With 777 Kanji, 90% Coverage of Kanji in the Wild | Hacker News Reader

With 777 Kanji, 90% Coverage of Kanji in the Wild

japanesecomplete.com

54 pointssova29 comments

This is a nice site, but it's worth mentioning what exactly 90% coverage entails in practice. It's more painful than people might realize.

https://www.sinosplice.com/life/archives/2016/08/25/what-80-...

PERFECT. Exactly what it felt like trying to learn Japanese in my 30s going to class only 1-2x/week and having no native speakers to converse with daily. Being sort-of proficient or knowing XX% of kanji/vocab/etc sounds a lot more impressive than it is in practice!

At first I made a lot of progress. Eventually I got into the middle where every lesson taught me both some new small thing and opened an entire dimensions of other things I didn’t know.

I eventually settled on being a functionally illiterate holiday traveler with the vocabulary of a kindergartener.

I’m never going to work there, so this is as far as I’m going to go..

I have a few Japanese friends on social media.. When I travel to Japan, and especially in a situation where not in a hurry, I can entertain a taxi driver or make small talk with some older folk in a quieter town I am passing through.

In a cafe / pub in a city I can make a quick order in Japanese and if questioned for more detail I have to switch to English pretty quickly.

It’s a very humbling experience and everyone should try it. It gives you even greater respect for immigrants who come here and learn English... frequently speaking it better than native born!

Great article, thanks for sharing. I don't know if there's a technical concept/name for this, but I call it "legibility threshold" -- the threshold at which you are highly likely to be able to intuit the words/ideas you don't know, by the context. I think it's a valuable concept as it pertains to joining a new company/problem domain -- there are so many acronyms/words you don't know, and it takes time to hit the "legibility threshold" where everything starts to click.

I call it the "launchpad" because I can finally leave my native language behind and learn from consuming media directly.

Exactly. Just think about how bad OCR is when it gets 90% of the words right! With a 90% word accuracy rate, you have roughly between 0.5 and 1 error per sentence in a typical text. That's not great. The only reason we can interpret a text that's been OCR'ed at 90% accuracy is that humans are good at filling in what the right word should be in the face of a spelling mistake. If it put in random, nonsense words, we'd just struggle and say the OCR software was garbage.

The previous paragraph is 91 words. Imagine reading it with 9 random nonsense words in place of real English words.

Yeah, exactly. I passed N1 which requires ~2000 kanji, but I still frequently encounter words with kanji I haven’t encountered/don’t know the readings of.

A lot of meaning making is from understanding context. If you are familiar with a situation you can have near perfect comprehension where you only understand completely 80 of the words.

Thank you for that, that was a thoroughly enjoyable read.

This is remarkably useless, and the premise is misguided. You can't just "learn a kanji", because the common ones have many readings and rarely stand alone: you need to learn them in context as compounds, where the meanings vary wildly. Here's just one: https://en.wiktionary.org/wiki/%E7%94%9F

As a rough analogy, imagine a site that proclaimed you can learn English simply by memorizing these 777 Greek roots: astron, grapheia, logia, nomos, tele, phone, kryptos etc. Useful, yes; sufficient, hell no.

I agree. In my experience at least, learning high-frequency vocabulary first can indeed speed up the acquisition of second-language literacy. In the case of Japanese, though, individual kanji function mainly as subcomponents of vocabulary, not as entire words.

If you are still getting used to recognizing individual kanji—if they still look mostly like indistinguishable squiggles—it might not hurt to practice them in order of frequency. But if your ultimate goal is functional literacy in Japanese, it would be better to spend more time learning high-frequency words—nouns, verbs, adjectives, adverbs, etc.—and how they are written, whether that be in hiragana, katakana, kanji, or some combination.

    100 kanji - basic signs
    500 kanji - 50% of printed material
    1000 kanji - 85% of printed material
    1945 kanji - 97% of printed material
    8000 kanji - 100% of anything japanese kanji in electronic form (still only about 99.99% in printed material as a lot of ancient texts have some pretty obscure one time only kanji).

    You can judge your kanji learning speed through the first 100 kanji benchmark. It takes one unit of time to learn the first 100 kanji, then about 14-30 units of time to learn the 800-1000, and then another 7-15 units of time to finish off the joyou list (and name kanji). So if it took you 1 month to learn your first 100 kanji it should take you another year to 3 years to learn your next 1000, and then another half a year to a year and a half to finish the list. Again, faster learners will find themselves closer to the earlier stages of the spectrum, while casual learners and slower learners towards the further end of the spectrum.

http://yesjapan.com/YJ6/question/1394/how-many-kanji-does-on...

