Return-Path: <jimc@math.ucla.edu>
	(Sendmail 5.61/1.07) id AA06569; Mon, 1 Mar 93 08:11:54 -0800
Message-Id: <9303011611.AA06569@julia.math.ucla.edu>
Cc: lojbab@grebyn.com
Subject: Re: TECH: Lujvo Place structure paper 
             <9303010206.AA26177@julia.math.ucla.edu> 
Date: Mon, 01 Mar 93 08:11:54 -0800
From: jimc@math.ucla.edu
X-Mozilla-Status: 0011

Procrustes is offended :-)  But besides some disagreement about pounding
square pigeons into round holes, I think it's a very good introduction.

Here's a modest proposal:  Devise a set of rules for lujvo place
generation.  Go through all extant lujvo and identify those whose
meaning or utility  would be markedly degraded if they were forced to
follow the rules; assign a "badness score" so as to distinguish those
which might be improved a little bit by jiggering the places, from
those which are useless without noncompliant places.  Perhaps adjust
the rules to fit the lujvo being crammed into them, or re-order the
places of gismu that are "obviously wrong".  Are there really many
disasters remaining?  Maybe some of them have the wrong root.  Maybe
some of them really should be gismu.  

This is what I did to generate the first version of my dikyjvo rules. 
With the vocabulary of that time it was possible to get near 100%
compliance with no gismu changes, and a number of patterns became very
clear (such as the X2 for garments specifying what body part it is worn
on, now being discussed again).  If you sort the gismu by root word
(which I was not able to do, technologically), there will be many
blocks to which a uniform rule applies, greatly reducing the effort of
judgement and the clarity of pattern detection.  

		-- jimc