Gadgetory

I mean afraid with CNET news I'm here at Microsoft's 10 fair in Mountain View where we're seeing some of what Microsoft Research is up to I'm here with Frank Saida for Microsoft Research agent who is going to show us the translating telephones so I'm gonna ask him a few questions he's gonna answer us in German and then we're gonna see how the translator does so Frank what are we seeing here we show a system here for automatic translation of language so it did pretty good how long has Microsoft been working on disesteem community the transition in taking opinion after Marcus or fashion the system now three different technologies on Microsoft Research are annoying for nine years at Microsoft all nine years in Beijing in China and are you more of a soccer fan or basketball I started as a German and then of course one of football and what's your favorite food my thing seems into spaghetti labeling session or spaghetti didn't translate the word actually I recognized what are some of the things if the translation engine is best at and what are some of the things that a major problem is that strange written text translation system but written text is very widely spoken language so basically the hardest part you can say this in English solidest part is that the text-to-speech part is the translation and then it's hard to translate spoken and written language so there are a lot of heart problems right so speech recognition by itself is actually very hard especially if you speak you know in a natural pace you stutter you repeat yourself to correct yourself and so on that is something that we are addressing in this system by you know having the system adapts to your voice over time so longer use the system the better it gets and we have actually used the system in the English version for about the year now and we get very very reasonable accuracy chance for that on the translation side the problem is really that because it's all based on machine learning right we need to get parallel texts in two languages for conversational style and that is very very different home it's much easier to get the four written text documents that are out on the web are typically written style so the challenge that we are facing here is that my German for example when it's because Paisley is not from Attica it doesn't fit into the normal grammar rules and that's something that confuses the translation system quite a bit thank you so much for taking the time I really appreciate it ashamed like to happen huh