I had written about post-editing a few weeks ago, ending with the promise to update when something new has happened. Well, it has, although not quite as might have been expected.
Not I received a post-editing job, but rather my students. From me.
In order to make things a little more interesting in my translation class and to introduce them to this undeniably existing and hard-to-avoid area of language services, I took a text, ran it through three different machine translation services, split the class in three and had them post-edit the texts.
When we started comparing the different versions, it was pretty interesting to see how the results differed from one sentence to the next.
Sometimes, all three had almost the same translation, sometimes two were quite similar, and then again they all differed.
When it came to accuracy, that too was different, both from sentence to sentence and from MT to MT, sometimes showing a complete misunderstanding of the original, sometimes adding things that weren’t there or even leaving vital parts out.
And even if the translations were correct, the oftentimes didn’t sound very good; the text didn’t flow as if it had been written by a native speaker – or translated by a (good) human translator.
The sentences were nothing had to be changed were almost non-existent.
Basically, everything I had told them beforehand about the shortcomings of MT were proven beautifully – what more can a teacher hope for! 😉