Re: [lojban] Automatic Lojban -> English translation?

[Björn Gohla <b.gohla@gmx.de>]

On Tuesday 18 June 2002 19:51, robin wrote:
[...]
> I'd say readable and grammatical, but not idiomatic. We already have
> parsers and glossers; all that is needed is something that will generate
> the English syntax. Even so, it would probably be a long time before we
> had anything remotely resembling natural-sounding English. It's pretty
> easy to get a program to render "mi dunda le cukta le ninmu" as "I give
> the book to the woman", and you might even be able to, with much more
> advanced programming, get it to guess unspecified tenses with 70%
> accuracy (I wonder how current machine translation performs with
> Chinese). The problem would come with rendering stuctures which have no
> natural English equivalents - you end up with something that is
> understandable but weird, as with a lot of natural language translation
> software. OTOH, it would be pretty easy for someone to look at a
> machine-generated tranlsation and rewrite it in more idiomatic English
> (or whatever).
[...]

they are doing pretty well in the kpml project:
http://www.fb10.uni-bremen.de/anglistik/langpro/kpml/kpml-description.htm