How do you determine the radicals or sub-components of the kanji?

[Raeldor]

Because for example the kanji...

誨

Tries to use the radical/kanji for 'every' (which has 6 strokes, not
7) to describe the right hand side, instead of using the kanji/radical
for person and mother.

This data is very useful, but might be better auto-generated by
another means. I have written software to do stroke comparison and I
am thinking to have it work through this database to try and match the
strokes to radicals/other kanji by comparing the strokes instead of
using the data in the XML file. If it turns out to be more accurate I
would be more than happy to share the result, but I'd like to know how
you arrived at the current data set first.

Thanks
Ray

[Alexandre Courbot]

On Thu, Jul 14, 2011 at 9:22 AM, Raeldor <rae...@gmail.com> wrote:
> Because for example the kanji...
>
> 誨
>
> Tries to use the radical/kanji for 'every' (which has 6 strokes, not
> 7) to describe the right hand side, instead of using the kanji/radical
> for person and mother.

I do not know much about this (Ulrich is the man for this kind of
stuff) but AFAIK radical information is not chosen according to
graphical ressemblance, but rather according to ethymology/historical
matters. That's why you can find such things.

> This data is very useful, but might be better auto-generated by
> another means.  I have written software to do stroke comparison and I
> am thinking to have it work through this database to try and match the
> strokes to radicals/other kanji by comparing the strokes instead of
> using the data in the XML file.  If it turns out to be more accurate I
> would be more than happy to share the result, but I'd like to know how
> you arrived at the current data set first.

I am not sure if this is what we want to have - or maybe this could be
added to the existing data as a new attribute. Would like to hear what
Ulrich things about this, as I prefer to have him decide for this
stuff.

Alex.

[Torrential]

Raeldor <rael...@gmail.com> wrote:
> > Because for example the kanji...
>
> > 誨
>
> > Tries to use the radical/kanji for 'every' (which has 6 strokes, not
> > 7) to describe the right hand side, instead of using the kanji/radical
> > for person and mother.
>

Hi Raeldor,
I am no expert, but I have been looking at different fonts recently as
I introduce radicals to my own program, the Kanji Sketch Pad, and the
radical-combination you are referring to, which is also used in 'sea',
is one that shows major variation between fonts. I first noticed it
with the kanji for 'sea'... Some fonts use the horizontal 'person' and
then 'mother', some use something that looks like 'every'. I suggest
you write a script that prints each kanji in different fonts, before
locking into one particular system, so you get a sense of the huge
variations. The other alternative is to choose one font to work with
and accept that it will contradict expectations based on other fonts.
The vectors used in kanjivg do not closely follow any single common
font, as was discussed in a thread a few weeks ago. (I'll look for he
link),
Cheers,
Craig.

[Torrential]

Link to font discussion...

http://groups.google.com/group/kanjivg/browse_thread/thread/1784a2c8fc36f6d9/086d18fae4bd5e42?lnk=gst&q=font+sea#086d18fae4bd5e42