Real-Time Computer-Based Translation in Iraq

[TheBORG] writes "The U.S. military has been testing software on laptops that translate English to Arabic and Arabic to English to have conversations with Iraqis without the need to have a Arabic linguist on hand. 'This year the military's Joint Forces Command has been testing laptops with such software in Iraq. When someone speaks into a microphone attached to the computer, the machine translates it into Arabic and reads that translation aloud over the PC's speakers. The software then translates the Arabic speaker's response and utters it in English.'" (See this related story from last year about this daunting machine-translation task.)

I never thought before I spoke before...

[superlou (991322)]

With some luck it will translate my banal whining into cutting social commentary.

Re:I never thought before I spoke before...

[doti (966971)]

I will not buy this record, it is scratched.

Re:

[stunt_penguin (906223)]

For the clueless and impatient [youtube.com]

Re:I never thought before I spoke before...

[cluckshot (658931)]

The mechanistic translation of Arabic into English will further blind US troups to the social and other conditions in Iraq. They will get back the techical translation but none of the meaning of the speakers. As such the failure to have good translators will be a serious problem. I suppose the best example of this is in a silly film "Mars Attacks". "Don't Panic, We are your friends." ---> Time to start panic.

Re:I never thought before I spoke before...

[smitty_one_each (243267)]

Your assertion that US troops aren't going to bother to learn anything whatsoever about Iraq may hold true for a small minority.
However, the gadget will likely have a catalytic effect: given something that can ease some of the basic communication challenges, the bulk of the troops will likely become somewhat conversational rather quickly.
I base my remark on personal experiences of the US Navy in Japan and the Philippines--I wouldn't expect Iraq to be substantially different.
Your point about the need for good translators is not without merit, but the pessimistic tone elicits a yawn, sir.

Re:

[Xyrus (755017)]

Experiment:

1. Find an article, any article, that's written in english.
2. Go to babelfish or some other translation site and translate it into another language.
3. Now, translate that back into english.
4. Endless hours of fun, especially while drinking.

This will either piss the Iraqis off more, or make thm laugh so hard that they'll stop bombing each other.

~X~

Computer: AL YOUR QAEDA ARE BELONG TO US

[tepples (727027)]

From the article:

MASTOR's accuracy is not perfect, but "you can communicate a concept and you can elicit a response from someone"

Given that "Al Qaeda" is Arabic for "The Base", and machine translation is notorious for its poor grasp of grammatical structure and homonyms, are soldiers going to have to deal with outputs like "AL YOUR QAEDA ARE BELONG TO US"?

Re:Computer: AL YOUR QAEDA ARE BELONG TO US

[wired_LAIN (974675)]

Dear al-Qaeda, let's so double the killer delete wmd all

Re:Computer: AL YOUR QAEDA ARE BELONG TO US

[From A Far Away Land (930780)]

I bet the reason the Bush admin's estimates for civilian casualties is so low, is due to the numbers being translated from Arabic numbers into whatever number system we use.

What? Oh damn, never mind.

Translation through a computer in life or death situations makes me sad inside.

Re:Computer: AL YOUR QAEDA ARE BELONG TO US

[quigonn (80360)]

On a more serious sidenote, it is indeed interesting to know that the "Arabic numerals" are actually called "Indian numerals" in Arab countries.

Re:Computer: AL YOUR QAEDA ARE BELONG TO US

[pluther (647209)]

Even stranger, Iraq, like most middle-eastern countries, doesn't actually use Arabic numbers themselves.

They use Persian numbers.

Re:

[mpe (36238)]

On a more serious sidenote, it is indeed interesting to know that the "Arabic numerals" are actually called "Indian numerals" in Arab countries.

Arabic text usually uses Hindi numerals. Regardless of the numerals used, numbers in Arabic read from left to right.

Re:

[cvd6262 (180823)]

Obviously, Slashcode doesn't handle Arabic script well.

Re:Computer: AL YOUR QAEDA ARE BELONG TO US

[rolfwind (528248)]

I can't but wonder if it wouldn't be more effective to have a central call center for arabic/english translators for the soldiers to call when they to facilitate communication. Perhaps they could install a videophone system to provide the visuals. How feasible is this?

