I have been wanting to write this since a long time ago and never found the time or the motivation, but after some recent changes made by WK and reading the response of the community I felt I have to.
First of all I want to say that thanks to WK I have been able to learn and love lot of kanjis and I have always been very motivated. Anyways in the last year I felt that motivation has been decreasing because of the WK vocabulary.
We all know that the WK not only uses SRS to help to memorise kanjis but also uses some words which put the kanjis in context to help to retain them better and Iām totally fine with it. However Iām feeling that the WK is trying to āforceā users to learn words, rather than using them to learn kanjis, and imho itās been done totally wrong.
To put my opinions in context I have to say that I started WK at the same time I started learning Japanese (2,5 years ago) with private teachers, schools and using other resources like Anki. My struggles with WK have been:
I donāt always read the full usage of the vocabulary to get the nuance of them so I know itās my fault in this case but most of the time I try to use some WK word I have never used I do it wrong and the teacher looks at me like wtf. Many times he/she tells me that those words are hardly/never used.
Learning vocabulary is not customisable at all. I would prefer to focus on kanjis, learn maybe one or two words with each kanji and leave my teachers, classes and my own resources to teach me words according to my level, not to WK level.
My message to WK is, with no hate and all my love, that you as an app never offered or promoted a proper vocabulary learning experience but instead a good Kanji learning method plus learning words (almost as a side effect).
So WK, before forcing the users to some unpleasant changes and introducing new words, please hear all of the complaints the users have made in the past (lack of customisation, too many synonyms, improvements in mnemonics, not being able to focus mostly on kanjis, etcā¦) and do something about it to improve the app experience and after that maybe you can implement your own system to learn vocabulary, which again, due to the order of kanji learning implemented by WK is at the moment incompatible with the natural learning of the Japanese language.
This is a common occurrence within wanikanis system. There are many vocabulary that donāt actually serve a meaningful purpose and could clearly be replaced by more helpful vocabulary. I totally feel where youāre coming from, brother. Iāve been pointing this stuff out for years haha.
And not just reduced, but I think we could even up the usefulness of the words on this site without affecting learning of kanji as well! I canāt be bothered to find it, but I remember pulling up a random level like 2 or 3 years ago and showed how by using all the same kanji and readings, I could alter 5 words or so to where each of them was significantly more common or roughly the same.
Yes, weāre here to learn kanji, but what good does kanji alone do if not applied into real words and if WK uses weird words to reinforce them instead of more common ones? At the end of the day itās the latter weāll see and use. Anyway, can never understand that logic. I fully get where youāre coming from.
WKās approach is definitely quantity over quality
which is fine for learning enough kanji so that you can start reading. itās fine for bootstrapping that kanji knowledge to the point where you can use reading to provide the context which you absolutely need to understand the words.
but there is no way you can teach a language with stupid 1 to 1 equivalencies between vocab, given with only the slightest of contextā¦
@darkvitae81 spot on. Big thanks for putting your thoughts in points so clearly. Alignment with JLPT levels and the natural learning progression is something Iām currently using as reference in my own kanji app project.
I recently wrote an API in JavaScript (node.js) to support pulling vocab from Jisho via their REST API endpoint (unlike existing Jisho libs in JS which were either old or would scrape the HTML results). Wild cards + filtering for ācommonā vocab are supported out-of-the-box so itās possible to take those WaniKani kanji and instantly pull better words .
Whoops, sorry for the double-post! Meant to edit the previous one.
If you had some automated way to generate sayā¦at least 1 entry containing whatever readings you wanted for all the kanji you wanted optimized for some frequency list, it would be interesting if you did that with jpdb.ioās top 30k lists. You can make a top 30k for fiction, one for anime and drama, one for non fiction, one for visual novels, etc. and it would be interesting to see the wordlists it generates to teach you all the readings for the kanji that you wanna know/reinforce while using the most common words for one of the respective frequency lists. It would be a pretty sick way to have the structure of wanikani while making sure all the words you use are actually quite useful (unlike wanikani) for the content you want to consume.
Hmm interesting. I havenāt thought of checking specific readings for a given kanji.
The Jisho.org REST API is just a more limited version of their āsearchā dictionary UI. It doesnāt cover kanji records, for instance, only vocabulary. Also, not all of the query tags actually work (for instance, #adjective doesnāt actually give you only adjectives). So one can query for words containing a specific kanji only.
I would then have to turn to jpdb.io as you suggest.
For now I would focus on words which are common and likely to appear in any media. Iām pretty sure one can derive a working corpus of ~4-6k words which would be useful to reinforce kanji readings and as a starting point for reading native content. That coupled with JLPT level sorting, for instance.
One thing I would be a little wary about in terms of content-based progression is that unless one takes the approach jpdb.io takes with kanji ā simpler kanji breakdown, it would be difficult to arrange the kanji in order of complexity to ease users into kanji better without making some assumptions.
