Shelf and Moreover

Some of the kanji numbers have changed with production of the last couple of lessons, mostly because I had a number of kanji marked as being dependent on ‘shelf’. These included ‘help’ (), ‘investigate’ (), ‘tatami mat’ () and ‘best regards’ (). Although these kanji do contain a radical often known as ‘shelf’, the appearance of that radical is usually the same as ‘moreover’ (), which is only reached as part of the second five-stroke core kanji lesson.

These kanji had been introduced prematurely because, confusingly, there is also kanji called ‘shelf’ (), which only needs ‘tree’ () and ‘moon’ ().

Moving the kanji around has changed the numbers, which will mean that the current version of the Android app will have slightly different numbers to the online lessons for now. The numbers will be reconciled with the next release.

Japanese Keywords Update

As discussed in a previous post, some kanji students using a Heisig approach like to go back over the list a second time, but using Japanese prompts instead of English prompts. This a great way to contextualise the kanji knowledge you already have and begin the transition to thinking of the kanji as a working part of your Japanese vocabulary, rather than as a bunch of shapes with English labels.

Heisig explicitly recommends deferring all vocabulary acquisition until all the joyo kanji have been learned in the limited, English-keyword-only sense, and his book series leaves Japanese readings and vocabulary until the second and third volumes. There is much that is good about this advice – learning the shapes of the kanji is actually the easy part of acquiring Japanese, particularly if the kanji are broken into radicals, a spaced repetition system is used, and you are comfortable using mnemonics. It makes sense to tackle this easy part and then approach the acquisition of vocabulary when you already know your kanji characters. The problem is that the approach assumes everyone is patient enough to finish the entire joyo kanji before they move on to the next stage, and students may feel that they are putting a lot of work in and still don’t know any ‘real Japanese’.

I suspect there is another problem with a pure Heisig approach, related to learning efficiency, though I do not yet have any evidence to back up my intuition. Spaced repetition and mnemonics are great for acquiring vast quantities of new knowledge efficiently, but deep knowledge of any factual item is greatly enhanced if the item is more than a passive bit of remembered text on a flashcard, and instead becomes a connecting element in a network of knowledge. If you already know that ‘rain’ is associated with the kanji ‘‘, and can reproduce it after an interval of a couple of weeks, you are probably ready to learn that the Japanese word for ‘rain’ is ‘あめ’ (ame). While you are reinforcing the stroke pattern at progressively longer intervals, you might as well use that time and effort to learn the word ‘あめ’ – and using the stroke pattern of as your link to the new vocabulary means that is no longer at the edge of your knowledge, it is a pathway to the next bit of knowledge you are trying to acquire. (That is, it seems as though you are no longer actively learning ‘‘, because you are now concentrating on ‘あめ’, but you are in fact reinforcing ‘‘ anyway, in the most valuable way possible.)

With these potential benefits of Japanese keywords, why wait until you know 2,042 English keywords?

The full release version of KSP will be offering a graded transition to Japanese keywords, as a configurable option. I envisage each item will pass through the following steps:
1) English keyword… to kanji (concentrating on stroke pattern)
2) English keyword… to kanji (Japanese keyword pops up on completion, and is absorbed passively)
3) Japanese keyword with delayed English pop-up… to kanji
4) Japanese keyword with English pop-up available as a hint.. to kanji
5) Japanese keyword… to kanji

Users will be able to configure each step to occur automatically, when the item passes a certain number of successful reviews, or the transition to the next level could be individualised per item, based on the number of times the item has been claimed as well known. That is, if the English keyword easily evokes the image of the kanji, a user might claim it as easy and from then on the Japanese keyword will pop-up on completion. If the user claims the item as easy again, it will switch to a Japanese keyword with a delayed English pop-up, and so on.

