Serdar Yegulaip wrote an interesting post entitled “Talk to Me, Openly” that is up on InformationWeek’s site, dealing with open source machine translation. While there are a number of very active open source translation memory projects out there, with OmegaT being one of the most popular among freelancers, open source machine translation hasn’t sparked the same level of interest.
Personally, I find machine translation more interesting than threatening. I think that the day when human translators will be put out of business by MT is far, far off, likely beyond the lifetimes of any of us who are working in the industry now. However, I think that MT’s use in areas where speed and cost are more important objectives than quality, or where writers are willing to use a very restricted vocabulary, is really interesting. For example, at a translation technology seminar that I attended this summer, the presenter gave an anecdotal example of MT being used to translate online knowledgebase pages for a large technology company. Apparently, when the company then surveyed people who read the machine-translated help pages (and who knew that the page they were reading had been translated by a computer rather than a human), their rates of satisfaction with those pages were only slightly lower than the satisfaction rates of people who read similar documents translated by humans. To me, that’s more intriguing than anything else, and I’m interested to see if the open source community jumps on the bandwagon.
Hi Corinne:
A very large knowledge base that has been translated with MT (and I think even without human postediting, although I’m not sure), is Microsoft’s.
“I think that the day when human translators will be put out of business by MT is far, far off, likely beyond the lifetimes of any of us who are working in the industry now.”
MT alone, maybe, even probably. But what I think will happen fairly soon is that in certain domains any translator unwilling to work with MT (either as a tool used by the translators themselves, much as TM is now, or as a tool used by the translation customers, who then contract translators to post-edit MT output) will find work draying up.
Riccardo
Riccardo, thanks for your comment! Yes, the knowledgebase that I heard of with respect to MT was Microsoft’s; and as you said, I think that they did not postedit it, but I’m not sure. That’s an interesting observation about translators in certain specializations needing to work with MT, I’m sure you’re correct! I remember reading somewhere (of course now I can’t remember where…) that although translators shudder at the thought of working with MT, editing fuzzy matches produced by MT is not really that different from editing fuzzy matches produced by TM. I’m not sure that’s totally correct, but I thought it was an interesting take. Thanks for visiting!
Hey there,
This is a bit of a spam — and about a year late!… so feel free to delete it, but there are a few open-source / free-software machine translation projects out there, for example Apertium (http://www.apertium.org), Moses (http://www.statmt.org/moses) and OpenLogos (http://logos-os.dfki.de/).
Apertium at least allows you to integrate your translation memory with the MT system so that matches in your TM are translated as is, and the rest can be done with the software.
Fran