lost in translation...

Started by Perkele2012, Thu 05/07/2012 17:10:05

Previous topic - Next topic

Perkele2012

does anyone know of a good translation software or website ( except google translate ) for translating documents from swedish to english...

Perkele2012

GAAAAAAAAAAAH!!! I ( lots of bad words ) hate ( even more words that would shame a group of sailors ) computers right now!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

10 pages of translation down the drain because chrome thought i meant "go to previous page" instead of "erase the previous word"!!!!


Snarky

Probably easier to just do it yourself. Even in the best case, automatic translation is only going to give you a clumsy, often ungrammatical version of what you want to say. That's enough to be able to make sense of something in a language you can't read, or for basic communication, but if you already know both languages it's a waste of time.

Anyway, there's the old Yahoo! Babelfish, currently branded as Bing Translator. And Paralink offers online access to a number of different translation services, though the only one that covers Swedish is Babylon Translator. Finally, Reference.com offers a really basic translator (with a very restrictive character limit).

In my experience, none of these are really competitive with Google Translate, though it's sometimes worth running the same text through multiple ones, because they tend to make different errors. And they all seem to be pretty crappy at Swedish-to-English translation, probably because no one is working on it since Swedes speak English anyway.

Andail

#3
Oh, I believe the Swedish-English (and vice versa) translation service isn't much worse than any other, despite Swedes' alleged fluency in English. My students use it all the time to dodge manual translation, which I routinely reprimand them for.

Perkele2012

#4
Quote from: Snarky on Thu 05/07/2012 17:53:35
Probably easier to just do it yourself. Even in the best case, automatic translation is only going to give you a clumsy, often ungrammatical version of what you want to say. That's enough to be able to make sense of something in a language you can't read, or for basic communication, but if you already know both languages it's a waste of time.

Anyway, there's the old Yahoo! Babelfish, currently branded as Bing Translator. And Paralink offers online access to a number of different translation services, though the only one that covers Swedish is Babylon Translator. Finally, Reference.com offers a really basic translator (with a very restrictive character limit).

In my experience, none of these are really competitive with Google Translate, though it's sometimes worth running the same text through multiple ones, because they tend to make different errors. And they all seem to be pretty crappy at Swedish-to-English translation, probably because no one is working on it since Swedes speak English anyway.

i know it's better to do it yourself but im feeling quite lazy at the moment, and the thought of translating everything makes me even more lazy, lol.
well thanks anyway

EchosofNezhyt

Quote from: Andail on Thu 05/07/2012 18:45:32
students.

?

@ Perkele Google always seemed the best for what I've used it for.

Perkele2012

well the main problem when using google translation ( especially when you´re about to translate near 300 pages ) is that the program has got no clue wich one of the 10 words possible for each word you want to use, so the result will to 80% times lead to quite funny sentences.
( and ofcourse when you these damn electronic devices people call computers ( i call them hellspawn from now on! ) always find some way to **** your work up )

nah, i think it´s better to do it manually anyway, it´s going to take awhile but it will probably be for the best.

EchosofNezhyt

Ah... I've never translated anything I only speak English. :\

Snarky

#8
Quote from: Andail on Thu 05/07/2012 18:45:32
Oh, I believe the Swedish-English (and vice versa) translation service isn't much worse than any other, despite Swedes' alleged fluency in English. My students use it all the time to dodge manual translation, which I rutinely reprimand them for.

I use Google Translate a lot for French and German, and sometimes Spanish, and it performs much better on those than it did on some Swedish texts I just tried. It's probably mainly a matter of the sizes of the dual-language corpora, which are likely to be much bigger for the more prominent languages.

Quote from: Perkele2012 on Thu 05/07/2012 19:26:25well the main problem when using google translation ( especially when you´re about to translate near 300 pages ) is that the program has got no clue wich one of the 10 words possible for each word you want to use, so the result will to 80% times lead to quite funny sentences.

Actually, Google uses the sentence context to distinguish homonyms, and to determine the most appropriate synonym to use. It certainly doesn't always get it right, but it has got some clue.

Anian

#9
You realise that if Google Translate was able to accurately translate 300 pages of text from Swedish or any other language to English, there wouldn't be any need for translators...EVER. I really can't figure out why you are you frustrated by this pretty good software.

