+- kanji koohii FORUM (http://forum.koohii.com)
+-- Forum: Learning Japanese (http://forum.koohii.com/forum-4.html)
+--- Forum: Remembering the Kanji (http://forum.koohii.com/forum-7.html)
+--- Thread: Onyomi goals (/thread-11002.html)
Onyomi goals - ktcgx - 2013-08-01
So, after 2 weeks, I'm still averaging 70 frames a day. I just did frame 1401, and I'll try to get through about another 100 today because I'll be travelling Sunday-Wednesday, then Thurs and Fri will be busy days, so I'm a bit worried about how many kanji I will actually be able to get through in the next week.
Onyomi goals - ktcgx - 2013-08-12
Well, after about a week's break, while trying to still get through about 100 reviews a day, I got back into the Kanji yesterday, but only got through only about 50 kanji. I have over 900 cards due for review today T_T... Still, I think I'll reach my goal. I'm on frame 1530, and have another 10 days to go.
Onyomi goals - ktcgx - 2013-08-22
Another update:
Well, after the week's break I took, it's been very hard to get back into it. It's been very hot and I haven't had the motivation.
After today, there are 3 more days til the summer holidays are over, and school starts again. I'm on frame 1866, so I think it's pretty safe to say I won't make it. However, even so, by pushing myself hard at 70 a day for those first 3 weeks, I've made far better progress than I would have otherwise, and even though I will fail my goal, I feel like there's still success I can take from this, because I came so close, and I know so many onyomis already now, that I didn't before.
Onyomi goals - Vempele - 2013-08-22
How were the non-pure groups?
Does Heisig ever explain why you're not supposed to review kanji->reading directly, apart from the issue of multiple readings? I've been making (randomized kanji)->reading cards for pure groups where the signaled reading is the main on-yomi (= the only on-yomi not in parentheses on Wikipedia's list) for all the kanji in question.
Onyomi goals - ktcgx - 2013-08-22
Vempele Wrote:How were the non-pure groups?It's harder to remember them when there are a lot of exceptions to the signal primitive, but I think anki will take care of those eventually.
Does Heisig ever explain why you're not supposed to review kanji->reading directly, apart from the issue of multiple readings? I've been making (randomized kanji)->reading cards for pure groups where the signaled reading is the main on-yomi (= the only on-yomi not in parentheses on Wikipedia's list) for all the kanji in question.
I think that reviewing kanji compound -> reading is better in the long run, because you get a tonne of vocab at the same time as doing the readings. It seems like 90-95% of kanji have only 1 onyomi, so it doesn't make a difference, and I think it's easier to learn them in words, because you feel like you're improving a lot, so it helps you keep motivated.
Onyomi goals - mmhorii - 2013-08-22
Here's a quantitative analysis of the power of phonetic components in kanji.
http://namakajiri.net/nikki/testing-the-power-of-phonetic-components-in-japanese-kanji/
As the author notes, it gets even better when expanding beyond the non-jouyou world of kanji, and phonetic components tend to be used more with rarer kanji.
Onyomi goals - Vempele - 2013-08-25
Thanks, that's a great link!
I've been trying to improve on the results. I don't have much yet, but here's a new pure group:
夭 妖 殀 沃 飫 ヨウ
Onyomi goals - mmhorii - 2013-08-25
@Vempele:
Glad you like the link.
I just wanted to mention that near the end of the author's post, there are download links for several tsv files, including one called components_phonetic.kanjivg.tsv. This file has the group for 夭 that's identical to the one you listed (in the file, search for 夭 with reading ヨウ and look in the fifth column, Kanji in phonetic series). I think the author already did what you're trying to do, using the KanjiVG database.
The link to the file is here:
http://namakajiri.net/data/kanji/components_phonetic.kanjivg.tsv
Onyomi goals - Vempele - 2013-08-25
I know. I used the tsv files as a starting point. I got that group's kanji coverage up from 29% to 71% by excluding the pure groups 喬 and 忝, at which point the only exceptions left were 呑 and 笑.
Onyomi goals - mmhorii - 2013-08-25
Ok, now I see what you're trying to do--I didn't understand at first. If you can post your improved results and/or script when you finish, that would be excellent.
Onyomi goals - Vempele - 2013-08-26
The following groups are perfect unless otherwise noted:
千+灬: 勲 勳 熏 燻 薫 醺 クン - 熏 plus 薫 and 勲.
先-贊: 先 洗 濳 筅 跣 銑 セン - In RTK2. -贊 removes 讚 贊 鑽 サン, a perfect group.
戍+幺: 幾 機 畿 磯 譏 饑 キ - This is really just the 幾 group with 畿 added.
見+臣: 攬 欖 纜 覧 覽 ラン - The 覽 group plus 覧 纜.
兩-廿: 倆 兩 裲 輛 魎 リョウ
兩+廿: 懣 滿 瞞 蹣 マン - Only a pure group.
筑: 筑築 チク
凡+工-筑: 恐 蛩 跫 鞏 キョウ - The top part of 恐 on top of anything.
