和英辞典Vocabulary dictionary

漢字辞典 Kanji dictionary

文法辞典 Grammar dictionary

例文検索 Sentence lookup

test
 

掲示板 Forums - [Discussion] Kanji reading vector "redo" (fork)

Top > renshuu.org > Feature Requests/Improvements



avatar
マイコー
Level: 346

Note: this is at initial discussion level - it may become a feature, it may not. If it does become a feature, it could be by the end of the year, it could be a year or more down the road.


Both from current or former (but longtime) users, as well as users who gave renshuu a try, but said "this is absolutely not for me", one of the more common complaints/criticisms has been around the way that kanji is learned in the system. I'm extremely happy with how kanji integrates into vocabulary schedules, as well as the way that kanji/vocab helper schedules work.

The big gap, as I see it, it with the contents of the actual kanji schedules themselves. As you probably know, there are several vectors for the schedules, and they basically break the kanji up into constituent parts:
Onyomi, Kunyomi, Definition, Stroke count/order, Radical, Writing

Everything other than the first two - I feel they do a pretty decent job. A good argument can and has been made for splitting out radicals and making them first-class study terms alongside words/grammar/kanji/etc., and that is something I hope to do one day.

While some people can and do get benefits out of the current onyomi/kunyomi (we'll call them reading vectors), many people find them either offputting, useless, or both.


Until a particularly feisty email I received a few days ago, I had thought "I cannot really improve this without redoing the whole system", but an idea popped into my head, and I'd like to get your feedback on it.


Proposal

In addition to the current vectors, have a "new" vector that will be called "readings through vocab". Internally, it will be the same two vectors, onyomi and kunyomi. It'll most likely be an either/or thing.

If this new system is chosen, then (this is still the fuzzy part), I'm foreseeing two main differences:
1. the "new terms to study" page will get a bit of an overhaul. Specifically, there will be more focus on specific, common words that use the vocabulary, instead of "readings first, oh, and here are some words that use them". Instead of

"Here are readings", and "here are words with this kanji", it will be closer to "here are some words that use this kanji", with notations under each word that show how the reading is used (to help with okurigana, inflections, っ usage, etc.)

2. The quiz question types would be different as well. They would primarily revolve around showing those same vocabulary, and having you choose/enter the readings for the kanji you are studying. So you might get 食べる with a "blank" over the kanji.
This would also be good for the current question form that has 食/たべる as the question, and you have to choose 食べる or 食る.

There could also be question types more focused on the underlying readings, although this would not apply to all readings. So if you had 食器(しょっき), it might ask you what しょっ actually is, to help you understand that the reading is not that, but a collapsed/contracted form of しょく). Or, asking if it is kunyomi or onyomi (I could see this getting less usage, though, so probably off by default)


In order to avoid any MAJOR changes to the internals, these questions would still grade in the same way - if a question involves the reading た.べる, then the results would affect the mastery level of "kunyomi" for that kanji. renshuu actually does (surprise) track a "shadow" mastery level per specific readings on a kanji, but it's never been developed enough to really integrate into the current schedule system.


Anyway, I think this could possibly answer most/all of the complaints that many users have had about kanji learning on renshuu, while also allow those who benefit from the current system to continue to do so.

What do you think?

1
1 day ago
avatar

As someone who uses the existing system extensively, I don’t understand how this would change my day to day interactions with the SMS.
Number 1 sounds like the existing vocabulary list on the kanji card. Number 2 sounds like some of the existing advanced question types available to pro users.

I might understand the proposal better if there were mockups of the new screens.

2
21 hours ago
avatar
Shamugan
Level: 910

After hurting my brain reading a dozen studies about learning kanji: Yeah, it would help. But it probably won't wipe out all the complaints.

Initially, I wrote a long text (as per usual) because I had a slightly different opinion, but... I was too confident about how much my own approach would help. There's definitely a lot of evidence in favor of your proposal, too.

My initial idea actually went a step further than your proposal: learning kanji readings through short sentences and context, rather than vocabulary alone. Like Japanese children who already know the pronunciation and meaning of many words through spoken conversation before learning the kanji used to write them:

  1. You repeatedly hear いえにかえる in context.
  2. Through repeated exposure, かえる gradually becomes recognizable as "to return/go home" in that context, and that particular reading-meaning association becomes more stable (distinct from the other かえる words).
  3. You start associating that already familiar reading-meaning representation of "かえる" with its written form, 帰る, for example through copying practice.

