[YG Conlang Archives] > [Latejami group] > messages [Date Index] [Thread Index] >


[Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: Welcome!



Yes, I agree.  Still, one of the reasons I think the
Latejami-to-English translation software is so exciting is that for
every Latejami sentence, it produces a corresponding English sentence
with the same meaning.  Therefore, it seems that even though a full-on
English-to-Latejami translation program would be impossible, a 
"constrained English"-to-Latejami translation program that **only
handled the subset of English sentences that the Latejami-to-English
program could produce as output** would be much more tractable.


Therefore, if a clever interface were to help the user input English
sentences that the parser could actually handle--maybe via some kind
of interface that continually displayed acceptable sentence
completions, just as Google Suggest displays possible query
completions--then you'd have an easy-to-use "constrained
English"-to-Latejami translator.  If others constructed such
Latejami-to-natlang and "constrained natlang"-to-Latejami programs for
other natlangs, you'd have essentially created the first universal
translator.

(E.g., if monolingual Japanese speakers wanted to translate something
into another language, they'd only have to formulate their sentences
in a way acceptable to the "constrained Japanese"-to-Latejami
translator, which wouldn't necessarily be too hard if the interface
helped them out.  Then the Latejami-to-(other natlang) translator
could translate their passage into the target language.  Even if it
was a hassle to formulate utterances in "constrained Japanese", there
are definitely lots of important cases in which a good translation
would be worth the work!)


Obviously this would be a huge job probably requiring multiple people,
and I know you're taking a break from all things Latejami for now. 
It's also hard to say how good the resulting translation would be
given that the into-Latejami and out-of-Latejami translation steps
would both likely introduce inaccuracies in the semantics and
pragmatic meaning of the original utterance.  But if you ever consider
releasing your Ada code either as open source or under some sort of
commercial license, please let this list know!  Your work is
fantastic, and could one day go a long way towards helping people
bridge the language barrier.



--- In Latejami@yahoogroups.com, <ram@...> wrote:
>
> "rexxia" <grecchia@...> wrote:
> > 
> > Back in the day, I remember hearing mention of a Latejami-to-English
> > translation program, presumably being written by Rick Morneau.  I
> > found this pretty impressive, given that Latejami certainly seems no
> > "closer" to English than any other natural language.  Did anything
> > ever happen with this?  Was such a program ever completed?
> > 
> 
> The software exists, written in Ada.  One of the reasons why I stopped
> working on Latejami is that the basic design algorithms were becoming
> too complex because I didn't anticipate some of the problems I would
> eventually run into.  A complete redesign is needed (which will also
> require some morphological changes to the langauge).
> 
> You can judge the quality of the software yourselves - the English
> translations of all of the dictionary examples and all of the lesson
> examples and drills are the unedited output of the translation software.
> 
> It's not fair to compare translation software for Latejami-to-English
> vs. natlang-to-English.  Latejami syntax is simple, precise, and without
> exception, and Latejami morphology is totally regular.  Because of this,
> software to parse and analyze Latejami is much, much simpler than for
> any natural language.  For example, the Latejami parser is short, simple
> and 100% accurate.  The equivalent software for a natural language would
> be incredibly complex and fraught with error.
> 
> 
> Regards,
> 
> Rick Morneau
> http://www.eskimo.com/~ram
>