Microsoft displays your system to translate English to Chinese voice Mandarin
The head of research at the company, Rick Rashid, said the company expects to have in the near future systems “completely demolish the barriers of language.” In a video demonstration of a conference held in Tianjin, China, Rashid spoke in English, and then rang his own voice, but speaking in Mandarin Chinese.
The system has been called “Deep Neural Net” (DNN) and has been investigated for several years.
“The first step takes my words and find equivalents in Chinese, and although this is not trivial, is the easy part. The second step rearranges the words to be suitable for the Chinese, an important step for a correct translation between languages. Of course, it is likely that there are still errors in the text both in English and in Chinese translation, and the result can sometimes be funny. Still, the technology has been developed to be very useful, “Rashid wrote in one of the blogs of Microsoft .
To create the final Voice version, Microsoft had to record a few hours of a Chinese speaking their language, and about an hour of Rashid speaking English.
The voice translation technologies are not entirely new and have been integrated into our daily lives through tools in smartphones, televisions and consoles, however, these systems are based on a statistical technique known as hidden Markov model and tend to having an error rate of between 20% and 25%.
According to Rashid, DNN has a much lower error rate. “This means that instead of having a word of four or five wrong, now the error rate is one word in seven or eight,” he said. “Although this is still far from perfect, is the most dramatic change in accuracy since the introduction of hidden Markov modeling in 1979, and as we add more data to the training we believe that even better results,” he said.Deep Neural Net, Mandarin Chinese, Microsoft, translation, Voice