Re: sumti categories

[jimc@MATH.UCLA.EDU]

Jorge Llambias <jorge@PHYAST.PITT.EDU> writes:
> Let's see... who disgrees with this...

> (2) Each slot of every selbri can be filled with sumti from only
>     one of those categories.

> This restriction is not grammatical, because the grammar allows
> any sumti to fill any slot, but rather, it's a semantic restriction.

Aha, the hydra of dikyjvo raises its fearsome head yet again!  A key
point in my conception of dikyjvo was that one could in fact alter the
definitions of "all" gismu so that their arguments were semantically
restricted in just about the ways Jorge proposes: they were divided
among "things", "events" and "texts".  "Numbers" were assumed to be a
special case of "things" but they did clearly have a disjoint set of
argument places.

Then one could unambiguously apply various transformation rules to a
dikyjvo to extract a canonical form containing only gismu (and
structure words), according to the pattern of arguments of the
component gismu.  For example, in one case an abstraction appears
because the main gismu "expects" one, while a regular sumti pops up for
a transitive gismu expecting regular sumti.  When the definitions are
radically hacked, the resulting patterns work almost by mental
telepathy to deliver accurate and credible meanings, with good coverage
of what one might want to say, and very few instances where
circumlocutions are needed to evade an unwanted interpretation.

Unfortunately, at the time Lojbab was moving the Lojban definitions in
another direction, and the distinction between semantic categories was
already blurred and became more so.  So the answer to Jorge's question
is, yes, there is a core of semantic regularity to argument places, but
no, it is not regular enough to support higher level constructs like
dikyjvo.  Should this situation change (thereby delaying the dictionary
substantially)?  Let's hear some clamor...

James F. Carter        Voice 310 825 2897       FAX 310 206 6673
UCLA-Mathnet;  6221 MSA; 405 Hilgard Ave.; Los Angeles, CA, USA  90024-1555
Internet: jimc@math.ucla.edu            BITNET: jimc%math.ucla.edu@INTERBIT
UUCP:...!{ucsd,ames,ncar,gatech,purdue,rutgers,decvax,uunet}!math.ucla.edu!jimc