The Power of Machine Translation

Recently the BBC held what they called Super-Power Nation Day during which they asked the question; If everybody in the world could communicate freely with each other, no matter which language they spoke, what would happen?

My somewhat simple answer to their question is ‘The Tower of Babel‘, but that probably isn’t what they were looking for. A subsequent report on the day makes fascinating reading.

By using a specially created website, users from around the world could post and reply to each other’s messages, even if they did not share the same language.

The experiment was part of the BBC’s SuperPower season, a series of programmes, online reports and events designed to examine the extraordinary power of the internet.

Representatives from more than 20 of the BBC World Service language services translated for people who attended the six-hour event at Shoreditch town hall, or called in by telephone.

Meanwhile, comments online were translated using software created by Google, allowing users to write in their own language before seeing it translated into six others instantaneously.

English, Arabic, Chinese, Portuguese, Persian, Indonesian and Spanish were all supported.

It seems as though the day was a great success and the BBC and Google should be congratulated for what they achieved. But…

For twenty of the languages they had to rely on human translators, only six could be translated by computers. And that leaves something like 6,000 languages which were not represented at all. Now it is true, that a large proportion of the world’s population speaks on of the six languages that were included in the computer experiment. But what of those who aren’t? I become increasingly concerned about the digital divide that we are creating between the world’s haves and have-nots. As I’ve argued more than once, machine translation will probably never extend to the significant number of minority languages in the world. And these are the languages spoken by the poor, the dispossessed, the ones without water, internet access or adequate political representation. What are we going to do to ensure that they don’t simply get left standing at the side of the information superhighway?

6 thoughts on “The Power of Machine Translation

  1. I helped creat speech-to-speech MT systems for several less-priviledged languages (Haitian Creole, Croatian, Korean) 12 years at the Language Technologies Institute of Carnegie Mellon Univ. We recently released data (carefully collected from others and created onsite during a period of 2 years) of the Haitian Creole project which as used by Google and Microsoft to create their online translators in just a few days. Once language data is available, then MT is possible.

  2. Here is the way to make it MT possible for minority languages, with a technique that as already used.

    The Bible as a Resource for Translation Software: A proposal for Machine Translation (MT) development using an untapped language resource database. By Jeff Allen. In Multilingual Computing and Technology magazine. Number 51, Vol. 13, Issue 7. October/November 2002. Pp. 40-45.
    http://www.multilingual.com/articleDetail.php?id=614

  3. Hey there!

    Just thought I should chime in to clear something up. We have written up a more detailed report here (http://www.bbc.co.uk/blogs/bbcinternet/2010/03/superpower_nation_an_experimen.html). But during the 6 hours our audience actually had the opportunity to post in 51 different languages which were automatically translated. The 6 languages represented (English, Arabic, Chinese, Portuguese, Persian, Indonesian and Spanish) correspond to our newsroom operations and audience, hence why we only offered those 6 views.

    The BBC World Service take great pride in the fact we broadcast in over 30 different languages (http://www.bbc.co.uk/worldservice/languages/index.shtml), especially to areas with undeveloped or restrictive media markets. The internet is a fabulous enabler for languages which don’t have natural representations in traditional media. The greater the body of documents in a particular language, along with a corresponding translation – the better these huge, statistical automated translation systems can be.

    So I don’t think its all doom and gloom.

    Best,
    tom

    1. Hi Tom, thanks for stopping by and for the clarification. I’m a great fan of the World Service, it was our lifeline when we lived in rural Ivory Coast. I certainly don’t think it’s all doom and gloom for minority languages, we are working in about fourteen hundred of them at the moment and most of what we do will eventually end up on the Internet, which will help build up the sort of corpus you are talking about. However, from our point of view, no one else is likely to develop the corpus, so computer based translation is not likely to be a great help to us – but others can build on our work, which is no bad thing!

Comments are closed.