Artificial Intelligence and Translation

by Translation Guy on July 9, 2010

“Our young children and grandchildren will think it is completely natural to talk to machines that look at them and understand them,” says Eric Horvitz, a computer scientist at Microsoft’s research laboratory in a recent interview with the New York Times on speech recognition and medicine.

Horvitz has been working on a medical avatar project that uses artificial intelligence systems to listen and talk to patients in order to find what ails them and to recommend what to do. And naturally, thanks to natural language processing, they can translate between languages too.

“For decades, computer scientists have been pursuing artificial intelligence―the use of computers to simulate human thinking. But in recent years, rapid progress has been made in machines that can listen, speak, see, reason and learn, in their way. The prospect, according to scientists and economists, is not only that artificial intelligence will transform the way humans and machines communicate and collaborate, but will also eliminate millions of jobs, create many others and change the nature of work and daily routines.”

Meanwhile, translation software being tested by the Defense Advanced Research Projects Agency is fast enough to keep up with some simple conversations.

So, are expensive telephone interpreters on their way out? Sure, but I predict we’ll be among the last to leave the room to the computers. Translation is the final wrinkle for AI systems trying to navigate human relationships. Easy comes first. “Basic work that can be automated is in the bull’s-eye of both technology and globalization, and the rise of artificial intelligence just magnifies that reality,” says Erik Brynjolfsson, an economist at the Sloan School of Management at Erik Brynjolfsson.

The customer interaction is a logical place for AI to be applied in the call center business, but the first big advances are coming in the way artificial intelligence is used to manage traffic on networks. Real-time systems can adapt in real time to customer concerns before an agent even picks up the phone. And as the technology improves and becomes more widespread, we have become better as using it.

An interesting case study here. Panasonic designed a call center system to identify keywords so that an automated assistant could reply intelligently instead of having customers punch through such a dense call tree that 40% hung up in frustration.  “Today, Panasonic resolves one million more customer problems a year with 1.6 million fewer total calls than five years ago. The cost of resolving a customer issue has declined by 50 percent.”


  1. Brian P says:

    It might be nice, in the future, to leave the machine-like tasks to computers so that we can focus more on the creative and artistic. But if people grow up without ever having to do basic tasks, like building things, cooking, cleaning, translating languages, etc., I’m afraid of what will happen to their minds. They might miss out on basic understanding of how the world works.

    The experience that comes from doing machine-like tasks eventually becomes inspiration for creative thought that drives society forward.

  2. markowe says:

    I find it very hard to believe computers will be able to simulate human thinking and reasoning any time soon, perhaps ever. This kind of thinking assumes the human brain is just a kind of very complex computer. But what if it’s nothing of the sort? For a computer to “understand” and to interpret, it needs to be sentient, i.e. a being conscious of it’s own existence and perhaps to have some higher purpose driving it. Even as humans we often don’t really have an answer to the question of WHY we do things, think things, feel things. How do we instill something similar in a computer? We would have to understand ourselves first, and I don’t think we are even close!

  3. Glen Tomboc says:

    No, machine translation can’t be accurate as of today because machine translation of human languages or natural languages is still a field of research in Artificial Intelligence and still a lots of research is going on in many first world countries. But Google translator is still better than other translators but for a particular set of languages only like English, French, Spanish, German etc. Its Hindi translation is not up to the mark and the translated piece requires a lot of human effort to make it meaningful though small sentences can be translated well into meaningful sentences.

    • Ken says:

      Very interesting comments all. Thanks, you thoughtful posters!

      It was an MT guy, Glenn Akers, the CEO of Language Engineering Corporation, who said, “Language is the sum of all human knowledge.” Can’t argue with that, right? So when you get there, you’re done. So when the lawyer doesn’t need the translator but can use a machine, its likely that a machine will be a better lawyer than the lawyer.

  4. Aaron Hunt says:

    The problem lies in placement of adj. as well as many other factors. Like if I told you I want some blue jeans. In french it would attempt to translate it exactly as you typed it in. However in the french as well as spanish languages among others it would be correct to say jean blue instead. It is very obvious when I am in an online chatroom and translate someone else’s dialogue or it is obvious to the other person when I use it.

