Google is developing offline version of Google Now: 7 times faster times-Google Now, voice Assistant, Siri,Cortana, Cortana-IT information

On March 16, Google is building a smart phone for offline operation of voice recognition systems. Because smart-phones have limited storage space, the company database has been drastically reduced, while retaining the accuracy of speech recognition.

Today, Apple's Siri, Microsoft Cortana and Google's speech recognition systems such as Google Now has become more and more perfect, allows smartphone users to more easily and quickly, and interact with their device, but they all need to connect to the Internet to use.

In this regard, the company has developed a solution, they said in a recent research report, if speech recognition algorithm concentrates to a smart phone, instead of running on the server, to make speech recognition much more reliable, and does not consume a lot of power.

It was reported that Google is working on this offline version of Google Nexus 5 Now tested speech recognition system, it runs faster than the network versions of 7 times faster, it can now to 3 million from Google searches for the identification of anonymous speech samples.

Researchers found that the offline version of Google Now the chance of error is 13.5%, 5% more than the network version. The study still has room to continue upgrading, also proved the feasibility of the scheme.

谷歌正开发离线版Google Now：速度快7倍 - Google Now,语音助手,Siri,Cortana,小娜 - IT资讯

3月16日消息，谷歌目前正在构建一个能够在智能手机上实现离线运行的语音识别系统。因为智能手机存储空间有限的关系，该公司已经对语音库进行了大幅度削减，但是同时保留了语音识别的准确性。

现如今，苹果Siri、微软Cortana和谷歌Google Now等语音识别系统已经变得越来越完善，允许智能手机用户更加方便快捷地和他们的设备进行交互，但是，它们都需要连接到互联网才能使用。

对此，谷歌公司已经开发出了一种解决方案，他们在最近的一份研究报告中表示，如果将语音识别算法浓缩到智能手机上，而不再在服务器上运行的话，就能够让语音识别变得更加可靠，也不会消耗大量的功率。

据报道，谷歌正在Nexus 5上对这个离线版的Google Now语音识别系统进行测试，其运行速度比连接网络的版本要快7倍，它目前已经能够对300万个来自谷歌搜索的匿名语音样本进行识别。

研究人员发现，离线版Google Now的出错几率为13.5%，比连接网络的版本要高出5%。这项研究仍有继续提升的空间，同时也证明了这一方案的可行性。