Microsoft Research’s Chief Research Officer delivers a promising speech recognition keynote that reveals the company’s advanced translation technology.
Human language is highly complex. Different accents, pronunciations, idioms, slang, grammar exceptions and patterns -- all these things make it virtually impossible for a machine to translate accurately until now. Understanding that every human speaks differently even when they utter the same word in a different situation or context, Microsoft Research has developed a sophisticated program that can detect the nuances in the human language as it’s spoken and convert it in real-time to the language of choice. After compiling research and data of various speech patterns around the world, the software demonstrates miraculous progress in these programs.
In Rashid’s speech recognition keynote, he gives a brief demonstration of the mechanical translator’s ability. As the software translates his presentation in real-time, audience members can see that although the interface isn’t perfect, it has come a long way since translating sites like Babylon, Bing and Google, improving the accuracy by 30%. In his final demo, the speech recognition software displays the speaker’s words in English as well as Chinese—the chosen language by Rashid—and is able to mimic his voice while delivering an accurate translation of his sentences in the Asian language.