That way you keep the translators out of harm's way, making them easier to retain/hire.

Big worry

[Beryllium Sphere(tm) (193358)]

Arabic is even worse than most human languages for being contextual and ambiguous. It's superb for writing poetry but betting lives on translating it automatically?

Re:Big worry

[CHESTER COPPERPOT (864371)]

Good point. And like the article states ... it hasn't been tested in a real setting yet. How's it going to go translating a screaming, aggressive arabic speaker? What about a stressed out, crying arabic speaker that has just had his family shot and/or blown up? Sounds like just another technological band aid to something that is better off solved with investing in real linguists.

Re:Big worry

[lawpoop (604919)]

Yeah, sounds like more failure-prone technological solutions to the war on terror, like gait recognition, face recognition, headline scanning, which all are failure-prone, technological solutions to a human problem. What we really need is people skills, like actual fluent translators, experts with experience, covert agents, and inside guys.

Re:Big worry

[StikyPad (445176)]

What about a stressed out, crying arabic speaker that has just had his family shot and/or blown up?

Not to mention a panicked, confused english speaker who just had his leg blown off by an IED.

Re:

[Sancho (17056)]

So how do you translate "DO NOT RUN WE ARE YOUR FRIENDS" into Arabic?

Re:Big worry

[EugeneK (50783)]

How to do you translate "PLEASE GET THE FUCK OUT OF OUR COUNTRY RIGHT NOW" into English?

Oh wait, that's already in English...

Re:Big worry

[glowworm (880177)]

How to do you translate "PLEASE GET THE FUCK OUT OF OUR COUNTRY RIGHT NOW"

I'm not sure about the IBM solution but good old google gives a phrase that when retranslated back to English reads as "Hope of the exploited in our country right now"*. I don't see this being any real use except for the most basic translations, like which way to the mosque, or do you need a doctor. After all "The spirit is willing but the flesh is weak" :)

* I would paste the google translated Arabic, but for some reason /. seems to delete arabic unicode from it's posts. Try it and see.

Obligatory Python...

[tetsu96 (790788)]

...It is not before one hapless American, searching for the nearest terrorist, blurts out to a startled passerby "Please fondle my buttocks"

Subtitles

[seanadams.com (463190)]

Why don't the iraqis just use subtitles? [google.com]

Reminds me of old experiment

[Ruvim (889012)]

Reminds me of experiment I read about in old computer book... Program was created to translate from English to Russian and back. As a test, a phrase "Time flies like arrow" was translated to Russian and then back to English. It came back as "There are types of flies, called 'Time Flies' that enjoy eating arrows.

Wrong

[camperdave (969942)]

The phrase that was translated English->Russian->English was "The spirit is willing but the flesh is weak" and it came back as "The wine is good but the meat is rotten".

The phrase "Time flies like an arrow. Fruit flies like a banana." is a Groucho Marx quote. I'm not sure of the original context, but it is an example of how English (or any other natural language) is notoriously difficult to handle. For example, the sentence "Time flies like an arrow." may be justifiably interpreted in a variety of ways:

• time moves quickly just like an arrow does;
• measure the speed of flying insects like you would measure that of an arrow - i.e. You should time flies like you would time an arrow.;
• measure the speed of flying insects like an arrow would - i.e. Time flies in the same way that an arrow would time them.;
• measure the speed of flying insects that are like arrows - i.e. Time those flies that are like arrows;
• a type of flying insect, "time-flies," enjoy arrows (compare Fruit flies like a banana.)

May I be the first to say

[snuf23 (182335)]

My hovercraft is full of eels.

I used to work for a translation company and I've seen how much confusion can arise from even human translation, it makes me wonder really how prone to error this will be.

Re:May I be the first to say

[snuf23 (182335)]

Well now that was the poorest troll I've seen in a long time.

