The big breakthrough in machine translation occurred when Google lead the way in abandoning a rules-based system and switching to a statistical-based system in 2007. By trawling through a trillion web pages they identified masses of pages which were translations of other pages. They trained their system on thousands of pages of parallel text thus enabling the computer to come up with a range of possible translations of any new text. These alternative texts are then run through a model of the target language, thus enabling the computer to select the most likely interpretation of any cruxes. The Economist‘s recent (7 January) supplement on language processing describes this research in Section 3, Machine Translation: Beyond Babel.

“The time is coming when machines will be better at transforming a text from one language into another than a human translator.” Jaap van der Meer tells us at the TAUS blog. Most of the voices in this post caution that we not allow our expectations to run away with us. Giving with the left hand while taking away with the right, Khalil Sima’an says “Some of the more exciting kinds of translation, for example high-end literature and poetry, might remain in the hands of a few gifted human translators for some time to come.” Might remain? For some time to come? Not a lot for the anxious translator to hang onto.

But, isn’t this great news? We don’t need to imagine downloading a digital text of Crime and Punishment into Google Translate, which overwhelming dump may never get to the level of digestibility, to envisage a vast freedom of access to foreign literature. If, as I have been intermittently playing around with, I sit down to make a translation of Jean Giono’s Un Roi sans divertissement, how much of a leg up would it not be to have the whole thing rough drafted at one click of the mouse? The translator would then be freed up to polish infelicities, and important in this case, to establish a stylistic distinction between multiple different narrators — but maybe Google Translate or son of Google Translate, because most of the discussants don’t see this happy day arriving till 20 long years have elapsed, will actually be able to do this stylistic sorting more easily than me. For many readers and purposes I dare say an unedited first cut of a translation might be enough. Some publishers will regard polishing as a matter of pride. But either way, are we not inevitably going to be getting a huge increase in the availability of translated literature from all languages?