Thursday, 23 September 2010

Fight the machine? (1)

It was during the 1970s in a school staff room in England. A teacher was preparing audio material for a language lesson. The original was on an old-fashioned tape spool, so he connected a tape recorder to his cassette recorder to transfer the material from one to the other. He turned the volume right down because he didn't want to disturb anyone.
A colleague saw what he was doing and quipped: "Two machines talking to each other, and we can't hear what they are saying. Now that is scary!"

This evokes a whole range of scenarios that are familar from science fiction. But machines today can do much more than they could then. Now, many people have smartphones with more computing power than a whole roomful of equipment back in the 1970s.

This march of technology has also reached the translation business. There is much research and industrial development in a new discipline which is known as "MT" or "Machine Translation". There are romantic dreams about achieving technical inventions which would overcome the language barrier for all forms of communication, rather like the "Babelfish" in Douglas Adams' novel "Hitchhiker's Guide to the Galaxy" (although Adams himself suggested in the novel that this does not lead to global peace, but to even more bloodshed).

Out of curiosity, I fed my first article (Blogging the miracle) into Google Translate and asked it to translate it into German. The result is, of course, full of grammatical mistakes and questionable written style, and definitely not fit for publication. But much of it is more or less comprehensible.

Of course it is easy to make fun of Google Translate by quoting some of its more blatant mistranslations. One way to do this is by the "translation party". Here, you enter an English sentence and it is then translated back and forth between English and Japanese until it reaches "equilibrium", i.e. the English version remains the same on every "round trip". The sentence "Once upon a time there were three bears, Daddy bear, Mummy bear and baby bear", reaches equilibrium as: "Bears 3, Dadikuma, he was a mummy bear and baby bear". Great fun, and only one of the many ways to poke fun at machine translation. But does this really do justice to the subject?

There are a number of serious ventures using machine translation for real translation work. One advocate is Kirti Vashee, who blogs at "eMpTy Pages". He believes that the volume of machine translation will increase to cope with the enormous volume of material needing translation, especially from multinational technology manufacturers, and that many translators will need to move into a new field: post-editing of machine translated output. In his blog he covers many aspects of this topic, which I can't do justice to in a short paragraph here. His blog provides an interesting and thought-provoking (and sometimes controversial) perspective on the subject.

Another advocate of MT is Jeff Allen. He mainly works on French to English machine translation and post-editing systems. He has also done much work on the Haitian Creole language, and this is currently being developed to support disaster relief work after the earthquake in Haiti. I haven't yet seen any reports on how efficiently this works out "on the ground" (does anyone have any recent news?). A good place to start looking at Jeff's contributions is his profile at, which offers a number of links for further reading.

There are other serious users of MT, including large organisations such as Microsoft and the European Union. But MT also meets with much opposition from professional translators. This may be partly due to a fear of losing the market for human translation. But in many cases there are also doubts about how much MT can actually achieve, and whether it can really handle the subtleties of language.

So how do I, as a professional translator, regard MT? That is a long story, so I think I will have to make this entry into a series, and continue in part 2. Some day soon (I hope).

1 comment:

  1. Hi Victor,
    I have my own thoughts on this and I wanted to thank you for providing a little guide to the advocates of MT on the web.