@Vanilla@ResFort@Mrs_Diss@llinchou thanks for all your comments and specially for letting me know about jpdb.io which i just discovered, started patronizing and loving it already. much more useful to learn vocabulary and kanji than WK
im really sad to know that many of you have complained and nothing was made so im kind of giving up with WK, at least partially
will start filtering out words/radicals with reorder omega script and learn and review only kanjis (at least i will give a meaning to my lifetime subscription) and will learn useful and meaningful vocabulary with anki and jpdb.io
Glad to hear something came out of this discussion!
Despite all of my complaining (and there is a lot of itā¦ ), and especially now that I keep working on my own kanji app, I still think WaniKani has some merits. The problem is that it made some very poor design decisions (choice of vocabulary, for instance, as highlighted by Vanilla) early on and instead of trying to iron those out, itās now trying to reinvent itself (major code base refactor, summary page downed and forever in process of being replaced, kana vocab, etc.) despite not having the developer capacity to execute.
One reason I think many other apps are running in circles around WK is because the devs behind them are both very talented and have a clear, realistic vision of what they want to achieve.
Speaking of, Iām going to need you to hurry it up Linchouā¦ Iām ready to flip tables dealing with this program.
Lack of leech management is a killer. As well as I just failed 100 items out of 200 in a review as they were all past leaches I totally forgot (oops). The timing should be based on how often in general you fail the same word, even if you get it right consecutively a few times later. Or something.
Edit: Makes it worse about 1/3 of these words arenāt even commonā¦ so whatās the point honestly?
Apologies that itās taking a while . 2k kanji is no joke
On a serious note, I now have 4 levels worth of kanji sorted and am working on the designs of the UI views (kanji learning page, vocab learning page, main/landing page, etc.). Next up is setting up the database table models and the app logic itself. Later down the line the āfunā part - writing mnemonics .
I have a little more time on the weekend so hope to get at least some parts of the app running so I can start prototyping it better.
And since weāre all here, I just had a revelation today morning - one doesnāt actually need a study session. Reviews can be done on a per-item basis. A well designed backend should be able to handle the requests.
That part one could still have. The accuracy metrics are a completely separate business . Itās just that managing a study session on-the-fly + building summary stats from that 1 study session is not necessarily as relevant as you being able to see your overall accuracy (and other stats, like leeches!) over time.
EDIT: Need to add an extra view for the summary page it seems
This is actually the main reason why I personally disliked the kana addition. You canāt really acquire new vocabulary purely through low-quality out-of-context exposure. Kanji words are mostly fine, since they still help you drill the reading and provide at least minimal context through the meaning of the kanji itself, but for kana words that just doesnāt apply at all. To me it feels like WK is trying to be more complete than it has the potential to be, rather than expanding and improving upon what itās designed to be and making sure itās fulfilling its role as well as it can.
I will say the whole āWK doesnāt teach common wordsā is very relative because a lot of the time thereās stuff like åäŗŗ which isnāt as common as åé but it is a word youāll see used
However, I think this is also part of why people got so up in arms about kana vocab even though only 60 were introduced originally
(Althought Obviously it was a pilot run before they added more, so people feeling inconvenined by Just 60 were right to speak out lest they just stay quiet, be misinterpreted as being ok with it and having to deal with 500 more)
The kanji vocab in WK already feels bloated
Like theyāve recently introduced a level up banner, which is fine
but like levelling up always feels weird to me bc I know its going to be weeks before I actually get to that next levelās content
partially this is my fault for being slow, but even when I went faster, itād take two weeks from the start of getting to radicals to level up, then itād be two more weeks of vocab because thereās so much of it
One thing thatās been annoying me recently is when a word is introduced as a noun (like ē®å®) and then the same word will appear again as ē®å®ćć
A few of these could help with internalising grammar but at a point it feels like just padding
Also to be clear, I like the structure wanikani provides, so Iām not looking to move elsewhere but so often I see talk of adding more
and itās like, unfair comparison, but I think sometimes of Duolingo, which people who arenāt really in the Japanese Learning Community, will say āoh wow Iām making so much progress so fastā to refer to taking several months to learn kana or a few years for N5
When really they just donāt know that thereās better ways
And donāt get me wrong, Iām fine with taking my time, I donāt think MAX SPEED is always best
but I do appreciate some efficiency, its why I liked tofuguās kana guides to begin with, because mnemonics worked to hurry up the recognition process and I could start reading it as sounds then words after just a few days rather than rote memorising shapes for weeks on end
Think so as well. Originally it shouldāve been only the most common vocab to support kanji readings - ~3 words per kanji when itās introduced and automatically more as that kanji is used in higher level words. No need to keep on adding more or using rare words Japanese people donāt use. 3 words per kanji is already ~6k words in total. If these contain (extremely) rare words, thatās a lot of overhead. Weāre talking about ~2k kanji in a year here after all .
Yup, not quite sure whatās the reasoning here either. There is thousands of ćć compounds in general. Itās enough to keep only ē®å® + add āćć compoundā (or similar) as a label, maybe add English meanings which emphasize that the noun expresses an action. That would be more than enough. Having both ē®å® and ē®å®ćć is very confusing.
.
.
. 2k kanji is no joke