Context in 1) helps narrow down the possible meanings, so you don't have to deal with all the other かえる words at the same time. Repeated exposure then helps strengthen that particular association between かえる and "to return/go home". And copying can help associate that already familiar reading-meaning mental object with its written form. Which makes it easier to recognize later, even if people don't necessarily remember how to write even basic kanji.

The advantage is that you don't have to juggle 2–6 readings, several meanings, and an unfamiliar shape all at once. You start with simple sentences that narrow down the possible meanings, both between different words and between the different senses of a single word, instead of getting overwhelmed by dozens of readings associated with multiple meanings (because kanji meanings are often defined through simple Japanese words, especially those corresponding to their kunyomi. So learning those words is already learning part of the meaning of the kanji they'll encounter later. In a way, you're learning both at the same time, since those words are already some of the basic building blocks of kanji knowledge). And all of that interacts in the learner's brain and can interfere with each other. Learning 4–5 different かえる at the same time just makes all of them unstable. And if those associations don't get stabilized, they can interfere with each other, making them harder to remember and easier to forget.

But that's just one learning sequence. Among many others. I even found a study where learners started with writing and still got good results...

So, yeah, I think it would help, and there's research supporting the general approach too. Also, in one of the Japanese apps I use for the Kanji Kentei, readings are always practiced in context, even during writing exercises. And by context, I mean short sentences. For example, 大きな◯ (with おと as a reading hint) for writing practice, or 雨がやんだ with 雨 highlighted for reading practice. Another example is 右にまがる, with 右 highlighted. And in every case, the sentences avoid advanced kanji too. So you're also laying some groundwork for learning future kanji by getting familiar with the words first :D (well, for まがる at least. やむ probably isn't the best choice for a common reading).

At the same time, I wonder whether a first guided, positive experience with kanji could have a greater impact than any single feature. Found a study on that. People who have a positive attitude toward kanji tend to use a wider variety of learning strategies, while those who perceive kanji as difficult tend to rely more on rote memorization. Of course, that doesn't necessarily mean a positive attitude causes better learning strategies. It could also be the other way around, or a bit of both.

So maybe a basic premade kanji schedule with a few simple sentence or word companion schedules (like the linked grammar/word schedules) could help create that first positive experience and possibly influence how beginners approach kanji later.

Also, maybe without kanji like 生, 下, 上 at first. Because, yeah, they're common, but they can create a pretty terrible first impression of kanji. Their frequency doesn't necessarily justify overwhelming beginners with all those readings, even if they're really important. Especially foreigners. Japanese children already know many of the words associated with those kanji through spoken Japanese before they start learning their written forms.

1
16 hours ago
avatar
マイコー
Level: 346

Thanks for the responses. Let me see if I cannot clarify it a bit without making mockups - not really something I (unfortunately) have time to do.

1. "kanji/words in context" - this would be a dream addition, but not one I have data for. Unless they are using AI, then generating even 2-3 word snippets (which are also quite common in 国語 textbooks for elementary school kids over here) is a huge undertaking. Books "cheat" in a way because there is no expectation that you return to the same page and expect different snippets to come up, but I think a lot of the expectation of a platform like renshuu is to not have the exact same question shown to you each time, so it generates a lot of push to have multiple snippets for any given word, as well as coverage across a ton of different words (since renshuu gives you the freedom to study whatever you want). With kanken apps (or whatever), the app typically chooses what words it is going to show you (with no option to add your own stuff), so they can greatly limit the scope of where content needs to be created.

But yeah, a definite dream addition for me.


2. Most of the "advanced question modes" that are available currently would not work at all with this new mode, for a few key reasons.

A. A lot of the ones that involve vocabulary (lihe kanji crosses, for example) are restricted to 熟語 (2+ kanji compounds), so kunyomi are almost entirely unrepresented. Additionally, as they have 2+ kanji in them, you need to know both kanji, making them much less useful for beginners (I've gotten numerous questions over the years from N5/N4 level learners, with <200 kanji under their belts - wondering why so many of the advanced modes do not appear in their quizzes).

