Message-Id: <m0qQl7w-000023C@xiron.pc.helsinki.fi>
Date:         Wed, 20 Jul 1994 19:34:02 EDT
Reply-To:     jorge@PHYAST.PITT.EDU
Sender:       Lojban list <LOJBAN%CUVMB.bitnet@FINHUTC.hut.fi>
From:         Jorge Llambias <jorge@PHYAST.PITT.EDU>
Subject:      Re: ill-formed
To:           Veijo Vilva <veion@XIRON.PC.HELSINKI.FI>
Content-Length: 2627
Lines: 75

la kolin cusku di'e

> I think there is a hierarchy of levels of well-formedness in Lojban (probably
> in any language) and it would be worth trying to clarify them.
>
> Here is a first attempt:
>
> 1. Ungrammatical - fails to parse at all
> eg
>         *cu noi .e xamgu
>         *mi viska le gerku jo'u le mlatu
>
> (Impressionistically, there is a further distinction, between completely
> uninterpretable strings like the first, and nearly-valid utterances like
> the second, but I doubt whether the distinction can be made objective enough
> to be useful).

Definitely, the second one is human-parsable. I'm not sure what other
types of sentence are ungrammatical only due to the LALR(1) restriction,
but that could be a way to make the distinction.

> 2. Parses, but fails to be meaningful because undefined terbri are invoked
> eg
>         *mi gleki do le mlatu

This is grammatical, so I wouldn't use a * for it. Leaving aside the problem
of whether {do} can be a {se gleki}, I don't see much difference between that
sentence and

        mi gleki do do'e le mlatu

Other than the confusion caused by implying that {gleki} has a third place.
If it doesn't make sense, then it goes at the level of ill-formedness of
green ideas, which is the mildest.

(That sentence would go at level 3 because of the illicit raising, I think)


> 3. Parses, but fails category consistency (including illicit raising)
> eg
>         mi cusku li mu
>         do rinka zo gleki
>

Have you decided on a set of categories/features? I think that for a first
cut we should concentrate on mutually exclusive features, i.e. such that
if a terbri is +feature for one, then it must be -feature for all the
others. These are the easiest to deal with, because only one feature need
be assigned to each terbri. This is my list thus far (I give an example
terbri for each):

selcmi (set)
namcu (number)
fatci (predication: du'u)
selcusku (expression: sedu'u)

I'm not sure whether to include fasnu (event), and even less sure about
other abstractions. The problem is that they're not always separate from
objects. The +/-concrete is another one I have to think more about.
I'm sure there must be others, but the list probably doesn't go over 10.

> 4. Parses and meets consistency requirements but is semantically or
> pragmatically meaningless or self-contradictory
>         ko'a balvi le balvi be ko'a
>         lo skacau ke crino sidbo cu vilfenki sipna

In many cases, such (ab)uses can be accepted as poetic licence, in my
opinion. At worst they're false or confusing, but there's nothing
extremely wrong with them.

>
> Colin Fine
>

Jorge