 3”
“Guru 2 → Master” instead of just “Guru 2”
“Master → Enlightened” instead of just “Enlightened” [By the way, this one shows the new stage, instead of the leaving stage. I’m guessing because you didn’t have access to those data with your API.]
P.S. Not sure it's a safe practice, but here is mine: (reading only obviously)
xxx
Out of curiosity, how long do you need the API for? It is read only, but still I don’t want the whole community to know when I’m online Too paranoid, no one cares…I know… so I plan to make it expire some day
Not sure how you or everyone else studies vocab, so I can’t answer that.
TL;DR : Possibly (too much) exposure and listening to the words
A lot of the words I already knew (or had at least heard a lot of times so they “sound right”, not sure how to describe it better). Other than that, jukugo ones depend on whether you remember the kanji reading. I also listen to the two audios at least a couple times trying to say it the same way (and I listen to them during the review, there’s an option that plays one automatically), so when I’m in doubt about the reading I say it aloud and see if it sounds well. It is useful with the meaning too, because if I guess the right meaning and I know the word, I can write the meaning too. It doesn’t work to well with crazy readings that I couldn’t guess even in a hundred years though…
It would be interesting to see the data from people who live in Japan and know a ton of vocab. There may be a divide between the results of people who already know the words and people with little exposure.
I find it really interesting that nobody else has the Guru 1 dip like I do, and everyone has much higher Vocab accuracy in general. I wonder how people are studying them.
As for me, I don’t put any special effort into memorizing the vocab and just let the SRS work its magic. Unfortunately, that means I’ve often forgotten them by the time the first review rolls around, and also tend to forget them in the week between Guru 1 and Guru 2.
Very interesting tool! I plan to monitor my progress using it. Pity it’s so new/unknown, so not many people posted their numbers, and it’s hard to see if one is doing “ok” or not.
UX suggestion - round up percentiles to two digits after decimal. Also maybe reformat into three-column table to align SRS level, percent, and (n/m) (I get that building html with js is no fun
Here is mine, I'd say it's all over the place, hard to say which SRS level is the worst.
In case anyone is curious, here are my updated numbers. With an extra two weeks of data, my Guru 1 dip has almost disappeared, so I guess it was just a random fluke. There’s still a sharp drop from Apprentice 4 to Guru 1, but Guru 1 has gone up a bit and Guru 2 has gone down a bit so they’re now pretty close. There’s no longer a sharp drop specifically at Guru 1 followed by a sharp rise back at Guru 2.
The numbers are a bit skewed due to some time I took off last year, but fairly representative of how I’ve been doing. My apprentice 1 numbers are pretty low since I don’t have time in my schedule to do the review within the same day most times. And Master/Enlightened have suffered from the break I mentioned earlier. I’m looking at Kanji I haven’t seen in 4 months lol.
Funnily enough, the more you use WK the better you get at memorizing kanji.
For the last couple levels, I’ve been putting a lot more effort into learning the vocab, which has helped quite a bit However, it seems to help most with the apprentice reviews. I still miss a lot of the vocab once they come around for the Guru 1 or 2 reviews.
Well it depends, but kunyomi readings tend to be a lot harder to remember, as are rendakus and the like. Early on, I had a lot of trouble remembering all the X things and X days words until I specifically went through and made a chart to remember which ones use the little tsu. Also words with meanings that don’t really relate to the kanji can be difficult to remember. I also tend to mix up transitive/intransitive and words with similar looking kanji. Also, sometimes I just get careless and miss reviews even though I knew the answer.
(The same trend is present in the Kanji part, due to the first ~6 levels where I wasn’t overriding yet)