The problem with this approach is that there is no widely accepted list of Japanese keywords for the joyo kanji. I am currently going over available lists (as discussed in the previous post), and I have compiled a table in which Wrightak’s keyword suggestions and the one from kazemakase are listed side by side, in KSP order. About half of the time, the two lists are in agreement, making the suggested keyword an attractive choice for KSP. The other half of the time, those two lists disagree, and KSP will have to choose between them or come up with a third choice. Another issue is that there are a large number of homonyms, so some means of disambiguating these will be needed.

The combined Wrightak-kazemekase list is nearly ready, and will be posted soon as a downloadable Word document in the form of another update. Defining a final KSP list and implementing the changes suggested above will be a much more complex process.

Keyword Usefulness Test – Update

I am part way through scoring the vocabulary items found at the vocab link on this site. (See this previous post for details.) So far the results have been surprisingly reassuring. Given some of the criticisms of the Heisig approach that are floating around the internet, it would be easy to conclude that the Heisig keywords are arbitrary and largely useless. Instead, most of the words on the vocab list use kanji in a way that is consistent with their keyword meaning. There are many more scores of 5/5 and 4/5 than any other scores.

Once all the items have been scored, I’ll provide some quantitative analysis of the scores.

Let me know if you dispute any of the scores. The utility of a keyword is, of course, subjective, but if there is general disagreement with any of the scores I will change them

Japanese Keywords

It is already possible to use Japanese prompts with the KSP Android app. Currently, swiping the English keyword changes the default prompt language for that kanji to Japanese, with the kanji of interest masked. The prompt must be chosen from one of four dictionary entries, which have been farmed automatically from a publicly available dictionary of common Japanese words. The options are not always quite what you might choose if you were constructing a list of Japanese keywords. The biggest problem is the large number of homonyms in Japanese: some kanji appear in words that sound the same as other kanji, so the prompt you choose from the dictionary may not be unique. (Although the kanji itself disambiguates the definition, it is masked – a classic Catch 22.)

The final version of the app will make it possible to have one dedicated Japanese keyword for each kanji, which will appear as a kana pronunciation (as well as a kanji compound with the kanji of interest masked). This will be similar to the process described by wrightak, here:

https://sites.google.com/site/wrightak2/

… but with some kanji, KSP will choose Japanese keywords along the lines of the list offered here:

http://kazemakase.ca/2013/08/03/remembering-the-kanji-vol-1-with-japanese-keywords/

Both of these lists have their own issues, and no solution will please everyone, but one priority in making the KSP list will be to avoid duplicates, something not quite achieved with wrightak’s list. Matching the Japanese keyword to the English keyword will not be a priority, but KSP will note where the meanings diverge.

KSP will still be recommending that beginners start with English keywords, so that the task of assigning meaning to the visual representation of the kanji is as easy as possible, but the Android app will offer the option of switching to a Japanese prompt when the kanji is known at a specific, user-chosen level. For instance, kanji known at level 5 or higher could automatically switch to a Japanese prompt. The first time the kanji is shown after the switch, the English keyword will also be offered, but subsequently the English keyword will be suppressed. (Tapping the Japanese prompt will display a translation as well as the English keyword.)

From Heisig to Vocab – A Test of Keywords

I have just added a vocab list consisting of 669 entries, which has been recommended for the introductory level of the Japanese Language Proficiency Test (JLPT).

Over the next couple of weeks, I will be scoring each of the entries according to how well the English keyword associated with each kanji would prepare a beginner for the meaning of the actual Japanese word using that kanji. Scores will range from 0/5 (the keyword is no help at all) to 5/5 (a student armed with nothing but the keyword would guess the correct meaning). An example of a 5/5 entry is ‘‘, meaning ‘blue’, which is exactly what the keyword would lead us to expect. An example of a 4/5 entry is ‘‘, meaning ‘afterwards’, which is related to the keyword ‘behind’ but not quite the same.

I’ll publish the results here, for people who are wondering how well a Heisig-style approach will prepare them for learning real vocabulary.