On Sun, Feb 10, 2013 at 03:32:32AM -0800, la gleki wrote:
> Saying that unfilled places are {zo'e} and can
> be easily derived from context makes Lojban not so intuitive for computers
> because they don't know this context.
Sorry, but that's no argument for me. Contextual interpretation
is apart from the symbol grounding problem perhaps _the_ research field
in all major AI disciplines(though in very different settings),
so expect some progress there.
> In fact it just suggests that Lojban is a normal natlangish conlang.
> Of course there are no problems perceived by humans in ordinary "vulgar" speech.
Lojban is _supposed_ to be a language for human communication.
It just cleans up a lot of problems like audio/visual mismatch and syntactic
ambiguity, which make it even harder for computers as well as humans to
understand a language.
> The problem is that Lojban is too loose in allowing what can be inside
> abstraction-sumti ({mi kakne lo nu do citka lo plise} is still gendra
> although absolutely nalsmudra).
A language (for communication) is _not_ typed lambda calculus.
> A fix to this problem might sound like this.
No problem, no fix needed.
v4hn
Attachment:
pgpS3GWF9of3B.pgp
Description: PGP signature