March 2016, go AI program AlphaGo beat Li Shishi, once again sparked concern about artificial intelligence boom. "Take a dozen years bench" cloud known sound CEO Huang Wei, along with the rise of the Internet of things sign, perhaps the best time to usher in the development of speech recognition.
CEO Huang Wei
cloud known sound
before and after 2003, the industry has experienced a climax, because the speech recognition accuracy reaches 90% above the level of the laboratory, stampede in the size of the company. 2005, the world’s leading voice recognition company Nuance listed, the highest market capitalization exceeded $9 billion. In 2008, iFLYTEK (002230) market, the domestic market share has accounted for more than 70%.
, however, only a handful of Companies in this wave in the tight encirclement. Speech recognition accuracy in the actual use of the scene is very low, a short time to solve the basic technical hopeless. Large companies have shut down the relevant departments of voice recognition, boom but short-lived.
platform to achieve corner overtaking
The rise of the mobile Internet
, the depth of machine learning (AlphaGo) of the main working principle of development, so that voice recognition is showing a rising trend again. Apple’s $200 million acquisition of Inc. Siri, 2011 release Siri voice assistant. In the same year, Google acquisition of voice communications company SayNow and voice synthesis company Phonetic Arts, Amazon acquired voice recognition company Yap.
June 2012, Liang Jiaen founded the cloud known sound. This time, iFLYTEK has built deep barriers in the domestic market of speech recognition. In order to achieve the corner overtaking, Liang Jiaen want to be very clear: the first algorithm is second. If the algorithm is the engine, then the data is fuel. Even if the algorithm has been greatly improved, if there is no oil, then the car can not run up.
left: cloud known sound chairman and CTO Liang Jiaen CEO Huang Wei
, cloud known sound
Liang Jiaen decided to do the developer platform. For a start-up company, the platform is undoubtedly heavy burden, the team also had tangled. "We must take the platform to do it, let the market see cloud known sound." Voice recognition technology is the most important thing is to have sufficient data accumulation, not only to find someone to record the voice, but also the user’s real data. The platform can be done in a short period of time to complete a large amount of voice data accumulation, in order to achieve corner overtaking.
Taobao to buy the machine, ordinary game graphics, 10 people less than the team of engineers, self-learning depth learning algorithm, the platform on the line for 3 months.
"no one thinks that such a small company can really do it." Cloud known sound chairman and CTO Liang Jia said it is also quite well now.