巣: 剿勦巣樔 ソウ - Only a pure group
巛+田-巣: 緇 輜 錙 鯔 シ - 巛+田 is the right side of 鯔.
舌+氵: 活 濶 闊 カツ - The 活 group plus 濶.
敢-厂: 敢 橄 瞰 カン - Most kanji containing 敢 put it under a cliff:
敢+山: 巌 巖 ガン
敢+厂-山 : 儼 厳 嚴 - Only a pure group.
祭-宀: 祭 蔡 際 サイ - -宀 removes the perfect group 察擦 サツ. Both groups are in RTK2.
升-飛: 升 昇 陞 ショウ - Not quite a perfect group, as there's also 枡 (no on-yomi).
呈-士: 呈 程 逞 テイ テイ - -士 removes 鐵 and two kanji that distort 王 to 壬.
育+乂: 徹 撤 轍 テツ - In RTK2.
万+厂: 励 砺 蛎 レイ - In RTK3.
壬-廷-王-?+⺤: 婬 淫 霪 イン - Just a really complicated way of saying "the right side of 婬". The ? is probably 丿 (⺤ was originally also a ?).
瞿-矍: 懼 瞿 衢 ク - 矍 is a pure group 攫 矍 钁 カク
左-月: 佐 左 サ - Only a pure group. Listed as semi-pure in RTK2 with 惰 as the exception.
左+月+辶: 膸 隨 髓 ズイ - In RTK2.
左+月-辶: 墮 惰 楕 橢 隋 ダ - Only a pure group.
Pure groups:
可-奇-阿-竒: 何 可 呵 哥 彁 柯 歌 河 渮 珂 舸 苛 荷 訶 謌 軻 カ - 奇 is nearly pure for キ, 阿 is pure for ア and 竒 is perfect for キ (just two kanji).
少+貝: 嬪 擯 檳 殯 濱 瀕 繽 蘋 賓 頻 顰 鬢 ヒン - A pure group 賓 and a perfect group 頻 walk into a bar. Out walks a pure group.
我-羲-羊: 俄 哦 娥 峨 峩 我 莪 蛾 餓 鵝 鵞 ガ - Almost perfect. In RTK2.
工+巛: 剄 勁 徑 痙 經 脛 莖 輕 逕 頸 ケイ - Heisig notes in RTK3 that this is the old form of 圣, which is a semi-pure group for the same reading (confirmed: 怪 remains the only exception).
林+厂-广: 暦 櫪 歴 瀝 癧 轣 靂 レキ - In RTK2. The real purpose of -广 is removing the 麻 マ group (which has 4 exceptions now, 糜縻靡麾).
缶-丿+月: 徭 搖 瑤 窰 謠 遙 鷂 ヨウ - Almost perfect. Old form of 䍃: 揺 謡, a perfect group with the same reading.
回-亠-嗇: 回 廻 徊 茴 蛔 迴 カイ - Four of these also share the on-yomi エ: 回 廻 徊 迴.
屮+欠: 厥 獗 蕨 蹶 闕 ケツ - The 厥 group plus 闕.
圭+厂: 啀 崕 崖 涯 睚 ガイ
氏-民-昏-?-一: 帋 氏 祇 紙 舐 シ - Whatever the ? is, I'm fairly sure it's redundant.
氏+一(-民): 低 底 抵 柢 牴 砥 羝 觝 詆 邸 テイ - Exceptions: 岻祗胝鴟.
民: 岷 愍 民 泯 眠 緡 罠 ミン - Exception: 氓.
民: 岷 愍 泯 緡 罠 ビン - Exceptions: 民氓眠
昏: 婚 昏 棔 コン - Perfect group.
缶+勹: 掏 淘 綯 萄 陶 トウ - Almost perfect.
延-口: 延 涎 筵 莚 エン - "Almost" perfect. 口 removes 蜑蜒誕.
叉-蚤: 叉 扠 釵 靫 サ
蚤: 蚤騷 ソウ - Perfect.
則-厂: 側 則 惻 測 ソク - Almost perfect.
則+厂: 厠 廁 ショク, シ - Perfect.
武-斌: 武 賦 錻 鵡 ブ - 斌贇 is a non-phonetic group.
必+宀: 密 樒 櫁 蜜 ミツ - Almost perfect. In RTK2.
厄-危: 厄 扼 軛 阨 ヤク - All but 厄 also have the reading アク.
萬+厂: 勵 礪 糲 蠣 レイ - Almost perfect.
隶-康-示: 棣 逮 隶 靆 タイ - Almost perfect.
康: 康 慷 糠 鱇 コウ- Perfect.
隶+示: 隷 隸 レイ - Perfect.
臼+⺤: 滔 稻 蹈 韜 トウ - Almost perfect.
臣+戈-月: 臧 蔵 藏 贓 ソウ - The next group adds the missing readings.
臣+戈: 臓 臟 臧 蔵 藏 ゾウ - Exception: 贓.
That's not all, but I think I'll go back to improving the algorithm.
Onyomi goals - mmhorii - 2013-08-26
This is awesome; looking forward to seeing more!
Onyomi goals - Vempele - 2013-09-01
I need a better readings dictionary. Kanjidic even gets some Jouyou kanji wrong. Those, I can overwrite with data from the Jouyou list no problem, but the rest...
