Will language-translation software ever be perfected?

[BrainGlutton]

@Abraham Lincoln

Four score and seven years ago our fathers brought forth on this continent, a new nation, conceived in Liberty, and dedicated to the proposition that all men are created equal.

Now we are engaged in a great civil war, testing whether that nation, or any nation so conceived and so dedicated, can long endure. We are met on a great battle-field of that war. We have come to dedicate a portion of that field, as a final resting place for those who here gave their lives that that nation might live. It is altogether fitting and proper that we should do this.

But, in a larger sense, we can not dedicate – we can not consecrate – we can not hallow – this ground. The brave men, living and dead, who struggled here, have consecrated it, far above our poor power to add or detract. The world will little note, nor long remember what we say here, but it can never forget what they did here. It is for us the living, rather, to be dedicated here to the unfinished work which they who fought here have thus far so nobly advanced. It is rather for us to be here dedicated to the great task remaining before us – that from these honored dead we take increased devotion to that cause for which they gave the last full measure of devotion – that we here highly resolve that these dead shall not have died in vain – that this nation, under God, shall have a new birth of freedom – and that government of the people, by the people, for the people, shall not perish from the earth.

Plug that into Babelfish, choose “English to German,” and you’ll get:

Vor vier Kerbe und sieben Jahren unsere Väter weiter geholt auf diesen Kontinent, eine neue Nation, begriffen in der Freiheit und der Angelegenheit eingeweiht, daß alle Männer verursachtes Gleichgestelltes sind. Jetzt nehmen wir an einem großen Bürgerkrieg teil und prüfen, ob diese Nation oder irgendeine Nation also begriffen und so eingeweiht, lang aushalten können. Wir werden auf einem großen battle-fangen dieses Krieges getroffen. Wir sind gekommen, einen Teil von diesem einzuweihen auffangen, als abschließender stillstehender Platz für die, die hier ihre Leben gaben, die diese Nation leben konnte. Es ist zusammen passend und korrekt, daß wir dies tun sollten. Aber, in einer größeren Richtung, können wir nicht – wir können nicht widmen – uns einweihen können nicht – diesen Boden heiligen. Die tapferen Männer, das Leben und die toten, die hier kämpften, haben sie, weit über unserer schlechten Energie hinzuzufügen oder abzuziehen gewidmet. Der Weltwille wenig Anmerkung noch erinnern lang, sich an was wir hier sagen, aber es kann nie vergessen, was sie hier. Es ist für uns das Leben eher der unfertigen Arbeit hier eingeweiht zu werden, die sie, die hier kämpften, bis jetzt so vortrefflich vorgerückt haben. Es ist eher, damit wir der großen Aufgabe hier eingeweiht werden können, die vor uns restlich ist – die von diesen geehrten Toten wir erhöhter Hingabe zu dieser Ursache nehmen, für die sie das letzte volle Maß der dieser gaben Hingabe – beheben wir hier in hohem Grade, daß diese Toten nicht in nichtigem gestorben sind – dem diese Nation, unter Gott, eine neue Geburt der Freiheit hat – und das Regierung der Leute, durch die Leute, für die Leute, nicht von der Masse umkommt.

Now, whether that would make any sense to a German I do no know. But if you again use Babelfish to translate it back into English, you get:

Before four notch and seven years our fathers further gotten on this continent, a new nation, understood in the liberty and the affair inaugurated that all men are caused on an equal footing. Now participate we in a large civil war and examine whether this nation or any nation can bear thus understood and so inaugurated, long. We are met on large battle to catch this war. We came to inaugurate a part of this catch, when locking standing place for those, which gave here their lives, which could live this nation. It is together fitting and correct that we should do this. But, in a larger direction, we can inaugurate not – we cannot dedicate – us can not – this soil holy. The courageous men, which lives and the dead, which fought here, have it, to add or take off far over our bad energy dedicated. The world will little note still remind long, here of which we say themselves, but it can never forgets, what them here. It is for us the life rather the incomplete work to have been inaugurated here, it, which fought here, have up to now so splendid advanced. It is rather, so that we can be inaugurated the large task here, which is before us remaining – of this honoured dead ones we increased devotion to this cause take, for it the last full measure of the these gave devotion – repairs we here in high of degrees that these dead ones did not die in futile – this nation, under God, a new birth of the liberty has – and government of the people, by the people, for the people, not of the mass dies.