  5. Until we can program something capable of human thought patterns of learning, discernment, common sense, and extrapolation — effectively making artificial intelligence on par with the human brain, any true, worthwhile translation program is nothing more than a dream. Some of the romantic languages are similar enough that translation programs you are familiar with can give you a very rough idea of what the original text was saying, but every Japanese to English, or English to Japanese translator that I have seen horribly butchers and messes up even the most simple sentences — which is only to be expected.This is because languages — especially when comparing such languages as Japanese and English — are entirely dependent upon context and interpretation. In a majority of instances, for example, there are not even simple Japanese-English equivalents of words. There is usually a large number of Japanese words — each with slightly different meanings or nuances, that can be translated as any one English word. The vice-versa is also true. Furthermore, translation programs usually cannot even make out the subject and predicates of sentences when translating between Japanese and English! This is because Japanese is a highly context dependent language. Once a subject is stated — or even if it isn’t, it is often omitted and it is assumed that the listener knows what the speaker is talking about.
    To give an example, I will translate a short dialogue with varying degrees of directness and English interpretation. The below is taken from Erai Tokoro ni Totsuide Shimatta, with my own translations:
    Direct Trans.: Kimi-chan, something/or-something being-mad?
    Natural Trans.: Kimi-chan, are [you] upset or something?
    Direct Trans.: See-and-understand!
    Natural Trans.: Can’t [you] tell?
    Direct Trans.: Understood!
    Natural Trans.: [I’ve] got it!
    Direct Trans.: Again ridiculous work forced on I-think/I-guess/right?/deshou.
    Natural Trans.: [They] forced demanding work [on you] again, didn’t [they]?
    Direct Trans.: not-at-all/entirely not-understanding!
    Natural Trans.: [You] don’t understand at all!
    Direct Trans.: cause/source/origin is mother [emphasis].
    Natural Trans.: The cause is [your] mother.
    Direct Trans.: everyday everymorning everynight phone come-calling
    Natural Trans.: [She’s] calling every day, every morning, every night!
    Direct Trans.: can-have-stop to one’s-parent’s-home inviting?
    Natural Trans.: Could [you] have [her] stop inviting [us] over to [your] house?
    Direct Trans.: Mother did-for-be-pleased-with thing of Kimi-chan!
    Natural Trans.: Mother took a liking to you!
    Iさん「よかった!僕 安心したおぉ。
    Direct Trans.: Was-good!/Thank-goodness! I relieved [emphasis]. (note sleepy voice and poorly pronounced “yo”)
    Natural Trans.: Thank goodness! I’m relieved.
    Direct Trans.: don’t ”arbitrarily” be relieved! (note colloquial pronunciation)
    Natural Trans.: Don’t be!
    As you see, sentences often have to be entirely reworded, broken up, re-arranged, have words replaced, added, and ommitted, and in short, a helluva lot of interpretation to come up with natural English approximations. Yes; in translation, there is no “equivalent”, only “approximation.”
    Languages are simply too vast and, for lack of a better word, “適当,” to be translated by even the most clever programs. Until we can replicate the human mind, with all of it’s reasoning capacity, don’t hold your breath.

  6. John Winn says:

    I still believe that It’s not completely Impossible within, let’s say, 10 years… Science sometimes can create miracles : )

  7. One of the many problems that corporations face when expanding their business globally is not only the proper translation but the appropriate localization of their products’ names and slogans. Over the years, there have been many, many examples of directly translated product names, slogans, and catchphrases that have offended and/or confused consumers, and sometimes completely embarrassed the company responsible.

  8. It doesn’t matter with marketing but recently I was riding on a JR bus, the most important transportation company in Japan (I’m sure you’ve heard of them Ken), and found this plate in the small toilet:
    “PRESS TO FLASH” … is it a misspell, or they literally take a picture of you while…?!
    I know that Japaneses have a passion for taking pictures… but…

    • Ken says:

      I hope you didn’t press that button, Tristen. I mean JR trains run on time…. what if that translation turned out to be correct?

  9. David Dunn says:

    It’s easy to forget that not all household robot companions serve as cleaners or personal assistants, but the new Qbo robot is here to remind us they can provide some good entertainment as well.
    What’s cool about this robot is that it runs on the Linux operating system, meaning it has an open source code and can be modified in various ways. Some of its features include both omnidirectional and unidirectional microphones installed in its head, as well as high-def webcams that serve as eyes. The eyelids don’t do anything special, but they add a bit of that human touch. Very cool!

  10. Brian Mathis says:

    first of all artificial intelligence is way long concept…it is not only a program but it a also is like a brain which can think like a brain and perform like it…it is based on nano technology..whis can perform many tasks by itself at atime without any man power……like supercomputers…..AI and robotics is giving life to it

  11. CoargoPyday says:

    Nordegren told People ammunition she and Woods tried instead of months to accommodate the relationship. In the finale, a relationship “without safe keeping and ardour” wasn’t furnishings an gaze to anyone, she said.

    On Thanksgiving tenebriousness longest their Florida people’s home, Woods drove his SUV completely a cannonade hydrant and into a tree, mounting off revolting revelations that sports’ biggest outstrip had been cheating on his helpmate through multiple affairs. The span officially divorced Monday.

    my blog

LiveZilla Live Chat Software