TECH(?): can everyone write impeccably grammatical Lojban?

[Ivan A Derzhanski <iad@COGSCI.ED.AC.UK>]

>  Date:        Fri, 25 Sep 1992 16:58:55 +0100
>  From: And Rosta <ucleaar@UCL>
>
>  Obviously most of the Lojban written
>  at  the mo is in practise ungrammatical but  nevertheless communicative

Maybe not most of it.  Quite a bit of it is, obviously.

>  but in principle I suppose the technical lojbo line is that if it's
>  ungrammatical then it's not Lojban.

As is the technical linguistic line, for that matter: if it doesn't
follow the grammar of L, then it is not L, it is another language.

>  As Ivan no doubt knows better than me, the distinction is sometimes
>  (usually? always?) made between a *parser*  and (don't know the
>  technical comput ling term) a *grammaticality tester*.
>
>  A grammaticality tester looks at a sentence & if it's grammatical
>  gives you its structure(s), or, if it's ungrammatical just tells
>  youtopiss off.
>
>  By contrast, a parser strives tocome up with ananalysis whatever the
>  input.

I'm afraid you and I are using the terms "parser" and "grammaticality
tester" very differently.  In my experience an ordinary parser tries
to come up with an analysis, assuming in any case that the text is
built according to its grammar, and if it works, it outputs the
structure (here's the difference: I would expect a GT to simply smile
at this point), and if it doesn't, it says so.  Now there are special
parsers which are engineered for garbled input, but they are not what
people usually mean when they say "parser".

>  So basically if you
>  elide one terminator  too many or whatever you've had it.

Yes.  The reason being that it is damn difficult to formulate any
criteria for near-Lojbanness (or near-correctness) of a text.  Try to
imagine a grammar of near-correct English -- a set of rules that
generate utterances which may not be grammatical English, but from which
someone with competence in English would be able to derive a meaning.
(I didn't say write it, just imagine it.  :-))

Ivan