Onyomi goals - Vempele - 2013-09-05
https://www.dropbox.com/s/l46sgk1kh70shdf/phonetics.zip
182 groups (some of which are redundant)
429 perfect Jouyou kanji
593 pure Jouyou kanji (including the perfect ones)
949 matching kanji in the aforementioned groups (excluding exceptions)
119 exceptions.
1379 pure kanji in total.
4422 kanji (originally 6355) with 5484 readings, used 295564 times.
9177 readings in Kanjidic.
12402 readings in Kanjigen.
Readings not found in JMDict have been trimmed unless they appear on the Jouyou list.
phonetics.tsv is sorted by a rather complex set of criteria, but basically primitives that predict a lot of common (=non-parenthesized on http://ja.wikipedia.org/wiki/常用漢字一覧) readings of Jouyou kanji are preferred.
The sorting of best_phonetics.tsv is the same.
un_phonetic.tsv lists the 1311 kanji whose readings could not be predicted with >9% kanji coverage.
reading_counts.tsv lists the number of times each on'yomi appears in JMDict. 物 モチ is the most irregular reading, appearing in only one word out of 1151.
Thanks to mmhorii and toshiromiballza for the links and to netsplitter for making ryuujouji. And to the authors of the article, Kanjidic, Kanjigen (Gakken's Kanjigen), KanjiVG and JMDict.
Edit: Oops, the names of the columns in phonetics.tsv were messed up. Fixed. Also, typos in this post.
Onyomi goals - mmhorii - 2013-09-05
ドカーーーーン!!!!!
You just blew my mind. I love it.
Onyomi goals - ktcgx - 2013-09-05
What is the difference between perfect and pure groups?
Onyomi goals - mmhorii - 2013-09-05
For perfect groups, whenever the given phonetic component is present, the reading can be predicted with absolute certainty.
For pure groups, whenever the phonetic component is present, one possible reading can be predicted, but there may be other readings.
Onyomi goals - ktcgx - 2013-09-05
Thanks!
Onyomi goals - Vempele - 2013-09-06
With the caveat that some of the groups might not be pure/perfect with a bigger dictionary. Also, the groups of pure/perfect Jouyou kanji I mentioned earlier are only pure/perfect relative to Jouyou (though a majority of them are at least pure in general), hence 119 exceptions.
Readings coverage doesn't take exception kanji into account. Kanji coverage doesn't take kanji with no on'yomi into account (because those kanji are treated as if they don't exist). "Jouyou kanji that lack the predicted reading" can also contain kanji that actually do have said reading if the reading is uncommon and the kanji has at least one common reading, for example 情 (ジョウ is common) for the 青 セイ group.
Onyomi goals - toshiromiballza - 2013-09-06
Vempele Wrote:"Jouyou kanji that lack the predicted reading" can also contain kanji that actually do have said reading if the reading is uncommon and the kanji has at least one common readingI think it would be better if you didn't exclude those kanji, as long as the uncommon reading is an official Jouyou reading, no matter if it's rare or not.
Edit: Ah, I see.
Onyomi goals - Vempele - 2013-09-06
It only counts against them for sorting purposes. They're still included in the list of matching kanji.
Onyomi goals - DrJones - 2013-09-06
Great list! I'm inserting it right now into Anki using as a base this other list https://ankiweb.net/shared/info/2079428463
By the way, instead of 小+大 you could use 尞 to represent Ryou リョウ, 幾 instead of 幺+戍 to represent Ki キ, and 菫 instead of 三+艹 to represent Kin キン.
EDIT: You can also use 髟 instead of 厶 in 長-厶 and 夆 instead of 三+夂.
Onyomi goals - Vempele - 2013-09-06
DrJones Wrote:By the way, instead of 小+大 you could use 尞 to represent Ryou リョウ, and 幾 instead of 幺+戍 to represent Ki キ.幾 misses 畿 and 尞 is only recognized as a component of 尞療. There's also 侯 for 矢+亻(misses 候 but gets rid of 雉).
As for 三+艹, it's actually two components: 菫 and the one in 僅.
Onyomi goals - DrJones - 2013-09-06
Heh, as I'm using a chinese font I see both as the traditional form and can't see the differences.
Also, I found out that your 居 group is missing the kanji 裾 (2624 in Heisig order) which is also read as キョ/コ.
Onyomi goals - toshiromiballza - 2013-09-06
DrJones Wrote:Also, I found out that your 居 group is missing the kanji 裾 (2624 in Heisig order) which is also read as キョ/コ.Yeah, but the only official reading of 裾 is すそ, and there doesn't seem to be any word in JMdict that uses キョ/コ.
Although, including it would make the 居 group an imperfect jouyou group, which would be more accurate, I guess.
Vempele Wrote:Kanji coverage doesn't take kanji with no on'yomi into account (because those kanji are treated as if they don't exist).Maybe it would be better if you re-did the algorithm and take jouyou kanji with no onyomi into account this time (but not non-jouyou kanji with no onyomi).
How many "perfect" groups would there be then?