Seems translation software has a way to go. Will it ever be perfected, to the point of making human translators obsolete? Is that possible at all until they develop a strong AI?

[Sal_Ammoniac]

I’m not an expert, but that doesn’t mean I don’t have an opinion! I think the answer is yes. To some extent it’s a matter of raw computing power. Think of it this way – the Gettysburg Address has been translated into German somewhere, by somebody. So you take that translation, and begin building a database. The American Constitution, ditto. The Sun Also Rises, ditto. And so forth. Then you do the reverse, and get German into English. By the time you have a massive database, of all kinds of written expression translated from one language into another, you have a mine of information to which you can compare every single phrase or phraselet in the matter to be translated. Babelfish isn’t comparing to anything, much. It’s basically doing little beyond looking for a word-to-word match, I’d guess with some probability rules built in.

So for example, take the beginning of the translation (“Before four notch and seven years”) and google it, and you get basically no results (actually, there’s one person out there who did what you did [and maybe even is you]). What you’re doing there is proving, by going to a massive data source – all the pages compiled by Google – that the translation is bad. A good translating program would do the reverse, in effect – that is, confine its output to phraseology that is known to be good. And that takes some major computing horsepower, which is the direction that computers are heading in anyway.

[mrklutz]

I think algorithms like Sal describes will be a vast improvement and will probably suffice for day-to-day conversational translation. But if you look at more literary type of language, you will often find human translators disagreeing as to what the best translation is. Subtlety and shades of meaning don’t do well in translation, even in the hands of experts. Automating that level of nuance just isn’t likely, in my opinion.

[John_Mace]

Just to give another, more modern example, here’s a clip from Pat Buchanan’s culture war speech, which was supposed to sound much better in the original German.

The agenda Clinton and Clinton would impose on America–abortion on demand, a litmus test for the Supreme Court, homosexual rights, discrimination against religious schools, women in combat–that’s change, all right. But it is not the kind of change America wants. It is not the kind of change America needs. And it is not the kind of change we can tolerate in a nation that we still call God’s country.

English -> German:

Die Tagesordnung Clinton und Clinton würde Amerika auferlegen – Abtreibung Bedarfs-, ein Lackmustest für das Höchste Gericht, homosexuelle Rechte, Diskriminierung gegen fromme Schulen, Frauen im Kampf – der die Änderung ist, ganz recht. Aber es ist nicht die Art der Änderung Amerika wünscht. Es ist nicht die Art der Änderung Amerika Notwendigkeiten. Und es ist nicht die Art der Änderung, die wir in einer Nation zulassen können, daß wir noch Land des Gottes benennen.

German -> English

The agenda Clinton and Clinton would impose America – abortion of need -, a litmus test for the highest court, homosexual of rights, discrimination against pious schools, women in the fight – which is the change, completely quite. But it is not the kind of the change America wishes. It is not the kind of the change America necessities. And it is not the kind of the change, which we can permit in a nation that we designate still country of the God.

A few clunky phrases, but not too bad for free software on-line. Perhaps someone with access to a for-pay translation service can see what one gets when one ponies up some do-re-me, instead of just going for the freebie.

[Sattua]

First off, Babelfish isn’t anything like the state-of-the-art in translation. If you pay for a system, it will be better–if you pay even more it will be even better–and if you are the military of a rich nation, it’ll be the best available, and will be semantically intact though probably inelegant and obviously machine produced.

One will never be able to translate from language A to language B and then back to A and get exactly the same words. I, as a computational semanticist, am pretty sure that as soon as researchers and major funding sources stop acting like deer in headlights about it, though, there will be systems that can produce texts that will be mistaken as natural language produced by a native speaker. I’m sure that these systems will be able to produce texts in the correct registers, too–newspaper speak, politician speak, academic speak, whatever.

I actually work on ontological modeling for such translation systems, though we’ve gotten sidetracked into other applications in the last couple of years… because the military isn’t interested in improving the translation software it already has.

[Walter_Windchill]

Eventually (probably not that far in the future) translation software will leverage web 2.0 type operations to let users improve translations in context that will become part of the database, and the software will be self-improving.

I don’t know if there’s a such thing as “perfect” in translation, though… even when it’s done by humans.

[iamthewalrus_3]

The idea of massive-data driven translation algorithms that Sal suggests was mentioned by Robert Cringely at least once as a possible reason for Google’s attempt to digitize lots of paper books and make them searchable.

[Valteron]

As it happens, I work as a translator (which means written) French-to-English and English to French.