And what the hell is wrong with you trying to translate 10 pages before saving the text in Word or Writer or basic Notepad?  (roll) Translate is made to translate web pages so you can surf with ease, and it does a really god job of that. It can translate large chunks of official text, but you still have to rad through to check for faults.
I don't want the world, I just want your half

Perkele2012

Lol.

Because for once id like to be lazy!!!
And i was more frustrated over the situation than on google translator.:) :undecided:

monkey0506

#11
Quote from: Snarky on Thu 05/07/2012 17:53:35Probably easier to just do it yourself. Even in the best case, automatic translation is only going to give you a clumsy, often ungrammatical version of what you want to say. That's enough to be able to make sense of something in a language you can't read, or for basic communication, but if you already know both languages it's a waste of time.

I don't have very much cause to translate things very regularly, but when I have in the past, I always make it a point to translate the translated text back to make sure it still makes sense. I generally won't try to do an automated translation of more than about a paragraph, and sometimes I will break it down to individual sentences or phrases (even rewording the source text as needed) to find the best translation for the actual meaning (using another translator if needed, but generally Google can figure it out). It's a lot more time consuming this way than just plugging in an entire lengthy paper, but it gives much more reliable results, and is still significantly less time consuming than manual translation (especially if you don't know both languages).

Not a flawless method either, but translation as an exact science is debatable to begin with. :P

Quote from: Snarky on Thu 05/07/2012 21:04:09Google uses the sentence context to distinguish homonyms, and to determine the most appropriate synonym to use. It certainly doesn't always get it right, but it has got some clue.

This is why it's better (especially in the case of automated translations) to start with the most complete thought and work your way backwards as needed. ;) Most of the "bigger"/"better" translation tools available online will do exactly this. As you said, they won't always get it right, but that's when you can start narrowing it down (hence why I keep it to small chunks at a time).

Quote from: Frito Master on Thu 05/07/2012 19:07:09Google always seemed the best for what I've used it for.

Quote from: Frito Master on Thu 05/07/2012 19:34:20Ah... I've never translated anything I only speak English. :\

Yes, from what I understand Google's English-to-English translations are generally pretty accurate. :P

Perkele2012

This is the opening in the game/novel Hell Awaits that im trying to translate ( ive been using google translate now ) so is this a good translation or not.
would appreciate your views on it and is open for suggestions that would improve it.

Monday 10th of. May 2010
[The dream]

The pain was creeping, slowly, through the stomach, further around the spine and then up towards his throat.
the vomit was not far away now.
The breathing stopped for a while, it seemed like an eternity.
Heart is pounding hard in the chest and the pulse increases dramatically.
Frenetically, he tries to wipe the blood off his hands, it has dried ...
Now it feels like the tears are eating away the cheeks.
He wants to scream, he wants out of his body, the muscles turns in pain;
the only sound he manages to get out is a silent gasp.

Sweaty and disoriented, he sits up in bed, he has not quite grasped what is real or not.
The hands are still trembling when he reaches for the clock on the bedside table.
[02:46] lights up in red on the alarm clock that stands on the bedside table.
At the edge of the bed is Ellie, his daughter.
She stares at him, she looks sad.

- daddy! I can´t sleep, (she says anxiously)
- Are you having nightmares again?
- Mmmm.
- I know how you feel .
- Come ,lay down here between mother and i.

Jeff hugs his daughter tight and gives her a kiss on the forehead.
- I love you Ellie! Remember that

She did not answer, she had fallen asleep in her father's safe embrace.

Monday 9th of. May 2011

- Dad! Wake up (Ellie whispers in Jeff's ear)
Jeff opens his eyes and looks at Ellie and smiles.
She laughed and smiled at him.
She is so beautiful, like her mother in most of her features, the sun is shining through the window so that every single hair glisten when the sun lights up her golden-blonde little head,
Her green-blue eyes look at him and her gaze cuts through his heart.
He got a kiss on the forehead and then she whispered in his ear.

- I will return soon!
- I love you Daddy!


Jeff felt how he melted inside, that the three words in combination could give so much comfort.

- Iiiighhh .. he noticed how difficult it was to speak.
- I love you Ellie! (he managed to squeeze out at last.)

- WE ARE LOOSING HIM! READY!

The pain in his chest was unbelievable when the syringe settles into the chest and pumping his heart full of adrenaline.
The bright glow of the surgical hall strong lights flicker in his eyes.
A migraine-like pain cut through his head.
He hears a voice say that he will survive.
Then everything turns black.