I don’t know any Japanese but for Chinese, metrics like this are misleading, because the long tail of unknown characters are still enough to prevent you reading. If you know 80% of chinese characters by frequency you can’t read 80% of the newspaper, it’s probably closer to 10%. In other words, comprehension does not scale linearly with number of characters recognized, it’s more like a step function.

I don't understand - what do I do with this list? I can't click them, there's no definitions, there's no sign up button

Here's a MUCH better page:

https://kanjicards.org/kanji-list-by-freq.html

Much more complete and you can actually click on each kanji in order to read about it. Plus they're not pictures but actual UTF-8 characters so they can be zoomed into without aliasing and can be copied and pasted. Only thing that seems to be missing is a search feature.

It says something about the lengths we have been trained when we simply don’t know what to do in lack of a signup button

(It’s what it says; the list of characters. Those who find this useful presumably have a dictionary already. Or you can use eg jisho.org)

Scroll to the bottom and pay for the privilege.

Oh, it's just an advertisement.

Most of OP's submissions are also ads for "japanesecomplete.com."

It is worth noting that most of this posters submissions have been flagged as spam, as it is usually posts about this site.

As an addendum, this particular topic seems to have been posted 5 times in the last two years under slightly varying URLs as well: https://hn.algolia.com/?q=777+kanji

Is there a similar thing for the Anki or Memrise? Just the most popular words, unrelated to the JLPT levels.

But you need to have at least 2k to read anything serious.

will learning this list of kanji help with mandarin acquisition as well? Do the characters mean the same in Mandarin?

It'll help, but less than you'd think. In China many characters have been simplified and merged so they look quite different, eg 麺 noodles combined into 面 face, and meanings have diverged as well. A few famous examples include 手紙 letter vs toilet paper and 恋人 lover vs wife.

I would argue that it helps with more than the other comments suggest.

The difference in writing (which I don't think is simply a matter of simplification when comparing Kanji and Simplified Chinese), is typically not a problem because of the ways characters are made up on Chinese.

Consider (japanese/chinese) 記/记, 設/设, 語/语, 計/计 and 説/说, by learning that single simplification of the 言 part you already recognise a good amount of characters in Mandarin (there are many more of those in the list). In fact, in many cases a learned reader can just guess the meaning of the other without having to learn them in the first place.

Many of the meaning of the single characters are the same/similar between the two languages. Even in cases where meanings have diverged between Japanese and Chinese, having and understanding of the single characters that make up more complex phrases still helps a lot in terms of guessing meaning and context.

Also, 恋人 doesn't necessarily, and usually doesn't, mean wife/husband in Chinese, its meaning is very similar to the one in Japanese.

Edit: Forgot to mention that if you are interested in seeing the difference between Japanese and Chinese for the 777 characters, simply copy-and-pasting the list into Google Translate as Simplified Chinese does the magic. Running them through a `for` loop shows that there are 184 characters that are not the same between the two languages; I estimate that there are roughly 40% of them that's on the level of 麺/面.

As someone who has made the mistake of studying both at the same time, it is far easier to focus on just one or the other until it gets to a decent level.

It sounds like it might be useful, but in practice there's so much to do that whatever advantage it gives you is a blip, and the similarities tend to become hindrances when you are in the learning phase. Someone who is already a native speaker of either will get far more out of it than you will as there will be no danger of crossing wires. The only really significant advantage is that it familiarizes you with learning a character-based script in the first place.

Mostly yes.

Japanese people can vaguely infer the meaning of mandarin, given the simplified characters, without prior exposure to the language before:

https://www.youtube.com/watch?v=rzJqXd-1dEU

I found it interesting that Japanese people did better with traditional Chinese characters than with simplified Chinese. This makes sense because while some Japanese characters are simplified forms of traditional Chinese characters, they are often simplified very differently than the simplified Chinese forms. Japanese people might encounter the traditional characters in old books, but are unlikely to learn the simplified Chinese form unless they specifically studied Chinese.

Many of the most common Chinese characters used in Mandarin aren’t really used in Japanese, so especially for the most common words there isn’t that much overlap.

The ideograms are there but their meanings are not :/

Reply on news.ycombinator.com