Ryan Leavengood wrote:
I've had some forays into learning Lojban in the past, as well as a
bit of Esperanto. In fact I can recall a few years ago when I used IRC
more, my two channels of choice were either #ruby-lang or #lojban.
Unfortunately I never did get too far in learning the language, and
have forgotten most of what I did learn 
But part of my goal was always to try to create a decent
human-computer speech interface based on Lojban, which due to its
logical and regular nature would probably be easier than using
English. This would mostly be for fun, learning and my own personal
use (since I doubt the average person would learn a new language just
to talk to their computer.)
So, anyhow, let me know what you are planning Hal.
Well, I've thought of a few things.
1. An ri-like utility that works at the command line -- give it a
Lojban word, get back all info about it
2. A program (using PDF::Writer) which would hopefully make TeX
unnecessary in spitting out a nicely formatted dictionary
3. A program that would make what the Brits used to call a "trot" --
a text with the hard words translated in small print underneath
each line. And a page-oriented variant on this.
4. A nice parser.
5. A sentence diagrammer?? See my blog.
6. An annotator. Here's a first pass at it:
http://hypermetrics.com/pikmin.html
As for the annotator, I have bugs to fix and features to add. One
big feature I'd like is the ability to create an entire downloadable
annotated document (which someone could then host on his own site
or view offline). It also needs to be updated in where it gets its
content.
7. I thought of a text-to-speech thing (not voice recognition), but
I would do it by cheating -- convert the Lojban to phonetic gibberish
(e.g. "gismu" -> "geese-moo") and run it thru an ordinary TTS program.
Trouble is, cheating would sometimes get difficult, and many TTS
programs don't have certain phonemes like Lojban {x} (unless you
select a German voice or something).
Those are pretty much all my thoughts.
Hal