Monday, January 30, 2006

Babel - or Pentecost?

Well, actually, neither. There is an Italian proverb - unfortunately I cannot put my hand on the original - which is roughly translated as "to translate is to betray". And perhaps even more to the point is that most excellent line from Father Brown: "No machine can lie - nor can it tell the truth."

In this case, we may definitely add, "nor can it translate"... though it makes at least some attempt at it!

As I told you yesterday, I am no linguist. So in order to get some hint as to how well the mechanical translations worked, I did the only obvious thing, which will probably make you laugh... I fed the foreign language versions back into the machinery, but with the switch running the other way! so now I get back the English version of the other languages! Hee hee.

This proved to be VERY interesting:

From the French:
The Favourite Of GKC
Frances Blogg was devoted to G K Chesterton. Is thus this blog - and thus you can expect anything "of the pig to the pyrotechny" which "the truth of the only true philosophy illustrates..." [ GKC, The Thing ]

From the Spanish:
Favorite Of GKC
Frances Blogg was dedicated to G. K. Chesterton. He is so this blog - and so you can tell on any thing "of the pig the pyrotechnics" that "illustrates the truth of the only true philosophy..." [ GKC, the Thing ]

From the Italian:
Favorite Of the GKC
Frances Blogg has been dedicated to G. K. Chesterton. Therefore it is this blog - and so as to you can preview qualche.cosa "from pig to the pyrotechnics" that "illustrate the truth of only philosophy to align..." [ GKC, The Thing ]

From the German:
Favourite GKCs
Frances Blogg was inaugurated G. K. Chesterton. Is like that this blog - and thus you can expect everything "from schweinefleisch to pyrotechnics", "illustrate the truth of only applicable philosophy...", [ GKC, the thing ]

Of course this is a very short test, and was kind of unfair because of the meaning running through the quote-marks. But it is instructive, and I may wish to play with it again.

But I must call your attention to the very amazing result of the Italian, which is so good I think I may have to use it elsewhere: the truth of the only philosophy to align... then again, perhaps the computer has been reading Chesterton, for here we find that very same idea:
It is always simple to fall; there are an infinity of angles at which one falls, only one at which one stands. To have fallen into any one of the fads from Gnosticism to Christian Science would indeed have been obvious and tame. But to have avoided them all has been one whirling adventure; and in my vision the heavenly chariot flies thundering through the ages, the dull heresies sprawling and prostrate, the wild truth reeling but erect.
[GKC, Orthodoxy CW1:306]


At 31 January, 2006 07:59, Blogger rhapsody said...

:)Father Brown was right! So, would what you are writing come out meaning what you meant, in a different language?

Someone once told me that the different translations of the Bible could give a verse a different meaning- even the different versions in English that are available. I doubted that, believing that the translations from the original are done carefully- it's the interpretations that vary.

Although, I do wonder why there are those who swear by what I refer to as the 'condensed, abridged version' of one especially preferred Bible, which bears a man's name. Someone once said that the Americanized version of the Bible actually changed the meaning from that one- although that one is in English, which isn't one of the original languages either!

Pardon, Doctor Thursday-

I believe I babbled-on off topic again!

At 31 January, 2006 09:17, Blogger Dr. Thursday said...

It is not off topic. It is very important. I have just posted about words and their use over on the ACS blogg.

At 04 February, 2006 16:52, Blogger electroblogster said...

Darn! I just checked out a link that I had had to a great article on machine translation and found that it is now hidden to all but subscribers. I can't recall what was in the article vs. what was going through my head while I was reading the article. Which is to say it was a thought provoking article!

I do remember that it is a continuing challenge to make a machine translate between languages. The popular experiment is exactly what you have done - run it forwards then backwards and see what has been lost, butchered or worse. I did the same thing myself translating an email to a collegue in Taiwan into chinese and back. What a mess! I give them more credit for understanding me now.

I recall reading that the mahines, though starting with the words to be translated and with their programming to draw on, had to step farther and farther back from the words themselves in order to improve the translations. In other words they had to try to gather something of the meaning - to recognize patterns that would signify common thoughts (common because there would be no way to program them for all thoughts!!!) and then reconstitute the foreign phrase to replicate the thought.

... But even then you get to the stage I find most fascinating. The language is not only a reflection of reality but a reflection of perception and thought. And different languages belie the fact that different peoples perceive and think about things differently. It is not necessarily possible to SAY the English-speaker's thought in Mandarin Chinese. You may have to spend a lot of time and words in the effort to make that Chinese speaker share your idea.

This is why I find it so interesting and compelling to learn other languages. (This is why I found the Chinese book so insteresting )

I had a sense of this back in college philosophy classes when we kept resorting to the phrase "but people say..." in order to get a glimpse at how something real actually exists.

So good luck in your efforts to make these mahines translate Dr. Thursday!

Here is the article link should any of you happen to be able to retrieve it.

At 05 February, 2006 15:54, Blogger Dr. Thursday said...

How interesting - and how accurate you are - about the effort to make one's self understood - even when both have a common language!

The fundamental problem with trying to make machines do translation is that we don't really know how we humans do translation yet. We only think we know, in a simple sense, because we start out learning things like

canis = perro = "dog"

and so forth. This kind of thing (which we might call a look-up table, or an association list, etc) is handled in a straightforward manner by computers. And they can even deal with more subtle tricks like the (vocabulary word warning!) periphrastic verbs of English - that is, things like "you may be watching" which turn into one Latin subjunctive - they have an ending for the verb and say in one word what we take four to say!

