Do you think automatic translation technology will get good enough one day so that we won't need human interpreters or translators?

I remember when Noboru Takeshita was prime minister, he made a statement in the Diet and the TV news translator gave up trying to translate it and said, "The Prime Minister, with one of his untranslatable pronouncements."

Indeed they'res is....

i would like to see automatic translation evolve...But, it may have a problem with the english language..American english is always changing...Seem's like every years they'res a new word coming out...

Probably machine translation will never be identical to human one... but the robot will soon have a level totally acceptable for most cases. They will be like movie subtitle, not 100%, but people are ok with it. In contexts like hotel front desk and such, that will come very soon.

will not be able to translate, let's say poetry, teen-ager slang that is changing every month,

Google is already better at staying update with teen linguo than you and me.

I just want to know is just how some machine is going to be able to translate sarcasm, body language, gestures, and facial expressions,

The same way as you do... Some robots are already learning behaviors from humans , like kids are taught by adults, you face the robot and you say "when I enter, you say hello...", "er, say hello, only the first time you see me in a day..." and it gets better every day. Most jokes and joke patterns are not your invention, you re-use what you've heard. For body language and facial expressions, the robots I've seen at the robot fair were already quite good. They can do the expected conversation just as well as your neighbour as they can analyze you and say you're tired, stressed, merry... and pretend being concerned with your problems. They can analyze context (it's hot/cold/rainy... early/late... week-end soon...). You can input them 2 different cultural contexts, one linked to each language and they pass from one to the other.

along with being able to read between the lines.

You use your experience to do that...Powerful machines can browse an even wider amount of experience and guess suggested things. Probably the machines will detect things most humans fail to perceive... and will miss others that seem more obvious.

Weblio is much better than google translation, considering the ambiguity of the original English sentence.

Ambiguity (and obscurity) are the hallmarks of my native language! :-) We excel at ambiguity, obfuscation, and death-by-idiom! XD

I just want to know is just how some machine is going to be able to translate sarcasm, body language, gestures, and facial expressions, along with being able to read between the lines.

I've seen a computer translate mata ome ni kakarimasu (I'll see you again) as "I will hang before your eyes." Some of the photos on engrish.com are computer mistranslations and are hilarious.

I think we'll have extremely accurate electronic translators within 5 years and within 10 to 20 years we'll have almost perfect translation capability if Quantum computing can be realized.

One day, yes. It maybe 1000 years from now but surely that day will come.

One day.

For words and small phrases it is ok, but it is still a long way from being able to add nuance, colloquialisms and context to translations, which vary from country to country or, in some cases, city to city. A good example would be the word, thong, which is a very common kind of summer footwear in Australia, but it means something completely different in the US. Then, there are English idioms. Some of the more common ones like, 'It's raining cats and dogs' are correctly translated in Google, but some of the more obscure ones like, 'as busy as a beaver/bee' are quite funny when translated, 'Biba no yo ni isogashi desu' (The beaver is very busy). This is where electronic translations will always fail.

First to go will be the Eikaiwa ladies and gents

No. You only need to turn on auto-captions on YouTube to see why. It's hilarious.

Weblio is much better than google translation, considering the ambiguity of the original English sentence.

Still pretty crap. And write that sentence in proper Japanese, then translate it back to English - it will be even worse (it's easier for machines to go from E -> J than from J -> E).

Actually Google maybe the first to develop a reliable translation system with since they had adopted AI, deep learning and are the most popular as a computer translation provider. Basically practice makes perfect in which the more data is put into an AI, the more deep learning or practice is provided making it the fastest learner.

It would probably advance further if a scoring and correction feature by the reader is added.

FadamorOCT. 04, 2016 - 03:09AM JST

That said, things still are not to the point where you can't tell if it was translated by a human or a machine.

Translation by Weblio. http://translate.weblio.jp/

それは言いました、あなたがそれが人間か機械で翻訳されたかどうか見分けることができない所で、物事はまだ適切でありません。

Weblio is much better than google translation, considering the ambiguity of the original English sentence.

Not in the near future, but it will happen some day for sure.

Not if it uses Google Voice...I once asked "Did you put a 1000 yen in my wallet?"...to which it typed "Did you put urine in my wallet?" My wife got a laugh since I hit the send button without checking

That's a hard thing for us foreigners in Japan - the voice recognition systems can't handle the mixing of languages. That's still very far off, probably nearly as far as good machine translation.

If the answer ever becomes YES, hopefully I am no longer of this earth!!

