Re: [edict-jmdict] Re: Yomigata for Edict and Nedict

[Jim Rose <jim@*************>]

On Aug 17, 2007, at 12:10 PM, Paul Blay wrote:
> > > You seem a little confused as to whether you are asking or
> > > telling. ;-)
> >
> > Socratic method not taught in Britain anymore? (^_^)
>
> It caught on more in America.

All the more reason to allow us back into the Empire, and free us  
from the inestimable folly of our current government.  How does dual  
citizenship for any American who can trace his roots to 1776 or  
further sound?  It could be the start of eventual reunification, and  
we could pump some Socratic method back into the lifeblood of Britain.

> > > I shall give one reason for including an 'irregular' tag -  
> feedback to
> > > KanjiDict. The theory is that regular readings of kanji are  
> those that
> > > are given in KanjiDict for that kanji or are formed from regular
> > > euphonic changes from same. Thus all words using those kanji  
> should
> > > either have regular readings or should have readings  
> _acknowledged_
> > > to be irregular. It is quite likely that the process of assigning
> > > 'irregular' tags will thus end up plugging gaps in KanjiDict  
> readings.
> >
> > Interesting idea.
> >
> > But will there be real value when applied to idiomatic reading
> > compounds?
>
> What I'm doing with the Example sentences includes an analogous
> process. By eliminating (by properly indexing) all the known words
> with kanji/kana representations in Edict whatever is left is either
> not a word or a word that _should_ be in Edict but isn't. Quite a
> few of my submissions to Edict start from that point.

A powerful analogous example.

> > After you filter out 'regular' euphony as in 時々
> > [ときどき], numerals（２００７）, symbols（◎、 
> 〒）,
> > and kana kanji substitutes（ケ）, you're left basically,  
> perhaps
> > entirely, only with compounds which are not marriages of either
> > 'established' kun-kun or on-on readings (i.e., idiomatic readings),
> > which only appear to be irregular if you make the assumption that a
> > word's sound is always constructed from the "regular" sounds of its
> > parts.
>
> I think you're squashing too many arguments into one sentence
> there. Ignoring your aside about whether idiomatic readings are or
> aren't irregular you seem to be saying that "they aren't there so
> you won't find them".
>
> That doesn't seem to be a very scientific method to me. I have seen
> people (well, mostly 'person') submit new readings to add to KanjiDict
> thus it is proven that KanjiDict was not completely perfect. There is
> no reason to assume that the last such submission was the very last
> omission from kanji readings. It is also the case that some of those
> submissions were prompted by Edict entries with kanji that had  
> readings
> not represented in KanjiDict. It follows that a trawl through all
> 'suspected idiomatic/irregular' readings found on examination of Edict
> entries is entirely likely to prompt more such additions.


Then I have to break with Internet tradition and agree with you.   
You're right.  Good idea.  But should it be in the public version of  
JMDICT or just a development version.