Thursday, 11 March 2010

Where Machine Translations fits with Human - the Translation Continuum

Many thanks to clients and suppliers who took the trouble to send to us a piece in the New York Times recently about GoogleTM's free machine translation tool. The quality of machine translation produced has improved significantly over the past year as a result of some serious investment. Google is now our first recommendation when clients want a quick and easy gist translation.

So great news for society at large - as long as society uses the technology with great care! The improvement is such that in future people will be able to communicate across cultures more easily. But how do we view improved machine translation in the human translation industry? Good news for us too for several reasons:

  • Our clients have a real benefit in having a tool to give them a rough understanding of a document before deciding whether human translation is required.
  • In purely business terms, the translation market is expanding - partly due to this type of technology. Machine translation means that people are communicating more in foreign languages. This drives increased demand both for machine translation but also for complex and business-critical human translation.

I have to say that I am very much a linguist rather than a scientist, and human translation remains an art. From my viewpoint, machine translation is not going to replace human translation but it moves further along the translation quality continuum. Think machine translation at one end of a quality continuum with high quality human translation at the other end of a very long line. Machine translation is moving further along that line.

And a final thought for professional translators - I'm delighted to see Google amongst machine translation options integrated  as one source of reference in SDL Trados Studio 2009 SP2, our translation memory tool of choice. The critical question here is how can machine translation be best deployed as a back-up resource for those of us using translation memory, termbases and other technologies for human professionals. I think that SDL have adopted precisely the right approach.