Scaling the 6218 review wall

[l_l]

Today was pretty frustrating, with once again missing a bunch of reviews I should have gotten. For example, I remembered 仰ぐ because it was one of the reviews that I missed early on last week and it came back up today, but I still missed it again due to getting distracted.

Also, I missed the “look at” kanji due to putting “see” as the meaning, which is complete BS (yes, I add a synonym in such cases, but that doesn’t help retroactively).

And of course, transitive/intransitive pairs continue to be a huge pain. I’d forgotten just how annoying they are. The latest victim is 泊まる which I missed due to guessing wrong on transitivity for the meaning.

My review count actually went up compared to yesterday for the second time. Not that it’s too surprising. I have to run just to stay in place.

Tue May 18 2021:
Time spent: 34m
Reviews completed: 66
Reviews remaining: 6078
Accuracy: 40.91% (27/66)
Accuracy on previously burned items: 60.00% (18/30)

Current item counts:
Apprentice: 2913
Guru: 1200
Master: 1344
Enlightened: 3074
Burned 379

[2000kanji]

@l_l

I missed the “look at” kanji due to putting “see”

Yes, this bunch of look, see, view, viewing, look on, look at, look back on, … is a pain. If encountered in a text I am pretty sure to get the meaning, but seen alone I always mix them up.

[Mrs_Munch]

i know you’ve said that you don’t want to use a re-order script, but i think with so many reviews the SRS can’t function properly on its own.

under normal conditions, if you fail a review, the item comes back relatively soon. then you repeat the item, and reinforce what you’ve learned. this works because in normal conditions you can do all your reviews every day.

however, this doesn’t work when you’ve got this many reviews. if you fail an item, and do 60 reviews the next day, the probablility that that item will return is about 1%. indeed, it might be weeks or months before that item returns. the whole concept of spaced repetition crumbles.

by prioritising certain items (perhaps by level: top-down or bottom-up, perhaps by most recently reviewed, there might be other strategies) you can reduce the pool out of which WK selects the items it presents for review. and you can get your items at appropriate intervals again.

i hope you will take this post in the way it was intended: as well-intentionned advice. i can empathise with wanting to brute-force your way through this pile of reviews. however you attack this problem, i wish you the best of success! ^^

[l_l]

I don’t think it’s all that important since the SRS intervals aren’t perfectly calibrated to begin with, and I’ve already learned everything once, and the fact that there’s a lot of interlinked items (kanji vs vocab, related vocab words, etc.) means that the SRS system is often defeated anyway.

[l_l]

Wed May 19 2021:
Time spent: 42m
Reviews completed: 79
Reviews remaining: 6083
Accuracy: 44.30% (35/79)
Accuracy on previously burned items: 71.43% (20/28)

Current item counts:
Apprentice: 2926
Guru: 1206
Master: 1327
Enlightened: 3068
Burned 383

[l_l]

I was feeling pretty discouraged this afternoon, so I guess it’s fitting that I just burned 頑張る.

[l_l]

Oops, forgot to do an update last night. The stats should have been something like this:

My review count went up again, but on the other hand, accuracy was higher than normal, and I burned 頑張る, which was kind of cool.

Thu May 20 2021:
Time spent: 33m
Reviews completed: 55
Reviews remaining: 6117
Accuracy: 50.91% (28/55)
Accuracy on previously burned items: 77.27% (17/22)

Current item counts:
Apprentice: 2942
Guru: 1208
Master: 1316
Enlightened: 3065
Burned 386

[l_l]

Accuracy was pretty high again today.

Fri May 21 2021:
Time spent: 41m
Reviews completed: 74
Reviews remaining: 6106
Accuracy: 52.70% (39/74)
Accuracy on previously burned items: 78.79% (26/33)

Current item counts:
Apprentice: 2953
Guru: 1212
Master: 1302
Enlightened: 3059
Burned 391

[Aendur]

Something you might want to try is to review a bunch of kanji before reviewing vocab, even though you have already learned them in the past.

Because vocab are comprised of kanji, reviewing kanji first might help you recall both their reading and their underlying idea, which, by extension, might as well help you recall the words they form; and also, reviewing them first will increase the chance you will have already encountered them on that or on a previous session.

Then, since this might help your memory re-create better, stronger associations, you might see your accuracy improving faster (as a consequence of such associations, not as an end).

Regardless, good luck on your quest and keep up the great work!

[morteASD]

Have you thought about rearranging the items with a script from ‘most urgent’ to ‘least urgent’? Meaning you get to see the items with the shortest interval first? That way the SRS would work at least partially.

[l_l]

I spent a lot more time on WK than usual and did more reviews today. But progress is still very slow.

I also started calculating my accuracy rate on items that weren’t previously burned, so you can see how much my accuracy rate is affected by the easy resurrected items.

Sat May 22 2021:
Time spent: 88m (1 hour, 28 minutes)
Reviews completed: 156
Reviews remaining: 6078
Accuracy: 39.10% (61/156)
Accuracy on resurrected items: 63.08% (41/65)
Accuracy on non-resurrected items: 21.98% (20/91)

Current item counts:
Apprentice: 2973
Guru: 1233
Master: 1272
Enlightened: 3043
Burned 396

[l_l]

I decided to calculate my proportion of reviews and accuracy by level.

