[lojban] Re: jimpe: Language Processing in Lojban

[Rob Speer <lojban-out@lojban.org>]

On Fri, Jul 30, 2004 at 09:38:13AM +0200, Arnt Richard Johansen wrote:
> Also, I am personally curious about how you do Natural Language Generation
> in Lojban.

To be honest, we don't do much of it yet. We have some basic functions to turn
semantic objects back into Lojban, but it ends up in a rather obtuse form with
all the terminators in it (and if you have quantifiers, they're all in prenex
form).

So we don't actually get the output "ua la bab na crino" yet. We get something
like "ua naku la bab [cu] na crino [vau]". But we have plenty of time to fix
this. (Submitting a paper involves some amount of projection into the future,
because we have to submit the paper for approval six weeks before the
conference. If we're accepted, we get to revise the paper before the
conference.)

But if anyone has suggestions on how to determine which terminators are
elidable (for example, how does Nora's random sentence generator do it?) that
would be really helpful for the project.
-- 
Rob Speer