[The night before.]

The last fluid in the bottle of Jack Daniels slips down throat, his thoughts went back to that day a year ago, tortured him, haunted him.
Everything was about what if he had done so, what if he had not.
Had he not turned around at that moment she had never run out on the street…
The truck driver didn’t stand a chance, just a second earlier he had held her hand, and then…
What if he had ... he could have…
the guilt, the dreams, the eyes, oh my god her eyes.
He started crying.
He had almost been like in a coma for a year now, he did not know what was real anymore, he heard voices, had hallucinations, he never slept more than an hour and usually floored by fatigue.
All that was needed was a glimpse of him and you´d see, his soul was twisted by grief and guilt, his eyes told a story with more agony than anyone has ever been able to tell in words, it´s said that a picture tells more than a thousand words, a pair of eyes tells can tell us even more and also revealing a whole life, a window to your soul.
He had seen her several times, standing in just outside the eyes reaching view, in a street corner, a window or in the middle of a street, she would smile and wave at him.
He had decided that enough is enough now, he could not fight back the tears any longer, he could not be the strong one anymore.
tonight, he would escape all, tonight, everything would stop.
It was just after two in the morning when he sat in the car and drove away.
There was almost no traffic at all and he threw in a higher gear up and the speed was way above the allowed limit.
He saw the railing of the bridge approaching quickly,he did not care, the speed increased even more.
It's almost over, he thought, it will be done quick.
The sound of metal grinding against metal , glass that rips his face to shreds, every muscle in his body tense, getting ready to resist the enormous force that pushes him through the front window.
He sees himself smashing the windscreen with the weight of his body.
He see how the body turns into unnatural movements.
Weightless,
a smile
he sees himself hitting the water and sink ...
He looks at the man sinking, it's him, but someone else.

An indescribable pain cut through his lungs when the body through reflex trying to breathe and draw in the gallons of water that fills his lungs, yet he is calm, the lack of oxygen getting his body to produce quantities of adrenaline which acts as an anesthetic agent. He welcomes the warm darkness that wraps around him.
He has accepted his fate.
It was not his life that passed in revue, it was his daughter's death, which was played over and over again, for him it was like a confirmation that justifies his action.
A darkness sweeps in and embraces him, the eternal darkness, it's over now he thought before the last electrical impulses to what is interpreted as  a consciousness from the brain sends him a last message
A whispering voice echoes in his head.
- Dad ...

Anian

These parts suffered a bit from the translation style:
The pain in his chest was unbelievable when the syringe settles into the chest and pumping his heart full of adrenaline.

The last fluid in the bottle of Jack Daniels slips down throat, his thoughts went back to that day a year ago, tortured him, haunted him.
Everything was about what if he had done so, what if he had not.

But basically from the part [The night before] everything is a bit weird and in places incorrect.
I don't want the world, I just want your half

Tuomas

The whole idea of computer translation intrigues me. Yes, it would take my job from me, but at the same time, it'd be a great breakthrough in linguistics. The problem is, that a language is created by a man, and there's usually no rules to follow, the only way to learn one is to speak one for all your life. You can't really translate a text from one language to another because the whole ideas that the words represent are always different. Basically it's all about analysing the original text and writing down the interpretation.

Now I went to a symposium at our university to listen to an interesting presentation held by a polish researcher, can't quite remember his name though. While the whole academic community seems to have abandoned the idea of translating with a computer, this guy suggested, that a common language was created from all the languages, or the major language families in the world. Now someone asked if he'd count Esperanto as such, but the problem with that was that it was created to be pronounced and not read. Anyway, I suppose he was kind of a programmer trying to find a method to change one language into another, and it would have worked if it weren't for all the exception and the fact that "there are as many languages as there are speakers". Meaning we all have our sociolects etc, children speak differently, everyone has a different register when speaking to his boss or his brother or best friend etc.

A translation from Swedish to English would be extremely difficult to accomplish seeing as how they're two completely different languages. Yes, Swedish has a lot of English words, mainly nouns, but that's about it. Swedish into Danish, Norwegian and maybe even German might work better because they're both germanic languages. But again, there's too many differences. What you would have to do was to change the original input into one that's completely correct in syntax, with most common words in one language/dilect, then the output would be most correct (minimum of possibilities for misinterpretations etc), perhaps even with no syntax error if the program was good, but in the end, you'd have to change it into fuent language, and at this point instead of translating a sentence, you'd have translated the sentence 3-4 times. To pick the correct synonym the program would have to analyse the time the original text was written, the writer and his texts, probably his age and previous works, and hit in a rng to just match up to the human mind.