But once we add real semantics - once we start to look at the meanings of words - and how those meanings change based on the context - here we have opportunities. Opportunities which are far worse than forks or even knives in the road! We might mean something very clear - or a subtle, perhaps snide remark - or a personal, friendly joke - or a sinister attempt to mislead - or a wildly hilarious pun.

Oh, if we could only recognize the gift of our language - that we can read, that we can hear, that we can speak! How much more would we... well. Sorry - let me go back to the machine topic.

But then again - it is related. The horribly boring mechanics of mechanical data - I oversee the work of my software, handling some 20,000 plus files every day - coming and going, systole and diastole, inhaling and exhaling... they are important for our customer, but vastly uninteresting. They are in a horrible format (I did NOT invent it; I simply arrange that they go where they are supposed to go!) they say nothing more than (on the way out) what is to be done at a certain time, and (on the way in) what actually happened at a certain time.

(Sigh! They might be stones that fell and left their imprint on a muddy shore... curious but inhuman...)

No poems. No stories. No meditations. No discussions or essays or polemics. Not even lies, or diatribes or libels.

FOR THIS IS NOT MIND TO MIND. If I am to communicate with you, I must, absolutely must - Sauron-like, Tolkien-like, God-like - somehow put some of my very self into a something exterior to myself, and THEN you must take it as your very own.

And I think that Tolkien, word-master that he was, would agree. What greater technology is there? Words are truly "rings of power", and even the Great Ones might fear to use them...

To the extent, then, that rules can be constructed, a machine can follow them, and they will certainly help us with the simpler tasks...

But I think it was St. James who talked about the complexities of the tongue. If it's such a grave challenge to us, who have used it continuously for almost our entire lives, no wonder it is so hard to get a computer to do much with it.

Lest it be thought that I disclaim the idea of machine translation from non-technical grounds, I simply point out that computer science recognizes some things are not computible. As Gödel showed quite some years ago, not even all math problems are so!

But certainly this tool is fun, as far as it goes. I hope they add other languages soon - especially Latin. But then I think someone with more philosophy than me may begin to understand, and will comment on it better than I have.

At 05 February, 2006 16:17, Blogger Dr. Thursday said...

Two other points, after re-reading everything together.

It may be "possible" to represent all thoughts in a computer - in the sense that one might represent all thoughts in English - or in any human language - and therefore in ASCII or some other notation which transcribes language - but those things are not thoughts.

Show me green in ASCII. Show me orange - the scent, not the color! And roast beef, pine trees, Beethoven's 9th (choral movement!), fireworks, seashores... what do I mean when I say these things? What is in YOUR head? Not the words. The ideas! Maybe you're a oboist, and hate the Ninth symphony, and you hear a dissatisfied conductor... Maybe you only think of nearly frozen fruit-groves, or your nearly empty tube of chrome oxide, or sand between your toes, or touring a paper-mill, or ordering more potassium nitrate .... it all depends what we mean... but again, all this is coming to YOU in ASCII (the symbols of the machine!) What's in YOUR mind?

It's because there's an old Scholastic axiom:

Quidquid recipitur, in modum recipientis recipitur.

"Whatever is received, is received according to the mode of the receiver."

But I will defer this for it will take too long to explore just now, and I still have one other point to make.

That is, a computer does its work based on instructions I (or another programmer) give it. It is formally impossible for it to do something which we cannot "describe" by means of those instructions. Another way of putting it is this - which I tell my colleagues at work:

If you want the machine to do something, I have to know how to do it first.

I cannot be any means express an action to the machine which I do not already grasp. I don't have to carry out the action, but I have to understand it in all its detail - or I cannot instruct the machine to carry it out.

Someday, perhaps, more on this as well.

Thanks , "Electroblogster" for a very stimulating response! Let us continue this talk.

At 05 February, 2006 18:12, Anonymous Anonymous said...

the French and German have "favourite"

the Spanish and Italian have "favorite"


At 06 February, 2006 14:04, Blogger electroblogster said...

Now you have given me a whole batch of new things to think about. Our conversation grows wider instead of narrower.

But your axiom is so true:
Quidquid recipitur, in modum recipientis recipitur.

Thoughts ARE in the head. And they don't just describe oranges, or scents but also feelings, abstractions and selves. Even if a computer can analyze a smell (of say an orange) and even link it to a visual counterpart in shape and color, an entry unique to the taste, data about its weight and texture it isn't the way a orange is really known. In fact the orange is known differently by different people (as you alluded to) because of their subjectivity. Even if it is the same real orange people are experiencing doesn't mean they experience it the same. It's why that poor PC can't be expected to liken the rhind-wrapped sweetness to a charitable act.

There are things which we say that haven't yet been thought.

Thus ultimately, I don't think it is possible to represent all thoughts in any human language. The language can lead you in the right direction but you have to think. It is not mind to mind but rather mind to speech to mind. Sometimes mind 2 gets more out of the speech than mind 1 put into it. St. Thomas Aquinas was a student of Aristotle - but took the ideas to a whole new level! (Great story about Thomas and St. Albert I will have to tell some time). The Catholic encyclopedia backs me up: "Thomas could repeat the lessons with more depth and lucidity than his masters displayed"

It's funny that the orange may be an EASIER example since the orange exists all by itself. And it is genius that Christ spoke in parables for the same reason. City folk may not know fishing all that well off the top of their heads - but they are able to find out today just as clearly as 2 millenia ago what it takes to get a fish - and what it means to fish for men.

I would be curious to know how well our friend GKC would be understood by a native of Kabul. He depends heavily on the western mind in order to slingshot your thoughts to where he wants to go. How else can you explain the way (very literal) contradictions he uses to get you to think that not one but both things - and neither are true? Even I have to resort to this mode to describe it. Hmm.


Post a Comment

<< Home