Accurately identifying variant forms of kanji components

[karl rosvold]

As we seem to be on a roll with tidying up the non-visible data
contained in the KanjiVG files, I also propose that we accurately
identify which components are the standard forms and which are
variants.

For example, the two characters: 樸(6a38) and 査(67fb) both contain the
component for tree (木).
However, I would say that the final stroke of the tree element in 樸 is
shorter, and therefore "ki-hen"(tree when it appears on the left) is
a variant of "ki" (tree). The character for 樸 rendered from the SVG
file looks fine, but there is no mention that this is a "variant" in
the SVG file. I would think that if we used something like IDSgrep or
an equivalent algorithm, we could find when 木 appears on the left
(including when it is embedded in the character such as the second
tree in 森 for example). Since this is KanjiVG we are talking about
hopefully someone could even make a color-coded font, for example,
making the components not labeled as variants as black, and variants
in red. That would allow for quick scanning to find any components
that haven't been labeled correctly.

By the way, does KanjiVG distinguish between multiple variants? Or is
everything that is not the canonical form simply considered a
"variant"? For example, 手. I'd say 手(624B) is the standard form, as
is the 手 when it appears on the bottom as in: 挙(6319). The first three
strokes of 指 (6307) are one variant, and the first four strokes of
看(770B) are another (the last stroke being different from the last
stroke of the canonical form.

Karl

[Alexandre Courbot]

On Feb 27, 2012 10:32 PM, "karl rosvold" <karl.r...@gmail.com> wrote:
> hopefully someone could even make a color-coded font, for example,
> making the components not labeled as variants as black, and variants
> in red. That would allow for quick scanning to find any components
> that haven't been labeled correctly.

That should actually also be feasible as a web page, with a bit of Javascript. Yet another good idea to realize... We really need another coder for this project.

> By the way, does KanjiVG distinguish between multiple variants? Or is
> everything that is not the canonical form simply considered a
> "variant"? For example, 手. I'd say 手(624B)  is the standard form, as
> is the 手 when it appears on the bottom as in: 挙(6319). The first three
> strokes of 指 (6307) are one variant, and the first four strokes of
> 看(770B) are another (the last stroke being different from the last
> stroke of the canonical form.

I think the "original" attribute can be used to distinguish between variants, provided they have a codepoint associated.

Alex.