Snarky

Quote from: Tuomas on Sat 07/07/2012 01:19:49
a language is created by a man, and there's usually no rules to follow, the only way to learn one is to speak one for all your life.

Well, that's obviously false. All languages have tons of rules, and it's possible to learn a language you haven't spoken all your life.

QuoteYou can't really translate a text from one language to another because the whole ideas that the words represent are always different. Basically it's all about analysing the original text and writing down the interpretation.

... and we call that: translating! Of course you can't just take each word in one language and look up what word to replace it with in another language. No one thinks that.

QuoteWhile the whole academic community seems to have abandoned the idea of translating with a computer,

Eh? I'm on the same corridor as a computer linguistics department, and I can assure you based on the research posters they hang up (I remember one in particular that was looking at Swedish TV subtitles) that academic researchers are still working on computer translation.  Besides, computer translation plays an increasingly important role in the real world, with tools like Google Translate (including on YouTube subtitles) and Word Lens for the general public, and special apps e.g. for soldiers in Iraq.

Quoteit would have worked if it weren't for all the exception and the fact that "there are as many languages as there are speakers". Meaning we all have our sociolects etc, children speak differently, everyone has a different register when speaking to his boss or his brother or best friend etc.

So how can people communicate at all, then?

Let's be clear on what the goal is here. A working computer translator will not be able to translate absolutely everything with 100% correctness. People don't always understand each other, and some things are "untranslatable", e.g. puns, rhymes, deliberate ambiguities, or allusions. But let's not exaggerate either. Those are the exceptions. In practice, most of what people write will be comprehended by other speakers of the same source language, and in most cases there is a sentence in the target language that represents a good translation of it.

One of the big challenges is that a lot of perfectly clear sentences rely on semantic disambiguation. In other words, on people understanding the meaning of the sentence. (A not-so perfectly clear example of a sentence that is grammatically ambiguous and can only be resolved semantically - with effort, because it's deliberately confusing - would be the old "Time flies like an arrow; fruit flies like a banana.") This is hard for a computer: you can see Google Translate choking on even simple things like using the gender-appropriate pronoun depending on the name in sentences like "Lisa gillar inte sin chef" ("Lisa doesn't like her boss"; Google gives "his boss").

Getting a computer to be able to do semantic disambiguation at the level of a human would pretty much require creating a human-level AI, but that doesn't mean we can't do better than the current performance. Apple's Siri assistant is able to use context to disambiguate a lot of commands and queries that wouldn't have been possible a few years ago. And just last week, I heard a talk by one of the IBM guys who'd worked on Watson (the Jeopardy-playing computer). A lot of the work on Watson was on how to interpret the meaning of questions, and they made huge advances over the previous state of the art. One of the possible applications they have in mind for the technology is computer translation.

QuoteA translation from Swedish to English would be extremely difficult to accomplish seeing as how they're two completely different languages.

I just translated a sentence from Swedish to English a couple of paragraphs up. It wasn't difficult.

QuoteYes, Swedish has a lot of English words, mainly nouns, but that's about it. Swedish into Danish, Norwegian and maybe even German might work better because they're both germanic languages.

For the record, English is also a Germanic language. And in the big scheme of things (meaning compared to, say, Arabic or Chinese), Swedish has a lot more in common with English than just a few words.

QuoteWhat you would have to do was to change the original input into one that's completely correct in syntax

Google can already deal with common misspellings and grammatical errors, because it uses statistical methods rather than some hard-coded rule about how things should be written. They could certainly apply more of the autocorrection features from Google Search to Google Translate.

QuoteTo pick the correct synonym the program would have to analyse the time the original text was written, the writer and his texts, probably his age and previous works, and hit in a rng to just match up to the human mind.

A lot of the time readers don't know all those details about the writer, either. And again, let's not exaggerate the difficulty involved. Either the language has stayed pretty much the same (as English has for hundreds of years; so that it's quite easy to read e.g. Jane Austen today), and only a very few words have changed their meaning significantly; Or, if the language changes are greater, they are also more noticeable. So if a text is full of "thou" and "doth", those are context clues to use more archaic interpretations.