My job wasn't to translate. I produced translated websites and printed documents. The company I worked for was sort of a translation broker. We worked with in country translators across the globe. People who have never worked in the industry seem to think it should be simple - it isn't. Consider some of the complications:

1. Regional dialects - i.e. Canadian French is not identical to what is spoken in France. Same with English in the US vs. UK.
2. Language specific to a certain domain, such as medical, technical or legal.
3. Slang meanings
4. Humor/sarcasm
5. Analogies and metaphors

Consider a reference to Bible parable (say Noah and the Ark) sure most people in the US would understand it - but would you understand the relevance of a reference to a parable from the Koran?
We frequently had professional translators arguing amongst themselves as to the proper wording to use.

Re:May I be the first to say

[LordEd (840443)]

For those who do not understand the reference, the expression "my hovercraft is full of eels" is part of the hungarian translation book sketch from Monty Python's Flying Circus. It happens to be here [youtube.com] on youtube.

Or in text from http://bau2.uibk.ac.at/sg/python/Scripts/TheHungar ianPhrasebookSketch [uibk.ac.at]:

A Hungarian tourist (John Cleese) approaches the clerk (Terry Jones). The
tourist is reading haltingly from a phrase book.

Hungarian: I will not buy this record, it is scratched.
Clerk: Sorry?
Hungarian I will not buy this record, it is scratched.
Clerk: Uh, no, no, no. This is a tobacconist's.
Hungarian: Ah! I will not buy this *tobacconist's*, it is scratched.
Clerk: No, no, no, no. Tobacco...um...cigarettes (holds up a pack).
Hungarian: Ya! See-gar-ets! Ya! Uh...My hovercraft is full of eels.
Clerk: Sorry?
Hungarian: My hovercraft (pantomimes puffing a cigarette)...is full of eels
                      (pretends to strike a match).
Clerk: Ahh, matches!
Hungarian: Ya! Ya! Ya! Ya! Do you waaaaant...do you waaaaaant...to come
                      back to my place, bouncy bouncy?
Clerk: Here, I don't think you're using that thing right.
Hungarian: You great poof.
Clerk: That'll be six and six, please.
Hungarian: If I said you had a beautiful body, would you hold it against me?
                      I...I am no longer infected.
Clerk: Uh, may I, uh...(takes phrase book, flips through it)...Costs six and
              six...ah, here we are. (speaks weird Hungarian-sounding words)
Hungarian punches the clerk.

Meanwhile, a policeman (Graham Chapman) on a quiet street cups his ear as if
hearing a cry of distress. He sprints for many blocks and finally enters the
tobacconist's.

Cop: What's going on here then?
Hungarian: Ah. You have beautiful thighs.
Cop: (looks down at himself) WHAT?!?
Clerk: He hit me!
Hungarian: Drop your panties, Sir William; I cannot wait 'til lunchtime.
                        (points at clerk)
Cop: RIGHT!!! (drags Hungarian away by the arm)
Hungarian: (indignantly) My nipples explode with delight!

All your base all over again

[davidwr (791652)]

Soldier: Surrender now, we have you surrounded.
Computer: #All your base are belong to us#
Iraqis: [hysterical laughter]

I bet it doesn't work too well.

[Reality Master 201 (578873)]

You can probably have unbelievably simple conversations, like

"Do you want to kill me?" "No."

And for anything approximating a normal conversation, it's utterly fucking useless. Also, for the times when you actually need a very urgent, very good understanding of the language to prevent a lot of trouble, I bet it's beyond worthless.

At present, and for the forseeable future, there's no adequate substitute for humans that speak the language. I realize we throw Arabic speakers out of the military because they're gay and all, but maybe we could make an exception because their skills are necessary at present. No computer translation system is adequate for usage in a live military operation.

Oh, and IACL (I am A Computational Linguist).

Re:I bet it doesn't work too well.

[SuperBanana (662181)]

I realize we throw Arabic speakers out of the military because they're gay and all, but maybe we could make an exception because their skills are necessary at present.

Do you really think there are enough a)Arabic-speaking b)openly c)gay soldiers in the military, to make a difference? I bet you could count them all on two hands.

I think the military's policy is pretty stupid. However, I think if soldiers truly cared about "serving their country"(in quotes because I'm tired of "fighting in Iraq" = "defending freedom" in public discourse) above all else, they simply wouldn't tell the military they were gay. I'd also suspect that those that DO care about fighting for their country simply DO clam up and get the job done.