Not if it uses Google Voice...I once asked "Did you put a 1000 yen in my wallet?"...to which it typed "Did you put urine in my wallet?" My wife got a laugh since I hit the send button without checking

Yes, I do, but I always thought it would the Japanese electronic companies who would do this.

I am still waiting.

Seeing as the question leaves the time frame so open-ended, the answer is obviously "Yes".

それは、物事はまだそれが人間または機械によって翻訳された場合、あなたは言うことができない点にはない、と述べました。

(That said, things still are not to the point where you can't tell if it was translated by a human or a machine.)

(The translation by Google Translate)

As of my post right now, there must be 41 people here who earn money as interpreters and translators. Yes, ONE DAY, the article says, some day...

It's extremely hard to tell not to achieve the absolute automatic translation because 'one day' includes eternity:)

We can't underestimate technology. That's all.

@Patricia: I have also been wondering, how in the world did they get the job? It is pure torture to listen to them. Especially when they have to interpret simultaneously.

@CH3CHO: one meaning of the word "translator" is "interpreter".

Yes and No. For an AI to achieve perfect translation they will need to pass the Turing test with 100% efficiency. The present score is 30% efficiency acting as a 13 year old.

Patricia YarrowOCT. 03, 2016 - 07:34PM JST

I have been sure for quite some time that the female NHK translators

I think they are interpreters rather than translators.

I have been sure for quite some time that the female NHK translators of the news (except for one British sounding gal) ARE A.I. translators. It's the only explanation for their infernal and constants gaffes that NO native or highly trained translator would make.

Considering that most Japanese companies don't even consider translation ability as skill they would be better off with a crappy robotic translation. Honestly I don't think most Japanese companies care about the quality of translation.

cleo No, not for a long, long time. My job is safe.

Always thought your job was to come on here and spout about how wonderful dogs and the J-stone cops are

The i-phone gadgets I am seeing in use right now amaze me. Translations can be made from any language in an instant. They may not be perfect, but they get the idea across. Add another 10 to 20 years in developing translation devices and we will be seeing something more perfect ... perhaps even perfect itself. Me and my translation/re-write friends may not like it, but the translation/re-write industry will someday be a thing of the past.

The fact that so much writing is unintelligible is a dual-edged sword. On the one hand, it's hard for anyone to translate it. On the other hand, the bar is set low for the intelligibility of the translation.

I think better education in speaking a language which is understood globally is a more achievable goal. It's English at this moment in time but that could change.

I'd hate to see a smartphone application which could translate perfectly. Pretending to be completely ignorant of a language is a perk which I don't want to lose.

Define "one day".

No, not for a long, long time. My job is safe.

http://translate.weblio.jp/

Weblio is generally good. But it sometimes produce strange translation.

デザートはいかがでしたか。

味がくどすぎました。

How about the dessert?

Taste was too lengthy.

That is probably because a lot is abbreviated in Japanese, and there are Japanese words that do not have exact translation in English.

Machines already do a remarkable job at translating standardized, high quality speech and text, but the problem right now really is that machines cannot differentiate between people or adjust to people's deficiencies in language use.

I'm not sure if you've never used Japanese -> English translators, but they do not do a very good job even with high quality text.

If you disagree, please give us an example here, and we can run it through the translator and test.

Machines already do a remarkable job at translating standardized, high quality speech and text, but the problem right now really is that machines cannot differentiate between people or adjust to people's deficiencies in language use.

It's kind of ironic to see examples (such as above), where the speech is filled with disfluencies, and the argument is that machines can never understand it. That is, if a machine is built to be as accurate as possible, but the input is garbage, then the output is garbage as well.

For humans this is not a problem, but we'll have to wait a few more years for AI and machine learning folks to construct something that does well with low quality speech, or text for that matter. Until now, it has hardly been a worthy goal to build something that can parse garbage into comprehensible garbage in another language.

Heck, NHK has only one man that does a wonderful spontaneous job at it. Machines will not do it.

Automatic translation technology would be very wonderful and the most convenient because we/they don't have to hire/call interpreters/translators every time that costs a lot.

Eventually, yes, utilizing AI. AI is functioning in various field already and evolving in much faster pace than people can speculate.

I've been translating many languages for over 10 years and my opinion is the same Ishiwara pointed above. Sometimes I get tasks of making machine translated phrases more natural. More human the text, bigger the mess.

@dcog

You said it. Give it a hundred years and people will be amazed that there ever existed doubts....

Yes, easily. AI is still in its infancy and the potential gains are far, far greater than anything us humans can ever hope to comprehend. 100% translation of languages will be a shockingly basic function for future AI

kohakuebisu: "It should be noted that a lot of material that is very hard to translate is also of little commercial value. There are not that many translators out there earning a living thanks to "teenage slang" or ancient poetry. Most (yes, not all) translators earn by translating business documents and releases, financial reports, government stuff, patents, manuals, and academic papers."