Of course, the focus of efforts is on modern texts rather than archaic ones. It's OK if a Swedish-English computer translation service has trouble with Bellman or Shakespeare. We're not going to have computer systems where you can just feed it Dante in medieval Italian and it will give you a perfect poetic rendering in any language of your choice. Or any work of literature, for that matter. You need a good writer with a good ear to figure out what works and what doesn't, how best to phrase something. A computer won't do that for you. What it can do is provide a "mostly" readable, understandable and correct version of texts if they aren't too obscure.

Tuomas

#16
I'm thinking, do I want to go there, since I should be studying for my test and forum discussions are not really something I'd be interested in nowadays.

Quote from: Snarky on Sat 07/07/2012 11:29:38
Quote from: Tuomas on Sat 07/07/2012 01:19:49
a language is created by man, and there's usually no rules to follow, the only way to learn one is to speak one for all your life.

No spoken language was created according to any rules, but has developed into something, that follows certain repeating phenomena, which we call rules. But all of these "rules" have just as many exceptions and I would never call them rules but guidelines at best. And yes, a native most often does not speak correct written language, but foreign speaker might. This does not mean the foreign speaker is as good at the language as the native, he's basically got the grips of the guidelines but lacks the skills of creating his own personality through what he is saying.

Quote
QuoteYou can't really translate a text from one language to another because the whole ideas that the words represent are always different. Basically it's all about analysing the original text and writing down the interpretation.

... and we call that: translating! Of course you can't just take each word in one language and look up what word to replace it with in another language. No one thinks that.

Agreed, and this seems to be a big problem with computer translation.

Quote
QuoteWhile the whole academic community seems to have abandoned the idea of translating with a computer,

Eh? I'm on the same corridor as a computer linguistics department, and I can assure you based on the research posters they hang up (I remember one in particular that was looking at Swedish TV subtitles) that academic researchers are still working on computer translation.  Besides, computer translation plays an increasingly important role in the real world, with tools like Google Translate (including on YouTube subtitles) and Word Lens for the general public, and special apps e.g. for soldiers in Iraq.

What I meant by the academic community is not the same what you're referring to as Google or Word Lens, which two are obviously products of companies rather than university research. Anyway, the fact that a computer translation can never be correct renders the whole area of study useless, which is why it's been mostly abandoned. But as said, this polish researcher for example never gave up. And who knows when it'll come back. It just doesn't happen now with what we've got.

Quote
Quoteit would have worked if it weren't for all the exception and the fact that "there are as many languages as there are speakers". Meaning we all have our sociolects etc, children speak differently, everyone has a different register when speaking to his boss or his brother or best friend etc.

So how can people communicate at all, then?

Do you not understand any other languages than your own or even any dialects of your native language? Speak with a child about politics and you'll see what I mean. at the same time, it's easy to spot social classes from each other through word usage and pronounciation, and a lot of people choose their company according to what and how they speak, be it, that they all use the same guidelines and grammar.

QuoteLet's be clear on what the goal is here. A working computer translator will not be able to translate absolutely everything with 100% correctness. People don't always understand each other, and some things are "untranslatable", e.g. puns, rhymes, deliberate ambiguities, or allusions. But let's not exaggerate either. Those are the exceptions. In practice, most of what people write will be comprehended by other speakers of the same source language, and in most cases there is a sentence in the target language that represents a good translation of it.

Of course most of it is understandable, seeing as the original text is most often available. But a translation is worthless if it doesn't give out the exact message the original one would, and by this I mean translating literature, not separate sentences of correct grammar. At this point that does not happen. Translating from English and German into Finnish for example it's not uncommon, that you have to replace one sentence with several that explain the meaning of each conjunction or reference. Before that the sentence is incomprehensible, and a computer program does not, so far, know which parts of a sentence or a phrase are difficult to understand and locate and which indeed need more explanation. There are so many different kinds of languages, and I'm usually talking about literature, but for example law-texts are pretty much created to be understood exactly the way they are, and in their case a computer translation in my opinion would be very applicable (sadly about 70% of transaltors work on that field).

QuoteOne of the big challenges is that a lot of perfectly clear sentences rely on semantic disambiguation. In other words, on people understanding the meaning of the sentence. (A not-so perfectly clear example of a sentence that is grammatically ambiguous and can only be resolved semantically - with effort, because it's deliberately confusing - would be the old "Time flies like an arrow; fruit flies like a banana.") This is hard for a computer: you can see Google Translate choking on even simple things like using the gender-appropriate pronoun depending on the name in sentences like "Lisa gillar inte sin chef" ("Lisa doesn't like her boss"; Google gives "his boss").

