What's happening at the
Thursday 17 Aug 2017 | 16:32 | SYDNEY
Thursday 17 Aug 2017 | 16:32 | SYDNEY

Reader riposte: The limits of machine translators

By


This post is part of the Asian languages in Australia debate thread. To read other posts in this debate, click here.

COMMENTS

7 November 2011 11:32


This post is part of the Asian languages in Australia debate thread. To read other posts in this debate, click here.

Aidan Dullard:

Cameron's point about the increasing sophistication of technology like Google Translate is often seen as the death-knell for professional translators and interpreters; as machine translation gets more accurate and more widely available, the need for human translators will supposedly diminish. As a student of Asian languages, I'm not so much worried at the job-destroying part of machine translation's potential, but its usefulness. While the technology certainly isn't there yet — my personal experience with Google Translate using Japanese and Chinese has turned up some very weird results — the potential is very exciting. Google's methods (using statistics and probability to gauge meaning and a gigantic database of bilingual documents to 'train' their system) works much better than many previous efforts and will presumably become even more accurate and expansive in the future.

But I'd argue that because it's dependent on a finite bunch of bilingual words and phrases, machine translation as we understand it today will never have the flexibility or skill at interpreting context of a human. Machine translation is excellent at getting the (somewhat garbled) gist of a foreign language document, and even better with smaller words and phrases, but for face-to-face conversations, not to mention official documents or business meetings, a human thinking on their feet and with knowledge of idioms and the cultural context couldn't be matched by a machine at this stage — and possibly never will be, without genuine artificial intelligence.

Many contributors to this debate have mentioned the difficulty of learning Asian languages as a reason for their declining demand; I'd suggest that, rather than threatening future employment prospects, Google Translate and its ilk actually make eventual mastery of these languages much easier. Looking up a word takes seconds on the internet rather than minutes flipping pages, and a smartphone with a database of thousands of words is, after all, a lot less heavy than a paper dictionary.

Finally, I think that even apart from the hard-nosed economic calculations surrounding studying an Asian language, there's an enormous personal benefit — and not just being able to read the menu. Understanding languages as part of broader Asia literacy is incredibly important if Australia is to engage successfully and deeply with countries in our region, and for me this would be a compelling reason regardless of the future potential of machine translation technology to replace humans. In the same way that hoping 'everyone else will learn English' is distressingly parochial, I don't think that machine translation offers a panacea, and I'd suggest that at the very least language skills provide an excellent base for Australians to relate to our neighbours in the so-called Asian century.

You may also be interested in...