As I expected, most of my reviews have been from the low and high levels with few in the middle. When doing reviews, there’s basically two kinds of reviews that come up. There’s the items from the low levels that I burned last year and then resurrected recently. And there’s the items from the high levels which I never mastered in the first place before my subscription ended and thus are still bouncing around the low SRS levels. However, the middle level items are mostly stuck at Enlightened, and thus not coming up for review again.

On the other hand, my accuracy by steadily and dramatically decreases by level. Obviously, it’s not surprising that I’d do much worse on the higher level items. The only surprising part is how clear and dramatic the pattern was. I had expected it to be messed up somewhat just from randomness.

Proportion of reviews by level:
1-5: 8.1%
6-10: 10.9%
11-15: 10.4%
16-20: 11.3%
21-25: 9.8%
26-30: 10.2%
31-35: 7.0%
36-40: 4.1%
41-45: 3.8%
46-50: 6.7%
51-55: 8.1%
56-60: 9.5%

Accuracy by level:
1-5: 70.9% (61/86)
6-10: 69.0% (80/116)
11-15: 64.0% (71/111)
16-20: 66.1% (80/121)
21-25: 51.4% (54/105)
26-30: 54.1% (59/109)
31-35: 36.0% (27/75)
36-40: 20.5% (9/44)
41-45: 22.0% (9/41)
46-50: 12.5% (9/72)
51-55: 10.3% (9/87)
56-60: 10.9% (11/101)

[Heiwire]

Best of luck to you I woke up to 800+ reviews today due to me skipping a couple of days last week on Wanikani.

[l_l]

Sun May 23 2021:
Time spent: 61m
Reviews completed: 108
Reviews remaining: 6086
Accuracy: 50.00% (54/108)
Accuracy on resurrected items: 79.59% (39/49)
Accuracy on non-resurrected items: 25.42% (15/59)

Current item counts:
Apprentice: 2984
Guru: 1246
Master: 1251
Enlightened: 3037
Burned 399

[smellslikerain]

Well clearly the SRS worked it’s magic if you remember the older levels! Keep going at it and keep your chin up, you’ll get there!

[l_l]

Yesterday, I got 仲直り and 慶事 right, which is notable because they are ones that came up for review when I first started that I missed.

One of the reasons I initially thought I’d be able to get through the review wall quickly is that I figured that even if I didn’t remember any of the high level stuff any more, I’d remember it once it came up for review and I missed it and saw the right answer and would be able to get it right the second time. Unfortunately, this is taking a lot longer than I expected, partly because the extreme size of the review pile means that it takes a long time for missed reviews to come up again and partly because I often don’t remember them - there’s been quite a few words that I’ve missed twice or even three times these last two weeks. But hopefully this is a sign that my accuracy will just keep going up from here.

[l_l]

Too tired to do many reviews today. My accuracy on resurrected items was unusually low as well.

Mon May 24 2021:
Time spent: 36m
Reviews completed: 65
Reviews remaining: 6113
Accuracy: 43.08% (28/65)
Accuracy on resurrected items: 54.55% (12/22)
Accuracy on non-resurrected items: 37.21% (16/43)

Current item counts:
Apprentice: 2990
Guru: 1252
Master: 1238
Enlightened: 3034
Burned 403

[l_l]

I burned an unusually high number of items in this morning’s session. That’s mostly luck, but it’s still interesting. What’s really interesting is the factor that “instinct” plays.

For example, when I saw 熊本県, I immediately thought “kumamoto”. But I didn’t actually remember that reading of 本 and thought it had to be “hon”, so I assumed I was just misremembering. However, I couldn’t think of anything else that “felt” more right, so I ended up just guessing kumamoto and of course it was right after all.

I also very nearly burned 将来. When I saw it, I immediately thought “shourai” and “future”. However, I second guessed myself on the meaning half. I thought “it has the commander kanji, so there’s no way that means future. It’s probably just a homophone of some other word that means future”, so I got it wrong.

Of course, my instinct isn’t always right. When I saw 失恋, I immediately thought taihen (大変). Presumably, my subconscious mistook 恋 for 変 (ug, why are they so similar?) and didn’t look closely. It’s especially embarrassing since I just finished reading Yotsubato ch25, where 失恋 comes up very frequently.

[sporadic]

So, I’ve been watching this for awhile. I have to ask about your stats. You seem to be only doing 2-3 reviews a minute. Is that true? What’s your process during reviews?

When I’ve got large (read, 200+, not 6800!) piles of reviews, I give each a few seconds and then mark it wrong. I figure if I don’t know it well enough for it to be readily accessible, I’ll almost certainly get it wrong next cycle, so might as well fail it quickly to get through the backlog.

I don’t know if that would help or not with such a pile, though.

[schtitt]

I find that spending at least a little time seeing if I remember it more strongly reinforces the mnemonic, because when and if I do remember, it’s a stronger feeling of “aaah of course” and it seems to me I remember these better in later cycles.

Edit: I guess what I’m saying is that the ‘struggle’ of remembering actually reinforces better than just giving up and then looking at the answer.