Very true, and in those respects translation software helps, though is not a substitute. But let's face it, the travel industry is where this kind of thing is needed for many, and where it would make the most money if it were to succeed. But that involves not documents, manuals, patents, or academic papers so much as interaction between people, guides, shop staff, etc. THAT is where it's furthest away.

The translation industry has various jokes along the line of "automatic translation has been 10 years away for the past 30 years".

Computer assisted translation is huge though. It makes the job easier, but creates big downward pressure on rates.

It should be noted that a lot of material that is very hard to translate is also of little commercial value. There are not that many translators out there earning a living thanks to "teenage slang" or ancient poetry. Most (yes, not all) translators earn by translating business documents and releases, financial reports, government stuff, patents, manuals, and academic papers.

Maybe in 50 to 100 years. Maybe. Writing effectively is not that easy, and there is a great deal of art to it. Simple translations, though, where clunky writing won't make too much difference? Just around the corner.

No, never. Not unless it is a fully-functioning AI, because machines can't understand the nuance and idioms of another language in the context of a conversation, given that idioms are unique to each language. Have you ever tried using the Facebook translation software to see how everything is written in another language? or Google? Give it a try, and get a good chuckle -- and that's with many possible permutations and translations having been input into the machines beforehand. It'll get better, and in terms of vocabulary and specifics there will be room in certain fields for translation devices, but they just won't ever be able to replace human interpreters for regular conversation or to try and understand a person's feelings or opinion. I can just imagine a tourist talking to a local using some sort of translation device...

Local: Ano... sa... ii tenki, ne! (translation device: "That one... sir... nice weather, isn't it!")

Tourist: No kidding! I was sweating my balls off yesterday, so today is awesome! ("It is no joke. Yesterday because of sweat I could not catch the ball, so today is one of awe.")

Local: yarou? ("I will do it?")

Tourist: Ummm... ("Tastes good")

Local: Nnn? ("?")

Tourist: Is there something I can help you with? ("Shall we help"?)

Local: Ano... ne... chotto, ne... taihen moushiwakenai ga, 1000 en wo kashitekuremasuka? ("That one... ?... a little... ?... it is difficult sorry, could I be borrowed 1000 yen?")

Tourist: Yeah, right! You wanna take my money? ("Okay, certainly. The money snare is taken?")

Local: Mou ikkai itte kudasaimasuka? Imi wo sukoshi wakarahen. ("Oh! Could it be said one time? The meaning was not understood in Osaka dialect")

Tourist: Whatever. I'm outta here. ("It's fine. I'm outer here").

Nope... not for a while yet. Not with idioms, and not with grammar. Simple patterns will progress, but it will still be a computer selecting among possible translations, with no idea of what is intended or of context.

Short answer: No.

Long answer: Yes, but only by the point that machines are able to dynamically interact with humans through natural language to such a high degree that our interaction with both machines and other people has been irrevocably changed.

Contextual awareness and other pragmatics sense: these points need to be got right, and I am not holding my breath.

No.

Technology will be able to translate technical manuals and instructions for how to use an airport and so on, but will not be able to translate, let's say poetry, teen-ager slang that is changing every month, irony, etc.

Most definitely. Maybe not any time soon, it could take many years. But unless we collapse our society, it will definitely happen. Clarke's third law was "Any sufficiently advanced technology is indistinguishable from magic". If you showed a working iphone with a network to someone from 100 years ago, they would think it's magic. Show them how you can do a video chat with someone on the other side of the planet in real time, and it's magic. You could even try to explain it to them, but it would still look like magic, because it wouldn't be in most people's capacity to understand what they were looking at. Same as machine translation - it may seem like it's an impossibility to some people now, but with enough progress, we will get there. Look at that Watson computer on Jeopardy - it won the show, dealing with real-life language to find the answers to the questions. If that processing power was put to understanding languages and bridging them together, it wouldn't a far reach to suppose they could manage it in not so many years.

As I have the chance to translate for a living, I would say no. It will surely come close in many fields . However, as long as the human experience is distinct from the machine then there will never be a perfect synthesis.

Sure, if it understands idioms, intonation and irony. But another way is to fit our communication to the tech; teach us to speak and write machine-understandable and -translatable language. If previous technology is anything to go by I expect movement in this direction; we have to move to fit the machines more than the machines move to fit to us. See how people have to speak to Siri, for example. If you think it is natural then you are well on the way to being fitted, almost without knowing.