Back on topic again – it looks like things are humming all over. Posts at my usual circuit of daily reads have slowed down a lot as well.
One interesting post I did encounter though was at a blog called www.betterlocalization.com, by a guy named Denver Wu. His post, “Machine Translation, here now.” painted a very optimistic view of the abilities of Machine Translation. (Based on his other blog posts it seems like he recently started working at a company that makes their own MT solution.)
I certainly agree with the sentiment that Machine Translation has some place in the translation workflow but I take the promise of Machine Translation rivaling that of a human with a huge mound of salt. Admittedly I’m firmly in the camp that believes for most translation work a human translator will always need to be in the mix. Machine Translation will certainly continue to evolve and be more effective but realistically, with few exceptions, I think many companies would be insane to just let a computer run free with translation.
I also agree with his sentiment that Translation Memory is just another type of Machine Translation – at the end of the day, especially with something as complex and subjective as language any translation tool needs to be taught how to process language.
At the end of the day though I think there’s one aspect of translation that will always remain out of the grasp of machine translation – Context. Machine Translation can take two sentences it’s seen before and translate them. But what if those two sentences came from different documents? Sometimes you’ll get lucky and they’ll flow together properly – but the risk is they make a borderline non-sensical pairing. Have this happen across several sentences in a row and your translation, which although technically correct at a sentence or word level will lose it’s meaning in the grander scheme of things.
At the end of the day technologies like Machine Translation can really only be used as a tool in the translation process. Someone with the ability to process things like context, tone & voice – while considering localization considerations – still needs to be in the mix.
Machine translation certainly has a place in automating the mundane – translation tasks that are repetitive, consistent and have little room for misunderstanding and little change in context.
At the end of the day, our goal as technologists shouldn’t be to replace the translator – they’re an essential part of the process – but rather build and create tools that automate the mundane so they can focus on the exceptions. In the end I think it means more efficient, higher quality translation for all of us.
Technorati: translation, machine translation, translators, opinion