Today, while browsing for information for a paper which I am preparing for an upcoming international conference on computational linguistics, I came across this image that says a little more than a thousand words. A company which had depended on a machine to translate an advertisement jingle had apparently gotten spiffed. The result is that large poster bearing half information and plenty self-deprecating humour. The photo (culled from blogamundo.net) came as an apt comment on the problems of machine translations. In this case, not just an unhelpful clog, but a pathetic one.

The question posed here include whether machines would ever replace humans as ideal translators of language data. True, we now have perfect and near perfect free online translating machines/softwares that are almost as outstanding as a human polyglot. But in the end, what value do they add when faced with translations of data that are not only nuanced but also cultural? There are texts in my language, Yoruba, which will never make it into English in one piece and still make sense. In short, my language is likely to remain one of the few ones for whom the machine translators were not really intended.

Good news for human translators then perhaps. In the next ten years, it is estimated that human translators of nearly all European languages will either be jobless or else negligible. Translations softwares can now do their jobs nearly as perfect. In a few years, it is suggested that there might not be need for humans anymore in that department. I beg to differ. If anything, what these free online translators have taught us as a lesson is their incompetence when faced with jobs that require “communicative” competence as opposed to mere “linguistic” competence. In short, do they not just translate/transliterate rather than communicate.

Below is the result of a round-trip translation experiment that I concluded on my blog in February of last year.

Languages: English-French-English. Machine: Babel

Source Data: “What is worth doing at all is worth doing well”

First I keyed it into English-French translation. I get the following result, which I suspect might be correct in French:

“Ce qui vaut la peine de faire du tout vaut la peine de faire bien”.

Now, I translate that same result back into English, exactly as it gave me, using the same software, I get the following:

“What is worth the sorrow to make whole is worth the sorrow to make well.”

Conclusion? Humans will remain. Machines will improve, but would never supplant humans. Well, talk of a true lingua fracas!

More on the language translation experiment at Headfirst into the Meddle