I also work as a simultaneous interpreter, which means listening to a person speak one language and immediately repeating it in another, at conferences and other meetings.

I have seriously wondered if software could ever do a decent job with written translation. It is possible, I suppose, but it seems to me that it would take a greeat deal of ability by thge computer to deal with many, many possibilities.

This may seem like an odd way to prove my point, but look at some of the gems probably produced by** human ** translators (NOT SOFTWARE, although I wonder if one or two might not be computer-generted) and then ask yourself how a computer program could be designed to avoid this: After you have had a good laugh, let’s look at some of the reasons for the errors.

In a Tokyo Hotel: Is forbidden to steal hotel towels please. If you are not a person to do such a thing is please not to read notis.
In a Bucharest hotel lobby: The lift is being fixed for the next day. During that time we regret that you will be unbearable.
In a Leipzig elevator: Do not enter lift backwards, and only when lit up.
In a Belgrade hotel elevator: To move the cabin, push button for wishing floor. If the cabin should enter more persons, each one should press a number of wishing floor. Driving is then going alphabetically by national order.
In a Paris hotel elevator: Please leave your values at the front desk.
In a hotel in Athens: Visitors are expected to complain at the office between the hours of 9 and 11 A.M. daily.
In a Yugoslavian hotel: The flattening of underwear with pleasure is the job of the chambermaid.
In a Japanese hotel: You are invited to take advantage of the chambermaid.
In the lobby of a Moscow hotel across from Russian Orthodox monastery: You are welcome to visit the cemetery where famous Russian and Soviet composers, artists, and writers are buried daily except Thursday.
In an Austrian hotel catering to skiers: Not to perambulate the corriders during the hours of repose in the boots of ascension.
On the menu of a Swiss restaurant: Our wines leave you nothing to hope for.
On the menu of a Polish hotel: Salad a firm’s own make; limpid red beet soup with cheesy dumplings in the form of a finger; roasted duck let loose; beef rashers beaten up in the country people’s fashion.
Outside a Hong Kong tailor shop: Ladies may have a fit upstairs.
In a Bangkok dry cleaners: Drop your trousers here for best results.
Outside a Paris dress shop: Dresses for street walking.
In a Rhodes tailor shop: Order your summers suit. Because is big rush we will execute customers in strict rotation.
From the Soviet Weekly: There will be a Moscow Exhibition of Arts by 150,000 Soviet Republic painters and sculptors. These were executed over the past two years.
A sign posted in Germany’s Black Forest: It is strictly forbidden on our black forest camping site that people of different sex, for instance, men and women, live together in one tent unless they are married with each other for that purpose.
In a Zurich hotel: Because of the impropriety of entertaining guests of the opposite sex in the bedroom, it is suggested that the lobby be used for this purpose.
In an advertisement by a Hong Kong dentist: Teeth extracted by the latest Methodists.
In a Rome laundry: Ladies, leave your clothes here and spend the afternoon having a good time.
In a Czechoslovakian tourist agency: Take one of our horse-driven city tours - we guarantee no miscarriages.
Advertisement for donkey rides in Thailand: Would you like to ride on your own ass?
In a Swiss mountain inn: Special today – no ice cream.
In a Bangkok temple: It is forbidden to enter a woman even a foreigner if dressed as a man.
In a Tokyo bar: Special cocktails for the ladies with nuts.
In a Copenhagen airline ticket office: We take your bags and send them in all directions.
On the door of a Moscow hotel room: If this is your first visit to the USSR, you are welcome to it.
In a Norwegian cocktail lounge: Ladies are requested not to have children in the bar.
In a Budapest zoo: Please do not feed the animals. If you have any suitable food, give it to the guard on duty.
In the office of a Roman doctor: Specialist in women and other diseases.
In an Acapulco hotel: The manager has personally passed all the water served here.
In a Tokyo shop: Our nylons cost more than common, but you’ll find they are best in the long run.
From a Japanese information booklet about using a hotel air conditioner: Cooles and Heates: If you want just condition of warm in your room, please control yourself.
From a brochure of a car rental firm in Tokyo: When passenger of foot heave in sight, tootle the horn. Trumpet him melodiously at first, but if he still obstacles your passage then tootle him with vigor.
Two signs from a Majorcan shop entrance: English well speaking / Here speeching American.

Now if you can stop laughing for a bit, consider the reasons for these “bad” translations. Most are comprehensible, but many are based on a failure to realize shades of meaning and double meanings.