Armed Forces Desperate for Arabic Speakers

[PChuck (1012975)]

Of the more than 11,000 service members who have been dismissed under the U.S. military's so-called "Don't Ask, Don't Tell" policy, about 800 had "critical abilities, including 300 with important language skills. Fifty-five (55) were proficient in Arabic." (Emphasis added). This information comes from the blog "Shakespeare's Sister," which can be found at http://shakespearessister.blogspot.com/2006/07/arm y-dismisses-arabic-linguist-for.html [blogspot.com]; this blog also has been quoted with approval by the nationally known non-profit group, Servicemembers' Legal Defense Network. See, e.g., http://freedomtoserve.blogspot.com/ [blogspot.com].

The dismissal of this many Arabic-speaking military linguists *has* had an enormous impact on the military's ability to function efficiently in the Middle Eastern theatre. Believe it or not, the Army is now recruiting linguists on Craigslist with the following ad:

Your primary responsibility will be to interpret Middle-Eastern languages into English to help with rebuilding efforts. On a day-to-day basis, you might:

* Provide records of foreign language communications * Translate, transcribe or produce summaries of foreign language materials into English or target languages * Identify the language spoken in an assigned geographic area * Scan written foreign language material for key words and indicators * Translate written and interpret spoken foreign language material to and from English, while making sure to preserve the original meaning * Translate and transcribe Middle-Eastern language TV and radio broadcasts into English * Translate foreign books and articles describing foreign equipment and construction techniques

LANGUAGES

The Army Translator Aide Program specializes in the following languages:

Arabic

* Algerian * Egyptian * Gulf-Iraqi * Jordanian * Lebanese * Libyan * Maghrebi * Modern Standard * Moroccan * Syrian * Sudanese * Tunisian * Yemeni

Other

* Pushtu-Afghan * Pushtu/Pashto/Pachto * Kurdish * Kurdish-Behdini (Kurmanji) * Kurdish-Sorani * Persian-Afghan (Dari) * Persian-Iranian (Farsi)

I have also seen a classified ad from the Washington Post from the U.S. military, seeking Arabic linguists (among others) for training and employment. Clearly, discharging all those Arabic-speaking members of the military because of their sexual orientation was foolish, to say the least.

As for the argument that these soldiers should just "clam up" and "not tell the military" they are gay, many LGBQ people would love to serve their country this way. However, you should go to the previously mentioned Servicemembers' Legal Defense Network website at http://www.sldn.org/ [sldn.org] to read about the everyday harassment, "witch hunts," and physical and emotional violence inflicted upon gay people by the military in violation of its own "Don't Ask, Don't Tell" rules. I think you would be extremely surprised to find out how many servicemen have been killed in the past five years by members of their own divisions/units; the Armed Forced do not exactly issue press releases every time something like this happens. The bottom line is that the vast majority of LGBQ soldiers are forced out against their will, as they try to be quiet and inconspicuous and to serve their country.

What won't be making it into translations

[BeeBeard (999187)]

1. Inflection and emphasis of some words over others

This is very important. Ever have somebody tell you "It's not what you say, it's how you say it"? It's true.

2. Colloquial expressions and figures of speech.

Right now, I'm looking at this book [amazon.com] filled with conversational Arabic expressions I picked up in the U.A.E., most of which make absolutely no sense when translated into English. Do you know what "The son of a duck is a floater" means? Neither will U.S. troops or this device.

3. Body language

Many Arabic speakers in particular gesticulate while they speak. It is just part of their cultural identity and often, the body language is just as important as what is being said. U.S. troops in the field won't understand the importance of what they see, let alone what they hear, and this device certainly won't help them with that either.

This is just what I could think of in a minute or so. I'm sure there are many more fundamental problems with using the translation device. Note that with a real live translator, most of these problems are avoided. If the U.S. military kept its Arabic translators in their ranks instead of firing them based on their sexual orientation [wikipedia.org] then maybe they wouldn't have to resort to these ridiculous devices.

