Communication and digital assistant is a funny thing, it feels like a stubborn child. If you've ever against the Xbox or Siri yelling, you may have lost hope. But the researchers said that a breakthrough in the field of speech recognition and artificial intelligence, recently acquired soon significantly enhance understanding of these electronic products, to make it better and we started to communicate. Bloomberg wrote today, Google is ambitious to expand the field of speech recognition, which I hope to achieve beyond human speech recognition capabilities through technical means.
Communication and digital assistant is a funny thing, it feels like a stubborn child. If you've ever against the Xbox or Siri yelling, you may have lost hope.
But the researchers said that a breakthrough in the field of speech recognition and artificial intelligence, recently acquired soon significantly enhance understanding of these electronic products, to make it better and we started to communicate. Google engineer John Schalke Vic (Johan Schalkwyk) said that this new equipment will not only understand what we mean, but also the context and tone combined with nuances of understanding the deeper meaning.
Schalke Vick is Google in an ambitious research project, hoping to create a company to take advantage of the huge amounts of data voice system. He said that a project they are currently testing laboratory, making the computer can understand and "thinking" people's language.
Recently various inventions, will give a speech recognition in the field of speech recognition and machine learning has brought great changes. Siri's a major inventor says engineers are feverishly developing speech recognition technology, so that with enough intelligence to start a real dialogue with the user. "All areas of speech recognition are a lot of progress has been achieved." SRI International, vice president of the underlying technology development company Siri William Mark (William Mark) said, "This dialogue has become the forefront of interactive technology."
Tim Tuta Seoul (Tim Tuttle) have been waiting for this day for a long time. He received his doctorate in 1997 at MIT and worked at the school's Artificial Intelligence Laboratory. 10 years, he has worked in Silicon Valley companies, and ultimately founded his own company in 2010. Except Labs. 图塔尔company last year began designing a system to increase the complexity of voice commands to mobile applications. For example, when a user into a supermarket, you can buy this functionality informed his broom which is located in the corridor.
"A year ago, we are doing benchmarks, we thought it impossible. But everything has changed. Our company has doubled bet speech, mainly because of recent technological advances have seen a variety." Tuta Dole said, "and the human level equal to or higher than the level of human speech recognition system will be commercialized."
But first, or the first to look at history: two and a half ago, researchers at the University of Toronto and Google published a rather influential papers, the content is "deep neural networks" to guide computer speech technology. A few months later, Microsoft and IBM also co-authored another paper, the Google engineer Jeff Dean (Jeff Dean) is called "Voice 20 years of research in the field maximum progress."
These studies make digital neural networks invention born of a few decades ago resurfaced. The technology of the 1980s in the big data and predictive analytics to achieve a good performance, but were constrained by the speed of the computer. Neural network until recently become viable options, mainly due to speed up computer processing speeds, as well as the development of new software models.
Google Labs has conducted similar research. Six months ago, the team from the old method, called "feed-forward neural networks" to start, promote neural network technology resurrection. This technology allows the system to store more information and to deal with longer, more complex sequences. The breakthrough comes from Google to simplify the underlying code, you can retain more ideas and concepts in the same system, so make it easier for users to ask complex questions, obtain meaningful answers. "System complexity may harm the long-term development." Schalke Vick said.
Google's system is currently using the context, the physical location and other factors make assumptions in order to determine the true meaning of the speech - the whole process of the human brain mode of thinking is similar. Google's latest network technology can improve the efficiency of this process, which handle larger amounts of data than ever before, to answer more complex questions.
To explain the speech recognition technology in the future of work, Schalke Vic mentioned a senior Vietnamese restaurant Google headquarters in Mountain View, a few kilometers away. This restaurant is named Xanh Restaurant typical speech recognition constitutes a challenge because Xanh name (pronounced "Zahn") is difficult to identify. "If I can find its location on the map and say, 'This is a restaurant, which is located in California.' So it will narrow the range immediately." Schalke Vic said, "With semantic technologies, we can dramatically improve quality. "
It sounds simple, but computers, hear a word and then put it in the context of the sentence to the identification, and then combined with geographic information, it is very difficult and time consuming. Today, Google Voice Search has been able to correctly identify the restaurant. Schalke Vick said Google in the future will be able to deal with some other equally ambitious problem.
Schalke Vick said in an internal Google, speech recognition technology has achieved unprecedented progress. Although Google's significant progress have to wait a year or two to apply to the user's phone, but this project has spawned a lot of other projects can be applied to Google's technology. "Moon project development, while at the same time also devised another one hundred useful technology." Schalke Vick said.
Schalke Vick said Google voice recognition technology 3 3/4 years ago, only recognize spoken words. But thanks to accelerate the pace of innovation, Google mobile applications can now be correctly identified 12/13 of the word. According Tuta Er introduced, or how long, "we will live in a world without a keyboard."
No comments:
Post a Comment