Date: Sun, 01 Aug 1999 14:49:19 +1000
To: [log in to unmask] [SORRY: I sent the message haf-finished]
At 21:02 26/07/99 GMT, Marcos wrote:
>Tem Mon, 26 Jul 1999 00:36:02 -0400, Robin Gaskell
>>* So, your conception is of an IAL as a true 'meaning representation
>Yes, more or less.
* It's nice to be able to reach common understandings.
>>> ...... I suspect ... using Glosa words ... commands of Basic ....
>> There were two ways they explained it:-
>> . -----------------
>> . A single translation renders a valuable document into Glosa: from
>>here, it can be translated into any other language, readily.
>Well, and tell me, where's the word "machine" in all that?
* Guess I made an assumption that the reader would interpret " ... a single
translation renders ... " to mean "'computer' translation".
>>>logical language in computing, ..... what's its worthiness?
>>* To help in translation ... Planned Language that is intended for
>>translation, is best if it is logic-based.
>Ah, I thought you were thinking on something else than MT.
* Probably I was; I do believe that logic in human language is a good
thing, of itself - leading, as it does, to clarity of communication.
>>>I still doubt this. ...... I fail to see
>>>great advantages in using a Esperanto-like language as interlingua
>>>instead of a more formal (and non-human) linguistical representation.
>>* Well, the Esperantists ...... were very strong
>>on having their language as a 'meaning representation system'!
>Yes, they were very strong... but only they.
* It does seem a shame that a proper Feasability Study was not carried
out for this project. The end result, nobody wins: the Esperantists get
browned off because their project had its funding cut; and we lose, because
narrow-minded bureaucrats can point to the DLT, and say 'Pivot Language'
schemes have been proved not to work.
>>>Well, I think that in order to be an optimal source language for MT we
>>>basically need that in it:
>>>1. Every word, phrase, or sentence has one and only one meaning.
>>>2. Every word, phrase, or sentence delivers the necessary information
>>>to generate a proper equivalent word, phrase, or sentence in target
>>* Sounds like a very particulate philosophy of translation, to me. I had
>I think it's quite basic,
* Ok, if that's what the computer needs, thats what it's gonna get. But
I suspect that such a process, without room for idosyncrasies of
expression, will be a little stultifying, not to say sterile.
>>imagined that the Artificial Intelligence behind the translation process
>>would have been more sophisticated.
>and it doesn't entail that AI be not used, but certainly it would be
>desirable that it wouldn't be needed.
* Sorry! Apologies here: I was jumping ahead. Apart from getting my
wires crossed between MT and computer-run translation using a 'Pivot
Language', I was showing that I believe computers will, ultimately, play a
much bigger part, than they now do, in the processing of information.
But, of course, a human language medium that is highly machine tractable,
must emerge before the AI/NLP process comes together. Needless to say I
was bundling the concept of the IAL/MT and that of the IAL/Intermediate
Language together, because I feel that one leads naturally into the other.
This was naughty, and unrigorous, of me.
>> My studies of language convince me that "ambiguity" ...... cannot be
>>^bred out of^ human language by translation.
>I believe some amount of ambiguity is preserved in UTL, but just to
>the point that it doesn't affect its eficiency on MT. .......
* A fine line, if you can draw it.
>>>This may work perfectly when sentences respond to a simple, standard
>>>structure. But most sentences do not. I seriously doubt that a Glosa
>>>parser could get always the correct PoS of any word just from the
>>>analysis of sentence's syntax, .........
>>* ....... in most languages .. where grammar is mixed between
>>morphological structures and word order .. it takes some disambiguating to
>>work out which word is functioning as what.
>> But, in both written Glosa and machine-generated Glosa, I was thinking
>>that the rules of the language's grammar would dictate, clearly, where the
>>different Parts-of-Speech would fall; ....... down to a single probability.
>Maybe, but I still got the impression that using Glosa for MT .......
Having >PoS marking make things much more easy.
* External imposition of P-o-S markers is a possible compromise that
could be struck to make the basic Glosa 'clean lines' acceptable to
Programmers ... but it would spoil Glosa's elegance.
>>> Btw, do you know Dkcsac? ..... a guy I found at a.l.a. newsgroup. He seems
>concerned with MT, and he likes Glosa ..... he expressed, in a moment of
>conversation, his disappointment for that Glosa were not suitable for
>such a process.
* No, I have not met the chap, but you might let me have his e-address,
>>>> Robin Gaskell
>>>Btw, how should we (or the machine) translate that "saluta"? Is it a
>>>verb, a noun, an adverb...?
>>* There is a fairly well-established heirarchy of functions. ......
the >>most terse ..... is a one-word sentence ..... a verb.
>Ok, ....... The result is a bad translation (at least into the natlangs I
>know), because we normally use a (plural) noun in that place:
* As I see it, we can teach the computer that the solitary word, or small
group of words, before a person's name is a "salutation" not some other
part of speech!!! Let's innovate, here.
Iso mi pa previo dice:-
The Bottom Line is people NOT money.
___________ ^^^^^^ $$$$$