Re:

[LS (57954)]

Your points are well taken, but I think they are overstated. Let's look at them:

1. Inflection and emphasis of some words over others

This problem is mitigated by several factors. Speakers and listeners will likely be aware that the emphasis is lost, and will probably speak slowly and evenly, and listen more carefully to what is being said. Most anyone would not take offense or grossly misunderstand the translation unless it was in perfect Arabic or English. It won't be - it's going to sound like broken A

"I'd rather my son be dead than gay!"

[BeeBeard (999187)]

Oh please. Bleu Copas was by no means "flaming". Skeptical? See if you can "tell" if you didn't know to "ask" when you check out The Daily Show's hilarious take on the incident. [youtube.com] Since anti-gay conservatives are as desirous of pragmatic thinking as they are 16 year old boys, let me ask you this: What do you think is more important, the safety of U.S. troops or the fact that the straight-acting man who is interpreting their words prefers men?

The only people ...

[Woldry (928749)]

... who think that computers are anywhere near ready to do realistic translation are people who have no concept whatsoever how complex human language really is. We will never have a working, reliable computer translation while we are still unable to fully explain or describe the rules of our own languages. Language is remarkably fluid and idiosyncratic, and the rules change not only from language to language, and from dialect to dialect within each language, but from individual to individual, and from utterance to utterance with each individual. So far, we have yet to invent a computer complex enough for the pattern-recognition skills necessary even to parse a majority of sentences correctly, much less decode them and then reconstruct them in a different language altogether.

None of this is to say that we can't ever do it, or that we shouldn't attempt. But the people who think it's possible with today's computer technology really don't understand the complexity of the problem.

The process of translation

[demondawn (840015)]

As someone who has studied translation (Japanese/English) at the University level, I can tell you that interpreting in real-time in a heavily context-sensitive language like Japanese or Arabic is an incredible challenge for even people who have spoken both languages for -decades-. When tiny grammatical changes can affect the entire meaning of a sentence, and voice recognition is by no means perfect, and homonyms come into play, the entire process is incredibly difficult. On a -personal- level, as someone who studies languages and desires a career in either teaching or translation, I'm worried not so much that it's replacing the human element, but that people believe it can be used without human intervention. The difficulty of interpretation and translation (this would be the former, for the record) is related to the distance, in linguistic construction, between the two languages, and few languages are further apart than English and Arabic. The increases in accuracy of machine translation also grow logarithmically; the more development that comes out of it, the less benefit you get. What I do believe we should be doing is investing money in both language education AND language technology. I also have a bit of a bitter taste in my mouth regarding the fact that the U.S. military is discharging qualified linguists that happen to be homosexual, but then I say that as a homosexual language student that wanted to join the military when I graduated. Now I'm looking to move to Canada.

Ya but...

[fahrbot-bot (874524)]

will it fit in my ear and does it come with chips [wikipedia.org]?

Hmm, great idea..

[necro2607 (771790)]

OK, I'm pretty sure I'm not the only one who just thought of the inane results of translating things back and forth with Babel Fish [altavista.com] .... This better be some DAMN good translation software.

I can just imagine the "limitation of liability" portion of the end-use agreement from the company that developed the translation software...

Even worse, what happens when some on-the-edge person pulls out a hidden weapon and injures/kills a soldier (or whoever) because of incorrect translation? Oh, is this just part of the "risk of the business"? ...

Colorless

[hey (83763)]

Colorless green dreams sleep furiously.

The man in the uniform said...

[The Famous Druid (89404)]

"My hovercraft is full of eels"

and then poked me with the electric cattle prod again.

I'd tell him what he wants to know, if only I could understand the question. :(

An example

[rbarreira (836272)]

Just to show you guys how good automatic translation is, here is the simplest sentence I've found so far which is translated very badly by google translation (systran, also used in altavista's babelfish):

• "She's dead!" into Portuguese gives "Está inoperante!", into english again gives "It is inoperative!".
• "She's dead!" into Italian gives "È guasto!", into english again gives "It is out of order!".
• "She's dead!" into French gives "Elle est morte !", into english again gives "It died!".
• "She's dead!" into Spanish gives "Ella es muerta!", into english again gives "She is dead!".
• "She's dead!" into German gives "Sie ist tot!", into english again gives "It is dead!".
• "She's dead!" into Arabic gives (something I can't paste here), into english again gives "Are dead!".
• "She's dead!" into Japanese gives (something I can't paste here), into english again gives "She has died!".
• "She's dead!" into Korean gives (something I can't paste here), into english again gives "Her it dies! where".
• "She's dead!" into Chinese gives (something I can't paste here), into english again gives "She's dead!".

