The universal translator closer thanks to Microsoft Research
On the odd occasion we talked about Microsoft Research , a division of Microsoft dedicated to research and technology foresight that works with some of the world’s most prestigious universities. Projects like Holodesk or contact lenses that measure blood glucose levels are examples of projects that are developed in this division of Microsoft in which experiments with new interfaces and smart devices that can improve our daily lives. With the idea of eliminating language barriers, a Microsoft Research has long been working in collaboration with the Universities of Toronto in voice recognition system that is able to translate, in real time, the words into another language and, at a conference , staged a demonstration in which the error rate is one word for every 7 or 8, the lowest achieved to date.
Rick Rashid, head of research in China was giving a lecture and presented the latest results of the project in real time translator (who already has two years of existence). So far, all projects that have been developed in this regard had not been able to reduce the error rate beyond 20% -25%, that is, one in four was incorrectly translated words, but after two years of work , Microsoft Research has managed to lower that threshold to the point that the system is capable of simultaneous translation and error, on average, in a word for every seven or eight.
In the video, along with Rashid’s explanation, we can see a demonstration in which the system is able to translate the Chinese what they are saying in English, something that has been possible to get by with the use of neural networks in the system that after a learning process, can better recognize voice patterns and therefore recognize words and proceed to its translation.
And though it might seem simple, this translation “on the fly” is nothing simple two phases. On one hand, the system should recognize the voice, and make out the words, in this case, translated into Chinese but also the order of the words does not have to follow the same sequence as they must adapt to the grammar of the language.
Although the experiment may seem removed from the universal translator from Star Trek, it seems that we are not so far from breaking down language barriers and, in fact, services like Google Translate or Bing Translate (with its spectacular mobile application) every day to help us communicate in other languages and maybe in a few years, we can bring a mobile version of the “universal translator” in our smartphones.Tags: languages, translation, Video, Microsoft Research