Machine translation might seem like a new phenomenon, but it’s actually 58 years old this month. Researchers at IBM and Georgetown University began developing an automated Russian to English translator in 1954, with operators laboriously punching messages onto cards.

Today, tools such as Google Translate are lightning-fast, covering hundreds of language combinations. But despite the optimism of those early researchers, computers aren’t yet close to replacing human linguists. In our latest video, Jack Waley-Cohen explains some of the problems with machine translation tools:

One issue is computers take language too literally. Unlike humans, they can’t detect the context or  nuances of meaning, and are confused by multiple definitions. They can also get lost by idioms or figures of speech.

For example, in English, we keep our fingers crossed for something, while Germans squeeze their thumbs. An Italian doesn’t literally have a fly jumping on his nose – it just means he’s annoyed!

Humans can also navigate complex cultural norms. Many French learners are confused by whether to use the informal tu or formal vous. Getting it right in  Japanese is even more fraught with difficulties. If you’re doing business in the language, you’ll have to get to grips with different verb forms, depending on how polite and respectful you need to be!

Languages aren’t static. They evolve constantly, as speakers coin words or change their meanings. Spotting new words, such as “clicktivism” or “mumpreneur”, is a  popular sport for bloggers and columnists. And terms such as “Arab Spring” or “squeezed middle” can rapidly enter our common language.

And mistakes often creep in when languages differ greatly from English. Many machine translation tools stumble over Russian, which tends to add “extra information” to words, or Japanese, with its distinct word order.

Of course that’s not to say machine translation doesn’t have its uses. Google Translate research scientist Ashish Venugopal says their goal is to produce an “80 per cent solution”, that can convey  the gist of text in foreign languages. His example: it can help you understand the main points of a political speech, but not produce a polished version suitable for publication.

Our post-edited machine translation services are an accurate and cost-effective solution for translating large quantities of documents. The software can rapidly translate thousands of words, while a human editor checks it for errors.

There’s no doubt researchers will continue to develop more sophisticated tools. But we still think human translators will be around for the foreseeable future. Do let us know what you think!

tags: language tools, machine translation, Post-Edited Machine Translation