There’s been a bit of a buzz in the world of machine translation over the past few days. The reason? Well, Microsoft claims it is close to creating the real life equivalent of the universal translator used by Captain Kirk and co in Star Trek.

Rick Rashid, chief research officer at the software firm, showcased the technology at Microsoft Research Asia’s 21st Century Computing event, and it wasn’t long before the video of his presentation was generating a lot of attention. Live on stage, he was able to demonstrate how the system could interpret his words into Chinese just a few moments after he had said them, and relay them back to the audience in his own voice.

This is certainly a breakthrough in machine translator technology, and Mr Rashid said he hopes that in a few more years it will help “completely break down language barriers”. It definitely shows how far computer software has come.

Building the perfect translator

Microsoft developed the system by starting with a statistical speech model based on the hidden Markov modelling technique. This system was developed in the 1970s and collects data from multiple speakers. It uses this as the basis of selecting the correct intonation and cadence of the computer voice.

This has allowed speech technology to come on in leaps and bounds and today you have probably experienced it for yourself when speaking to the automated telephone systems used by banks and other businesses. iPhone users will also be familiar with it if they have ever asked Apple Siri a question.

However, a basic speech system based on this model typically has an error rate of 20 to 25 per cent. Then, a couple of years ago, Microsoft Research teamed up with the University of Toronto to work on a technique called Deep Neural Networks, which was modelled on the best speech recognition system of all – the human brain. This allowed the developers to reduce the word error rate by more than 30 per cent.

Yet these developments have all been in technology that converses in the user’s own language – what about a machine that can translate the words of a speaker live? This is the breakthrough Microsoft claims to have made.

First, Mr Rashid demonstrated a programme that could transcribe the words he was saying, transform each word into the Chinese equivalent and then reorder this so that it made sense. The next step was what generated the greatest positive audience response during his showcase – taking that translation and converting it into speech using Mr Rashid’s own voice.

“We may not have to wait until the 22nd century for a usable equivalent of Star Trek’s universal translator, and we can also hope that as barriers to understanding language are removed, barriers to understanding each other might also be removed,” he explains in his recent blog post.

No replacement for human linguists

This breakthrough is really not to be sniffed at, but by no means should it be seen as a replacement for human interpreters or translators. By Mr Rashid’s own admission, the translation programme is not perfect and “the results can sometimes be humourous”. Likewise, the text to speech system is “not perfect”.

OK, we realise as a language service provider we are always going to favour a human translator or interpreter over a machine version. But when the results speak for themselves it’s hard to believe there will ever be a time when a software programme provides a better service than someone who has trained for years in translating or interpreting.

With a human, you can expect far better accuracy than one error in every seven or eight words (which is the best Microsoft has achieved). In addition, if you’re speaking live, you don’t have to wait for your speech to be transcribed sentence by sentence before it can be interpreted, as your interpreter will only be a few seconds behind you. Of course, a human will also be able to interpret questions from the audience for you so that you can connect with the people who have come to hear you speak.

There’s no denying the Microsoft translator’s ability to speak a different language in your voice is impressive. However, an interpreter will not only match the speaker’s words, but also their intonation and the emotion behind their words.

In terms of text translation, Microsoft’s second step – reordering the words so they make sense – certainly is a breakthrough. If anyone has ever taken advantage of a free website translation and found themselves just as lost when they tried to read it as they would be if it hadn’t been translated, you’ll know what we mean. An effective translation means far more than simply changing the words from one language to another.

Unfortunately for Microsoft and other machine translator developers, it is also about more than just reordering the words. Translation is an art, which requires you to have an understanding not only of languages but also the culture, customs and regional dialects and vocabulary of your target readership.

A professional translator will also have knowledge of your industry, the topic you are writing about and the audience you are targeting. All of this is used to inform the translation. This ensures that it is readable in the target language, explains any technical terminology correctly, and – most importantly – that the original message you were trying to convey is not lost.

As you can see, a human translator or interpreter has a lot to offer. In our opinion, we could all be waiting until the 22nd century for a machine that is able to do the same.