Getting a computer to be able to do semantic disambiguation at the level of a human would pretty much require creating a human-level AI, but that doesn't mean we can't do better than the current performance. Apple's Siri assistant is able to use context to disambiguate a lot of commands and queries that wouldn't have been possible a few years ago. And just last week, I heard a talk by one of the IBM guys who'd worked on Watson (the Jeopardy-playing computer). A lot of the work on Watson was on how to interpret the meaning of questions, and they made huge advances over the previous state of the art. One of the possible applications they have in mind for the technology is computer translation.

Exactly the problem I and a lot of linguists see with computers. How Siri and Google translate work is that they create a text corpus with a search option for worth pairs or triplets (probably more) that appear together more often and basically use statistical analysis to choose which words & meanings to apply. Pretty much the same as the human brain actually, a lot of people take in sentences and words from say, tv. And this is interesting. The problem is, when running a "Type A" study like this, you only need one corpus that gives you the possible variables, but doesn't go any further than the list it has created by itself. And to be frank I believe, but I don't know for sure, if Google (for example) uses a comparative corpus (or many) to run a "Type B" study with different variations, aka types of texts. But in this case one would have to choose the category of the input while translating, to choose which comparative corpus to use.

Quote
QuoteYes, Swedish has a lot of English words, mainly nouns, but that's about it. Swedish into Danish, Norwegian and maybe even German might work better because they're both germanic languages.

For the record, English is also a Germanic language. And in the big scheme of things (meaning compared to, say, Arabic or Chinese), Swedish has a lot more in common with English than just a few words.

Yes, what I meant was Swedish and German are closer to each other because of all the influence of other languages to the English language, for example the abusive amount of words and romanic languages, umm, I could see there being lots of problems with for example word compounding for instance. Not to mention my language, where the grammar isn't even realtively close to anything :)

QuoteOf course, the focus of efforts is on modern texts rather than archaic ones. It's OK if a Swedish-English computer translation service has trouble with Bellman or Shakespeare. We're not going to have computer systems where you can just feed it Dante in medieval Italian and it will give you a perfect poetic rendering in any language of your choice. Or any work of literature, for that matter. You need a good writer with a good ear to figure out what works and what doesn't, how best to phrase something. A computer won't do that for you. What it can do is provide a "mostly" readable, understandable and correct version of texts if they aren't too obscure.
The problem with texts that I usually handle is that it's not enough, that you can "mostly" read it. with law, and interpretation is of course always possible after a lousy translation, at linguistics your essay would be thrown into garbage, and imagine an engineer getting instructions for building a bridge translated with Google translator. My sister actually had this problem a month or so ago. She's working on he PhD and needed to quote a German text about composites and metals working together, having to quote it in English, but the whole shit was never translated, so we worked really hard to find the correct nails and weldings and whatever together. This, though isn't of course enough because I'm not a lisenced translator. (thank god)
[/quote]

Perkele2012

to cut it short.

what you can read between the lines in the original text is lost in automatic translation because of the lack of personal input.
:wink:


Tuomas

And this I find to be a real problem

Snarky

Quote from: Tuomas on Sat 07/07/2012 13:17:49
And yes, a native most often does not speak correct written language, but foreign speaker might. This does not mean the foreign speaker is as good at the language as the native, he's basically got the grips of the guidelines but lacks the skills of creating his own personality through what he is saying.

I do a lot of transcribing of spoken interviews for analysis. Spoken language is full of false starts, corrections, incomplete utterances and interjected non-words (umms and ahs), but I would argue that once you strip all that stuff out, the vast majority of the time you're left with a grammatical statement. So yes, people do usually speak "correctly" (though of course they don't speak in written language, that's a contradiction in terms). Whether or not slang and other informal uses common in spoken language are accepted in the written form is a cultural question, not a grammatical one.

As for a foreign speaker "lacking the skills to create his own personality through what he is saying", Joseph Conrad is one of the greatest writers in the English language, with a very distinctive personal style, and he didn't learn the language until adulthood, and spoke it with a strong accent to the end of his life. You could also add Vladimir Nabokov, though he learned English from a young age. Salman Rushdie's first language is Urdu. So no, it's not impossible.

