A Google insider has revealed the search engine firm has successfully developed the hardware prototypes necessary for producing translations in real-time. In an interview with The Times newspaper this weekend, Hugo Barra, a vice-president of Android, said the technology giant has “tonnes of prototypes of that sort of interaction”. Does this mean we could soon have access to a Star Trek-like universal translator or a Babel fish, as made famous in The Hitchhiker’s Guide to the Galaxy? Indeed, this fictional fish works by being inserted into the user’s ear, much like how Google’s technology will work through a smartphone held to the ear. The new system will feature on people’s mobile phones and will enable them to hold conversations with people speaking in a different language. While one person speaks into the phone in their mother tongue, the person at the other end hears their own native language, meaning the phone itself acts as the Babel fish. Mr Barra notes that the universal translator is something he plays with regularly so he can monitor how far it has come. Already, some language pairs, such as English and Portuguese, are proving successful in producing accurate translations. However, it could still be some time before holidaymakers are able to use their smartphone as a portable interpreter to accompany them on their travels. Indeed, the biggest hurdle to overcome is that of syntax and the way it varies in every language. Syntax is the linguistic arrangement of words in a sentence. The order of words is different depending on the language being spoken, which is why machine translations often return gibberish results. While the technology has successfully changed each word from one language to another, it lacks the intelligence to re-order them so that they make sense in the target language. A further stumbling block is that of speech recognition. Google Translate and other, similar technology, uses the written word, but a universal translator on a phone would need to translate human speech as it is spoken. Regional dialects, accents, slang and other differences between the written and the spoken word can all serve to trip up a machine attempting to understand and correctly translate what it hears. Yet Mr Barra is confident in the abilities of the technology, telling The Times it works most of the time when used in “controlled environments”. Of course, this is not the first attempt to take the universal translator out of the realms of science fiction and into reality. Lexifone is an Israeli start-up company that recently launched a real-time telephone translation service. The system works by analysing the conversation and producing the most relevant translation based on the topic, with human translators employed to analyse data as it comes in and improve the system based on these requests. However, tests by reporters at the Associated Press found the technology is still far from a being a real rival to human interpreters. The finished translation took some time to come through and when it did the results were often unreliable, particularly if the tone of the conversation had been informal. Max J Rosenthal, writing for the Associated Press, said: “Star Trek’s universal translator will have to wait.” Meanwhile, in November 2012 Microsoft unveiled its own universal translator at Microsoft Research Asia’s 21st Century Computing event. Rick Rashid, chief research officer at the software firm, demonstrated the technology live at the event by using it to interpret part of his speech into Chinese, which was spoken in his own voice. He explained that a statistical speech model based on the Markov technique had been the starting point of the machine translator, and this was teamed with the speech recognition method Deep Neural Networks. Microsoft Research worked with the University of Toronto to develop the Deep Neural Network, which recognises and processes speech in a similar way to the human brain. Microsoft’s machine interpreter works by transcribing the spoken words, translating these into the target language at the most basic level, and then reordering the words so they fit with the syntax of that language. The results are then delivered back in the voice of the speaker. Clearly, machine translators and interpreters are improving all the time. However, they continue to be no substitute for the real thing. Microsoft’s Mr Rashid, Google’s Mr Barra and the owners of Lexifone have all separately admitted that their technology currently lacks the accuracy and efficiency of a skilled human translator. A professional linguist will not only tackle slang, syntax and accents with ease, but they will also localise the results so they make sense to the target audience. So, for now, a qualified human interpreter is the only guide we want by our side, whether we’re exploring the galaxy or a little closer to home.