What Can a Technical Translator Expect from Machine Translation - invited talk at the Swedish technical translators' conference

I was invited to give a talk at the yearly conference of the Swedish Professional Translators' association (SFÖ); see http://www.sfoe.se/sidor/konferens-goteborg-2013--valkomna.aspx for the conference web page. The talk was in Swedish, and so are the included slides. It was a very nice event, with 370 participants, highly qualified people (100 to 150 came to my talk).

My message was that there is no fear that computers will take over translators' jobs any time soon, because of the reasons that Bar-Hillel and Martin Kay once gave and which are still valid. And that it is actually very strange that anyone should even suggest that machines could do translations unsupervised, while in most other jobs (such as construction work, health care, journalism) machines are just helping, not doing the whole job. But even in translation, machines can certainly be helpful in routine and boring tasks, such as making sure that grammar rules are obeyed. In MOLTO, most of the translation tasks are ones that no-one ever bothered to do before, such as rendering thousands of painting descriptions to 15 languages. This is possible because the task is limited, but orders of magnitude below what would be needed in full-scale translation.

I concluded by suggesting that translators could do things higher up in the value chain than current translation jobs: help build translation systems (via grammar tools and example-based grammar writing) and not just individual documents or (as in translation memories) canned phrases.