“The Health inspector was in our restaurant and he passed us!” is perfectly good standard English. But “The manager has personally passed all the water served here” is hilarious.

Similarly, “If this is your first visit to the USSR you are welcome to it.” fails to take into account the colloquial meaning of “you are welcome to it”. If that expression did not exist to mean “take it, I don’t want it” there would be **nothing wrong ** with that translation. There would be nothing wrong with a similar grammatical construction in English such as “So this is your first visit to our country? I hope you like it.”

When it comes to simultaneous interpreting, where you have to catch moods and shades of meaning, I really wonder what a computer could do. Suppose a person who has just been corrected at a meeting throws his hands up in the air and says “Well, luckily for the rest of us, you are here to setr us straight on our profession.” This is clearly a sarcastic and hostile comment and as an interpreter, I have about half a second to think of an appropriate formulation in the other language and to say it, while listening to the next incoming sentence.

Finally, my favourite whacky translation which does not appear above: “Fall is a very gay season in Bavaria. It is when the firemen and policemen get together to hold their balls.”

[BrainGlutton]

I think I read some of those in Reader’s Digest once . . .

[carterba]

It will be perfected someday. It will take massive parallelization to do it as well as a human. In my opinion.

However, if somebody proves tomorrow that P = NP, then the day after that translation software will be perfected.

[Valteron]

@BrainGlutton

I think I read some of those in Reader’s Digest once . . .

So what is your point? This list has been around, and it may well have been in RD.

[Bryan_Ekers]

@Valteron

Now if you can stop laughing for a bit -

Done.

Translations won’t be perfect until languages are “perfect”, but I predict dramatic improvements in heuristics. I remember a novel called The Tomorrow File in which a mind-reading computer needed to be programmed with clichés. I figure a future computer translator will have a database of a few thousand common phrases and the speed to check them, making translation of all but the most flowery and symbolic of text translatable. Certainly business documents will be first. I wouldn’t bet on Shakespeare anytime soon.

[BrainGlutton]

@Bryan Ekers

I wouldn’t bet on Shakespeare anytime soon.

I had to try this, just out of sheer perversity . . .

@William Shakespeare

To be, or not to be- that is the question:
Whether ‘tis nobler in the mind to suffer
The slings and arrows of outrageous fortune
Or to take arms against a sea of troubles,
And by opposing end them. To die- to sleep-
No more; and by a sleep to say we end
The heartache, and the thousand natural shocks
That flesh is heir to. ‘Tis a consummation
Devoutly to be wish’d. To die- to sleep.
To sleep- perchance to dream: ay, there’s the rub!
For in that sleep of death what dreams may come
When we have shuffled off this mortal coil,
Must give us pause. There’s the respect
That makes calamity of so long life.
For who would bear the whips and scorns of time,
Th’ oppressor’s wrong, the proud man’s contumely,
The pangs of despis’d love, the law’s delay,
The insolence of office, and the spurns
That patient merit of th’ unworthy takes,
When he himself might his quietus make
With a bare bodkin? Who would these fardels bear,
To grunt and sweat under a weary life,
But that the dread of something after death-
The undiscover’d country, from whose bourn
No traveller returns- puzzles the will,
And makes us rather bear those ills we have
Than fly to others that we know not of?
Thus conscience does make cowards of us all,
And thus the native hue of resolution
Is sicklied o’er with the pale cast of thought,
And enterprises of great pith and moment
With this regard their currents turn awry
And lose the name of action.- Soft you now!
The fair Ophelia!- Nymph, in thy orisons
Be all my sins rememb’red.

Using Babelfish to translate the above into Japanese (“All your base are belong to us!”) and back again to English, you get:

Because of a certain, or because it is not, it is question: ’ The topping lift and the arrow of a nobler inhuman good fortune it suffers with heart, or or trouble, and tis which takes the arm vis-a-vis the sea by in the opposite direction those finishes thing, how? In order to die in order above this to sleep; And we finish thing mental agony and thousand natural impacts where the meat is to the successor with the sleep which is said. ’ From the heart which is Tis wish’d consummation. In order to die in order to sleep. In order perchance to sleep in order to look at dream: There is ay and friction! Whether perhaps, when we mix this death escaping from the coil separated any dream comes, for with the sleep of death, we must give pause. There is a point which makes the disaster of such long life. When perhaps, he makes him himself naked bodkin and that quietus, it withstands the rod because someone’s, ’ the wrongdoing of the pressure person of time and Th, is proud and contumely of the high person, despis’d love and lag of law, the intense of insolence of the office scorns painfully, Th does not have the acquisition which it repulses the value of the patient advantage '? Who these fardels the life which becomes fatigued, it groans under that, it withstands the fact that you sweat, on the other hand us there are no we in us where the traveler does not reset the will of perplexity from some dread bourn after the death, has those sicknesses withstand besides the fact that you know, because it has from growing, it is bitter the country of undiscover’d? Therefore conscience to make our all cowardly people, in this way, as for hue of original product of decision there be o’er where casting ones whose sicklied thought is thin have been attached, enterprise and the flow of the large medulla being warped, turn, name of behavior. At the time of this point which it loses - soft now! As for fair オフェリア! - The nymph of thy orisons my crime entirely is rememb’red.

Endless hours of good clean fun!

[TokyoBayer]

@Valteron

This may seem like an odd way to prove my point, but look at some of the gems probably produced by** human ** translators (NOT SOFTWARE, although I wonder if one or two might not be computer-generted) and then ask yourself how a computer program could be designed to avoid this: (standard list of mistranslations deleted)

From snopes,

Various lists of “real,” unintentionally humorous statements and questions from the mouths of doctors, lawyers, schoolchildren, college students, tourists, and John Q. Public are a staple of net humor. Some of the same lists have been circulating on the Internet for years, and many of them were making regular appearances in newspapers and magazines for decades before that. Are any of them real, or are they all merely clever inventions? The truth is nearly impossible to determine after all this time. All that matters is that they still tickle our funny bones.

Your list wasn’t one this, but the one on the insurance forms is:

Coming home I drove into the wrong house and collided with a tree I don’t have.

The other car collided with mine without giving me warning of its intention.

I thought my window was down, but I found it was up when I put my head through it.

I collided with a stationary truck coming the other way.

A pedestrian hit me and went under my car.
etc.

As Snopes says, there’s no way to tell if these lists are true or not, but I’ve seen a number of the “mistranslations” on various lists over the last 25 years. You can look at a number of the translations, and see that tese are more likely to be created:

In a Tokyo Hotel: Is forbidden to steal hotel towels please. If you are not a person to do such a thing is please not to read notis.

This does not make sense in Japanese. One does not say “forbidden” and “please” in the same sentence in Japanese, any more than one would say this in English. Likewise, the second sentence does not make sense in Japanese either.

From a brochure of a car rental firm in Tokyo: When passenger of foot heave in sight, tootle the horn. Trumpet him melodiously at first, but if he still obstacles your passage then tootle him with vigor.

the source for this and some others on the list seems to be from Anguished English by Richard Lederer. I’ve seen a number of strange translations, but this list is more for entertainment value than a reflection of reality.

For the OP, one of the largest problems with machine translation is context. Unfortunately, there usually isn’t a simple one-to-one exchange between words. When I started working in Japan in 1990, I was working for a publishing / documentation company and started a translation division. While I didn’t translate myself, I hired the outside translators and evaluated their work.

We looked into machine translation then, and found it dismal. The “I see” moment for me, was evaluating a sample translation. The English started off with something like “Introduction” and the machine had translated “introduction” as “doonyuu” which is the word used for “introducing” a new technology. Most humans whould be able to quickly tell that the original English was the introduction to a book, which would be “maemuki” in Japanese. Another possible choice would be “shookai” which is used for an introduction to a friend.

A good translation would be impossible without Strong AI. However, machine translations work well for manuals which have set sentences and set vocabulary in limited context.

[BrainGlutton]

@carterba

However, if somebody proves tomorrow that P = NP . . .

[Sattua]

@TokyoPlayer

A good translation would be impossible without Strong AI.

And there, sir, you are wrong. Weak AI is plenty good enough.

[Capt.Ridley_s_Shooting_Party]

However, if somebody proves tomorrow that P = NP, then the day after that translation software will be perfected.

No. It’s perfectly possible that any proof that P=NP (or otherwise) will be non-constructive. Just knowing that P=NP doesn’t give us a polynomial time algorithm for NP-hard problems.

Google came to our university last year and did a technology presentation. They’re a big funder of machine learning, and work on automated translation systems as well. They said that the problem is twofold: computational power and the amount of data to work from. Techniques that appear to be useless on small corpuses suddenly become very effective when you give it a corpus the size of Google’s database, for instance.