B. This is probably the biggest one, but the advanced modes are built with the presumption that the words that come up are ones that you already know.

C. Even if we are to remove kunyomi from consideration (as it is badly represented by the vocab-based advanced modes), there is no guarantee that all common onyomi can be represented in the advanced modes (which is why the system requires you have at least some of the multiple choice/fill-in-the-blank options chosen as a fallback). So currently, there is no way to fully avoid "direct" onyomi/kunyomi questions.


In my head, the largest change here is that the structure and flow of the questions ensures a few things:

1. You never need to answer a direct "here is a kanji, produce the reading" (or reading > kanji) question.

2. Every reading is displayed in the context of a word that you either know or are learning through the kanji schedule

3. There is still some language on the screen (namely in the results panel) that helps create a link between that reading and the kanji. (So if the word 直観 comes up, and you're studying 直, the results panel will show the "reconstruction" of ちょっ -> ちょく

(nothing is formalized, but I could see it also use that panel to pull up other words you've studied that share the same reading for additional reconstruction)


0
6 hours ago
avatar
Carnelian
Level: 1365
マイコー (11月17日, 8:04)

1. "kanji/words in context" - this would be a dream addition, but not one I have data for. Unless they are using AI, then generating even 2-3 word snippets (which are also quite common in 国語 textbooks for elementary school kids over here) is a huge undertaking. Books "cheat" in a way because there is no expectation that you return to the same page and expect different snippets to come up, but I think a lot of the expectation of a platform like renshuu is to not have the exact same question shown to you each time, so it generates a lot of push to have multiple snippets for any given word, as well as coverage across a ton of different words (since renshuu gives you the freedom to study whatever you want). With kanken apps (or whatever), the app typically chooses what words it is going to show you (with no option to add your own stuff), so they can greatly limit the scope of where content needs to be created.

But yeah, a definite dream addition for me.


This will certainly be a huge undertaking. If displaying collocated words rather than whole sentences is the goal, the Tsukuba Web Corpus can be used to find top results for, say ○○を食べる or ○○に帰る. They may have to be filtered for difficulty and for avoiding words like 何 or もの. In the case of ○○を食べる, top results could be something like ご飯, 料理, 肉, 野菜 and 魚. The corpus shows that を食べる and に帰る are respectively the most frequently seen combinations of the particle + 食べる and particle + 帰る as well, and just restricting it to those and keeping to the dictionary form (or a simple, frequently used form) may work. By not having sentences and just using words that are found in the dictionary, translation issues can also be avoided. I'm mentioning this as a possible way to realize an addition like this. It's undeniably a lot of work.


Another possibility is to make greater use of words containing case-marking particles which are already in the dictionary, such as お腹が空く.


In my head, the largest change here is that the structure and flow of the questions ensures a few things:

1. You never need to answer a direct "here is a kanji, produce the reading" (or reading > kanji) question.

2. Every reading is displayed in the context of a word that you either know or are learning through the kanji schedule

3. There is still some language on the screen (namely in the results panel) that helps create a link between that reading and the kanji. (So if the word 直観 comes up, and you're studying 直, the results panel will show the "reconstruction" of ちょっ -> ちょく

(nothing is formalized, but I could see it also use that panel to pull up other words you've studied that share the same reading for additional reconstruction)


​These changes could make the kanji schedules more compelling. Perhaps when the correctness is revealed, it could also be stated that ちょく is onyomi. From what I understand, some learners, even at N1 level, simply give up on making the distinction between onyomi and kunyomi. This could be a way to passively learn it/see this information instead of quizzing to learn it.

0
4 hours ago
avatar
クリス~エム
Level: 335

I like the idea of being able to learn kanji as the "primary" new-term schedule, but I would currently struggle with learning them in isolation. If it would present example words as part of the introduction proper (not just the examples tab like now), use those in the quizzes and maybe even automatically add those to a vocab schedule, then it could be useful.

0
3 hours ago
Getting the posts




Top > renshuu.org > Feature Requests/Improvements


Loading the list
Lv.

Sorry, there was an error on renshuu! If it's OK, please describe what you were doing. This will help us fix the issue.

Characters to show:





Use your mouse or finger to write characters in the box.
■ Katakana ■ Hiragana