Re: Natural Language Processing Using Lojban

[John Glossner <glossner@LUCENT.COM>]

[[ This is an inprogress private thread that is being moved to the
Lojban ]]
[[ Forum. If there is significant interest, I will summarize the
previous ]]
[[ correspondances.     Please join
in.                                   ]]
[[ Thanks, John
Glossner                                                  ]]


Saluton Lionel,

Lionel Bonnetier wrote:
>
> Saluton John,
>
> Vi do decidis esperantistighi?
>
Currently, my Esperanto studies have been limited to 2 weeks part time
and that includes a new baby! However, what attracts me to Esperanto
is the fact that I can understand quite a bit with practically no
time investment! This is not true of Lojban. Now, I realize that
I am of Euro-Germanic descent and that knowing some French and Dutch
makes Esperanto almost too easy. However, after one hour, I was able to
comprehend some sentences. Lojban is a different beast altogether. That
is going to take much more time.

> I do agree that Esperanto and Lojban are less tricky to parse at a
> skeen-deep level, but after many years of using esperanto I must acknowledge
> that it bears exactly the same semantic ambiguities as any other language on
> the planet.
>
Agreed! My reasons for learning the two are very different. What I hope
to
achieve with a "language compiler" is the ability to represent certain
semantic
ambiguities at various levels. From my limited understanding of Lojban,
I can
specify ambiguity. However, in Esperanto, the ambiguity is a normal part
of
the language. Let's suppose I want to convert a Dutch text to a German
text.
What I would look for is a conversion of

        Dutch->Esperanto->Lojban->Esperanto->German.

Now the obvious question is why go to Lojban?
What I am hoping is that just as a compiler can perform certain
optimizations at various level of Intermediate Representations (called
IRs), that
I can do the same. Where might this be useful? Suppose we have a
sentence such
as: The water can't be too high. I hope to resolve "meaning" ambiguity
with
Lojban. For example, if I can determine that the context is a nuclear
power
plant, then you always want lots of water :-) (some Saturday night live
fans may
remember a hilarious comedy stunt with this ambiguity as the main point)
However, if I can determine that the context is glass of water, then I
know
if the water gets to high it will overflow.

Now, I really don't know how to represent this in Esperanto (or any
other language)
unless I add a whole lot of words. I am hoping that Lojban can
"optimize" some
of these sentences and clarify the ambiguity. It may be only a hope but
if it can
be done (which I do not yet know), then I can have an interesting way of
representing knowledge. Plus, with Lojban, it seems I can determine the
truthfulness of falseness of a statement.

> Have you heard about the Distributive Language Translation (DLT) Project
> sponsored by the Netherlands in the '80s. It used Esperanto as the
> turntable, and was competing with another project using numeric entries,
> whose name I can't remember now. The money stopped flowing in at some point,
> and I don't know what all that became. You may find info at some esperanto
> web sites.
>
If this is now the program Ergane at http://www.travelang.com then yes I
know of it. The S/W is free, the support is excellent, and I have found
it very useful for learning Esperanto and Dutch!

>
> Several levels of representation might bring suppleness, the same way the
> neuron layers cope with complex signals.
>
I agree. This is exactly why I'm exploring multiple representations.
Esperanto
merely solves the O(n^2) problem of language<->language translation.
Lojban
on the other hand is a way of dealing with the levels of representation.

> But we have to go very deep behind
> the language to meet Minsky's "assembly of the mind".
>
Here is where the fact that I'm a Computer Scientist/Architect will
reveal
itself :-)

> I would be happy if we can continue these discussions -- I'm stopping now
> for I'm made awkward by a whole night & day of awful C++ coding.
>
And you're still sane ;-) My favorite times at UNC were the 3 straight
days of programming (and not even noticing that the days had passed :-)
Anyway, I'm at Bell Labs and code daily in C++ - let's just say that it
is
an expressive language! :-) (try to parse the real meaning of that one
in
a program!!!)

> I hope my English is understandable :)
>
As you have seen, your English is far better than my French. And,
judging
from what I've seen you write, I think it is better than my English!!!

>
> Maybe we should drive some of our threads in the Lojban list? Unless they
> have a technical problem, no article has been posted for long, but I know
> there are many computer scientists among them.
>
Agreed. Done as of this response.

> I guess you've downloaded their lojban parser?
>
Yes but I've not played with it yet. I will get to that soon though.

John
--
**********************************************************************
* John Glossner                           *                          *
* Adv. DSP Compiler / Architecture R&D    * 55E-212                  *
* glossner@lucent.com                     * 1247 S. Cedar Crest Blvd *
* http://einstein.et.tudelft.nl/~glossner * Allentown, Pa  18104     *
* Ph: (610) 712-2532                      * Fx: (610) 712-2790       *
**********************************************************************