[Valteron]

[QUOTE=TokyoPlayer]
From snopes,
This does not make sense in Japanese. One does not say “forbidden” and “please” in the same sentence in Japanese, any more than one would say this in English. Likewise, the second sentence does not make sense in Japanese either.

QUOTE]

Thanks for the contribution, Tokyo Player. I did indeed feel that some of the allaeged Japanese translations did not seem “real”. It occurred to me that even a people as polite as the Japanese would see that one does not say “forbidden” and “please” in the same sentence.

But I get the feeling SOME of these could be real translations. Especially things like “You are welcome to it.” After all, it is perfectly correct to say, “How was your trip? I hope you enjoyed it.”

And I have personally seen some really funny translations, so I know they DO exist. "Made in Turkey"on an item of clothing was translated as “Fait en dinde”. The French for the country of Turkey is “Turquie”. “Dinde” means the Turkey Bird, the fowl that people eat. So they were saying that the clothing was “made out of turkey meat”.

I remember another package where salted nuts were translted as “noix sales”. They should have written “noix salés”. With the accent, it is pronounced “sah-lay” and means salted. Without the accent it is pronounced “sal” and means “dirty nuts”.

I am very interested in your contributions, Tokyo Player. (I almost said, “I am very interested in you” but suddenly realized that that almost has a romantic/sexual connotation…see how subtle language is?)

Are you a Japanese-English Translator? I am a translator and simultaneous interpreter working in English and French. Is Japanese your mother tongue?

Is it true that Japanese has a special verb form that can be used only in speaking to the Emperor? Or is that an urban legend?

Your comment about “introduction” in Japanese reminded me of something funny in French. The verb “introduire” meaning “to introduce”. But it cannot ever be used to mean social introduction of one person to another, as in English. In French, it means literally to insert something into something else. So If you say to a woman that you wish to introduce a man to her, you are literally saying that you want to insert the man into her! The verb to use is “présenter”.

[Valteron]

I just tried something with Babel Fish. It is all English-to-French.

I tried:

I want an apple. Result: 100% perfect translation.

**I would like an apple, please. ** Still 100% perfect result.

Boy, an apple would hit the spot right now. I’m on a diet. Mostly fruit and vegetables. Result: The first sentence is garbage. It begins “Garçon” and then goes on to say something in French that means something like " an apple strike a blow to the stain at this moment". The second and third sentences are comphrensible but need a little tweaking to make them really natural French.

Now, looking just at the problem of “Boy”, I would like to know how if a computer could really handle this.

The computer program would have to recognize that the word “boy” at the beginning of an English sentence followed by a comma has three possible meanings in English. Then it would have to have ways of rendering them into the equivalent in each tarhet language, French, Russian, Chinese, etc.

1. A way of adressing a young male who actually is a boy, as in “Boy, bring the horseshoes here!” said the blacksmith to his young appentice.

2. An exclamation that has nothing to do with boys and really is the equivalent of “gee”.

3. A disrespectful and racist way in which some southern US whites addressed black men in the past. If you are translating a novel in which the sentence is “Boy, get your ass over here…” shouted the klansnman to Jim as he brandished a gun" I would actually in this context consider translating “Boy” as “Hé, le nègre. … . . .”. This means literally "Hey, nigger. . . . ". Because in the context of this scene, this horrible and racist expression in French would convey the hatred and denigration inherent in the statement. Since the French often say “garçon” to call a waiter, it would simply** not do ** in this context.

Now, I am not an expert in software and AI. It sounds like some of the posters here know something about it. So, on just that one word, “Boy,” would it seriously be possible to cover all bases and get an accurate and APPROPRIATE translation every time from a computer?

[Voyager]

I’m not buying that advances in computing power are going to solve the problem - but I remember reading in USA Today that the just introduced 80386 was so powerful that AI was now a done deal. :rolleyes:

As Valteron’s examples (real or made up) show, you need both semantic and cultural context to do good translations. (Look how Kennedy screwed up the Berlin speech.) That’s not going to happen automatically with raw computing power. Sattua, I’d guess that computational semantics has advanced quite a bit since I studied it 30 years ago or so. Do you know how good the DoD (or NSA, even better) translators are? Obviously not so good that the military can close the school in Monterey.

I don’t buy that a raw data-driven search is going to hack it. As the examples show, you get an explosion just in the meanings of various words. You still need a reasonable degree of semantic knowledge to sort through the search results. I’m also not convinced that every sentence worth translating into appears in a book already.