Quote
QuoteEh? I'm on the same corridor as a computer linguistics department, and I can assure you based on the research posters they hang up (I remember one in particular that was looking at Swedish TV subtitles) that academic researchers are still working on computer translation.  Besides, computer translation plays an increasingly important role in the real world, with tools like Google Translate (including on YouTube subtitles) and Word Lens for the general public, and special apps e.g. for soldiers in Iraq.

What I meant by the academic community is not the same what you're referring to as Google or Word Lens, which two are obviously products of companies rather than university research.

Well, like I said, my office is on the same corridor as a computer linguistics department (i.e. at a university), and they do still work on computer translation. Besides, most of these cutting-edge problems in computer linguistics and AI are collaborations between university research groups and companies (or government agencies like DARPA).

QuoteAnyway, the fact that a computer translation can never be correct renders the whole area of study useless, which is why it's been mostly abandoned. But as said, this polish researcher for example never gave up. And who knows when it'll come back. It just doesn't happen now with what we've got.

Right. Just like how the fact that we can't achieve eternal life renders the whole discipline of medicine useless, which is why it has been mostly abandoned.  (roll)

Quote
Quote
Quote"there are as many languages as there are speakers". Meaning we all have our sociolects etc, children speak differently, everyone has a different register when speaking to his boss or his brother or best friend etc.

So how can people communicate at all, then?

Do you not understand any other languages than your own or even any dialects of your native language? Speak with a child about politics and you'll see what I mean. at the same time, it's easy to spot social classes from each other through word usage and pronounciation, and a lot of people choose their company according to what and how they speak, be it, that they all use the same guidelines and grammar.

My point was that the fact that communication is possible at all means that there is a certain shared understanding of terms, a commonality in language. Or maybe you would prefer "The fact that one person trying to understand another can never be correct renders the whole concept of language useless, which is why it's been mostly abandoned."

QuoteOf course most of it is understandable, seeing as the original text is most often available. But a translation is worthless if it doesn't give out the exact message the original one would, and by this I mean translating literature, not separate sentences of correct grammar.

I'm... I don't even...
That's not even wrong. That's insane.

It's impossible for a translation of any particularly challenging text to "give out" exactly the same "message" as the original. So does that mean that all translations of the Bible, all translations of The Iliad and The Odyssey, of Beowulf, of Dostoevsky, of Kafka, of Asterix are worthless? Any movie with dubbing or subtitles, worthless?

You're a translator, right? So your job is being worthless, I guess?

Let's be a bit more specific about "worthless" computer translation:

I sometimes find websites, e.g. Wikipedia articles, that are only available in languages I don't know; but I can run them through Google Translate and (with some effort) read and understand them. That's not worthless to me.

I often get emails in German that would take me maybe ten minutes to interpret. With a single click, Google Translate gives it to me in English. Even if the sentences aren't elegant, or even all grammatically correct, it takes me a fraction of the time to understand. That's not worthless to me.

I have friends on Facebook who are Dutch, Israeli, Greek, Finnish, etc., and who sometimes post in their own language. But since Facebook added a translation feature, I can just click to see it translated, and even if it doesn't always work perfectly (or sometimes even at all), it's often good enough for me to know what they're saying. That's not worthless to me.

I've used Google Translate to help me read comics in French, since it's often quicker and more helpful than a French-English dictionary. That's not worthless to me.

I also have French books that I have scanned and run through OCR. I can do searches for particular key words, and auto-translate the paragraphs around each result to see if it's relevant, narrowing it down before I send it to someone who can do a more accurate translation. That's not worthless to me.

As recently as this morning, I was carrying on a correspondence with someone in a language I don't know. I'm sure my Google-translated messages had lots of mistakes (though I did try to verify them using some of the methods monkey mentioned, as well as others), but we could communicate. That's not worthless to me.

I can use Word Lens to help me understand signs when I'm on vacation in a foreign country. That could stop me from getting lost, help me select what I like from a menu, help me use a ticket machine... it could potentially save my life. That's not worthless to me.

Thinking that a system has to be perfect or else it's worthless blinds people from the huge impact that "good enough" computer translation can have, and already is having. There are things it's still not good enough for, given particular requirements (and alternatives being available), sure. But that's a completely different claim.

SMF spam blocked by CleanTalk