Will There Still Be Any Translators/Interpreters Left in 2107?

Now, let’s face. In about one century (early 22nd Century), our profession will probably be obsolete (shudder 🙂. It will all start with scientific translations (the more “scientific” a text, the easier it is for a machine to translate). But one thing is for sure: literary translation, especially poetry, will be a hard nut to crack for machines (the more “cultural” a text, the more difficult it is for a machine to translate).

So, fast forward to 2107 – the only translators left are literary translators… unless, of course everybody speaks the same language.

Now, be courageous, friend, and read this article in: DefenseNews.com

A few highlights:

DARPA (The U.S. Defense Advanced Research Projects Agency) says all (the five specific translation devices it is testing) perform in the 70 percent to 80 percent accuracy range.

In the short term — the next three to five years, she said — DARPA wants 80 percent to 90 percent accuracy (much better than Google translation software’s 60%) for specific task-related phrases (This of course is far from a 10,000 word-document, even when you use TM!).

“We are optimistic that in the near future, we will be able to deliver a device that will be able to translate successfully 80 to 90 percent of the time when speakers articulate carefully and stick to specific subject areas during the conversation,” said Mari Maeda, program manager for DARPA’s Translation Systems for Tactical Use program.

The long-term goal, said DARPA spokeswoman Jan Walker, is two-way translations across all subjects with 100 percent accuracy, with background noise, dialects and accents taken into account. (How LONG-term will that be? Well, I’ll believe it when I see it 🙂

At least this will go a long way towards reducing translator/interpreter casualties. (See “Decimation” post) Or maybe by the time they get to 100%, there will be Universal Peace, no more wars, and all that …

As a friend of mine likes to say: “Even if we all spoke the same language, there would still be a necessity for translators and interpreters, because some people would find a way to speak that common language differently”.

To which I will add: “And they will make their own version of that common language so complicated that no machine will be able to translate it”

So, don’t worry, dear friend and colleague, we’re still around, for some time yet!

A.M.Sall

P.S. Many of you have tried to sign up for our Free 7-Part PR ecourse, but the link had some problems. We are very sorry. But we’ve fixed everything: we’ve turned the ecourse into a free downloadable PDF and all you have to do is visit this link and instantly download it!

*******

Subscribe to Translator Power and successfully market your translation/interpreting services either by Email or if you prefer the feed, just look on your left and make your choice!

Advertisements

About translatorpower

The Cross-cultural Connector, The Global Mindset Advocate
This entry was posted in Amsall, interpreters, literary, PR, Translator Power, translators. Bookmark the permalink.

3 Responses to Will There Still Be Any Translators/Interpreters Left in 2107?

  1. Manny says:

    Scary but quite possible indeed.

    Like

  2. Yeah, Manny, but don’t you think we have enough time to get ready 🙂 100% accuracy is not for tomorrow (Of course you could say that even with human translators you hardly ever get (100% accuracy!)

    Like

  3. postediting says:

    As for DARPA, they have been funding translation system projects for a long time, and have been making the same statements about the goals. I’ve worked on such projects. See my article on this topic at:
    ALLEN, Jeff. 2000. What about statistical-based Machine Translation? In International Journal for Language and Documentation (IJLD), Issue 7, October/November 2000. pp. 41-42.
    http://www.geocities.com/mtpostediting/Allen-ijld7-updated.doc

    As for the statement: “We are optimistic that in the near future, we will be able to deliver a device that will be able to translate successfully 80 to 90 percent of the time when speakers articulate carefully and stick to specific subject areas during the conversation”, this was already being done 10 years ago. The reference here is to speech-enabled translation systems. The more prepared and find-tuned that the system is with regard to topic-specific vocabulary, and if the users of the systems already use that specific vocabulary, then the recognition level is obviously higher. However, the reality is that even the same user will produce different variants of the same terms and words in their speech over time. And this obvious happens all the time between different users. So, the challenge for such systems up till now has been how to deal with the high range of variants.

    With regarding to articulating carefully, this can be seen with any speech recognition system. The amount of speech data collection which is needed (and I’ve done a lot of that for several languages) to build an prototype system for a language is significant. The ongoing speech collection to fine-tune the speech systems to understand not simply 200-400 people, but rather anyone who speaks that language as a native speaker, is an endless task. The simple reason is that no single person pronounces the same sounds 100% all of the time. It’s the simple basics what what is taught in introductory phonetics and phonology courses. And then if you add to the formula the need for the system to understand non-native speakers of language, that make it even more complex.

    The additional statement of the long-term goal to handle “100 percent accuracy, with background noise, dialects and accents” will not happen. 100 accuracy in language is impossible in non-controlled environments because language is not binary and mathematical, but it rather subjective and influenced at every moment by social, psychological, and other factors. For background noise, I have worked in many projects on this, and it is the law of the lowest common denominator that prevails. You can have the most sophisticated software and background noise suppression programs, but if the hardware equipment is low quality, then the software cannot do its job. The ability for such speech translation systems to master all dialects of a given language is not a simple 3-month task. Anyone who works both in speech and language, and who understands the complexity of it, will first ask you which dialect you want to focus on.

    I’ve provided below a link to my articles on speech translation issues:
    http://www.geocities.com/jeffallenpubs/speechtech.htm

    Like

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s