Check radicals, strokes against KanjiDic/Unihan

[benkasminbullock]

For all characters, check whether a radical exists in the data.
Where KanjiDic has an entry for a character:

• Check that the radical in KanjiVG exists for each character
• Check that the Nelson/Traditional/General radicals are the same as those recorded in Kanjidic
• Check the stroke counts are the same
  Where KanjiDic doesn't have an entry, check against the Unihan entry (this doesn't contain Nelson information).

[benkasminbullock]

Report so far:

The element 白 is completely missing on the group of radicals for many, perhaps all cases. For example 皖 has the radical group correct but without any element.

Some radicals in KanjiVG differ from both Kanjidic and Wiktionary, for example:

• 鵈 耳 Kanjidic/Wiktionary have 鳥
• 辧 刀 / 辛
• 舊 艹 Kanjidic/Wiktionary have 臼
• 隲 is 馬 / 阝
• 杰 is 灬 / 木
• 蘯 is 皿 / 艸
• 辯 is 言 / 辛
• 着 is 羊 / 目
• 斈 is 子 / 文
• 戝 is 貝 / 戈
• 巨 is 二 / 工

Most of the above seem most likely to be errors in KanjiVG, they don't seem to obey the rule of left/top first and the obvious radical is skipped over.

On the other hand, some of Kanjidic's radicals don't agree with Wiktionary and KanjiVG:

• 竜 is 龍 in Wiktionary and KanjiVG but 立 in Kanjidic
• 歯 is 歯 / 止

Some but not all cases with element 月 are missing a kvg:original field of 肉

The element 日 is used for radical 73, but where this is used kvg:original field of 曰 should be added too.

[benkasminbullock]

I'm getting a bit stuck on some of the Nelson radicals. For example,

0557b-Kaisho: 啻: tradit but no nelson: 30 7 30
0557b: 啻: tradit but no nelson: 30 7 30

The Nelson radical is supposedly 7, but that is ニ, which doesn't appear in any form in the character. I might have to borrow Nelson from the library or something to check the rest of these. Lots more weirdness like this:

05f6b: 彫: tradit but no nelson: 59 4 59

The tradit radical is 59 which is 彡, which makes sense, but Nelson is 4 which is 丿, which doesn't really make much sense to me, since characters are full of these shapes of strokes. Another possibility is that, since these are the values in Kanjidic, the cause might be errors in Kanjidic.

[benkasminbullock]

The nub of things which I'm stuck on from KanjiVG/Kanjidic is

0557b-Kaisho: 啻: tradit but no nelson: 30 7 30
0557b: 啻: tradit but no nelson: 30 7 30
058f9: 壹: tradit but no nelson: 33 32 33
05f6b: 彫: tradit but no nelson: 59 4 59
07099: 炙: tradit but no nelson: 86 4 86
0723d: 爽: tradit but no nelson: 89 4 89

The top case is the ニ (7) mentioned above where it does not exist in the character. The second case is a difference between 土 (32) and 士 (33) which makes no sense to me since the top part is visually clearly 士. The next three cases are where Nelson seems to use a dash (4), but since the Nelson radical is not indicated in the file it's not clear which of the various strokes should be the "dash". Assuming that Kanjidic is correct, Nelson's choice of using a dash as a radical is a very dubious one.

[benkasminbullock]

Most of the work is done so I'll close this and leave it for later reference.