Now, being generous while categorizing those results gives:

Complete Success = 2 out of 9 = 22% (Spanish and Chinese)
Almost successfull = 1 out of 9 = 11% (Japanese)
Catastrophic failures = 3 out of 9 = 33% (Portuguese, Italian and Korean)
Serious failures = 3 out of 9 = 33% (French, German and Arabic)

How they get to sell software which fails more than half the times at translating such a simple sentence is truly beyond me...

Re:

[demondawn (840015)]

That's because you make the mistake of assuming that "She's dead!" is a simple sentence. Not only does it imply precedent knowledge on someone's part, which is handled differently in different languages, but it also involves the use of the copula (is), which doesn't always work the same way as it does in English. For example, the French is "Elle est morte!" (subject) (copula) (verb past). On the other hand, Japanese has "kanojo wa shindeiru" (topic) (topic marker) (verb continuative). (As an aside, this is

You chose a relatively difficult sentence

[kahei (466208)]

The sentence you chose is vulnerable to a lot of factors that make translation difficult. It contains a contraction, for one thing. Worse, it contains a pronoun, so all at once it's subject to problems among languages with gendered pronouns versus those without -- that's why the gender gets thrown away in the German, Italian, Portuguese, and French versions. It also contains a past participle predicate, which is another construction that has analogues in many languages but different actual meanings (henc

Re:

[rbarreira (836272)]

You're partly correct, but I'd like to remark at least two things:

• The Portuguese and Italian cases don't fail because of that
• Even if you give it context, it won't necessarily succeed (try "the woman is dead" in portuguese or italian, for example)

Re:Great Idea

[badmammajamma (171260)]

The troops learn simple, common phrases in arabic but that's it. If you actually expect them to learn to speak or read it fluently, then you're expectations are completely unrealistic. Your argument might have some actual bite instead of weak flaimbaitness if you made such a comment about the leadership of the country that sends the troops there in the first place.

Re:Great Idea

[iogan (943605)]

The troops learn simple, common phrases in arabic but that's it. If you actually expect them to learn to speak or read it fluently, then you're expectations are completely unrealistic.

Why is that exactly, though? Some of them must have been there for years by now, how is it they have been completely unable to learn the language?

I remember the same thing happening while I was teaching English in Prague, 99 percent of the americans there simply couldn't learn czech, while a good 80-85 percent of the rest of us did. I spoke better czech after about 3-4 months than most of my american friends, regardless of how long they'd been there. Never mind how well the russians did, most of whom picked it up in weeks or at most a few months (their language obviously being much more similar, but still)

Being an american who spoke the local language was in fact considered extraordinary, and usually these people would be very well known in the expat community.

I have a feeling the soldiers would be more welcome and more accepted by the locals if they at least made a token effort to learn a little bit of the language and try to understand a little bit about local culture and values. Like, you know, read a few books published by iraqis for instance.

Re:

[cliffski (65094)]

Agreed 100%. I think that using technology to translate could be pyschologically, pretty bad. We all know that iraqis and americans are from different countries, but we also know quite a few iraqis can get by in english, as its such a common language. You already have a situation where an arabic guy is talking to some (probably) white guy, whose in uniform and holding a gun. Then you have the situation that he obviously doesnt speak any of your language, now we have a situation where he finds your language

Re:Great Idea

[cvd6262 (180823)]

I remember a news report of US soldiers going door-to-door shaking down a "hostile" Iraqi town. My favorite moment was the first grunt through the door yelling, "Yimshi! Yimshi! Get down! Get down!" as he threw a confused Iraqi to the floor.

I know enough Arabic to know Yimshi means "Walk," or "Leave/Get out of here."

Re:Great Idea

[dwillden (521345)]

Actually we just can't cram soldiers through the Defense Language Institute's Arabic programs fast enough. It takes at minimum a year to achieve a basic conversational proficiency.