Google NowShort Bytes: Google has published a research paper titled “Personalized Speech Recognition In Mobile Devices” which points towards their intentions to give self-thinking power to their Google Now voice assistant. The goal is to create an on-device speech recognition system which could process commands without the need of being connected to the internet every time.

Google is pushing their AI ambitions one step closer to reality with every tick of the clock. Addition to this is a new update planned for the card-loving Google Now personal assistant. All you need to say is “Okay Google” and the digital butler will be there at your service within a matter of seconds.

The only downside is that some of the features don’t work if your device isn’t connected to the internet as Google Now sends your voice commands to Google server for better interpretation of what you’re trying to say.

According to a research paper authored by some Google minds, a futuristic variant of the Google Now voice assistant is being developed by the search giant. The new personalised Google Now won’t require internet access – up to some extent – to process your voice commands as it will be intelligent (artificially) enough to understand you. The commands will be saved for later use and sent to the servers if the voice assistant feels them hard to understand.

The onboard speech recognition system won’t eat up your system memory as it occupies only around 20 megabytes of space, 20.3 megabytes to be precise. It could deliver around 7 times faster performance against the real-time version when tested on a Nexus 5 smartphone, says the research paper. A “combination of vocabulary injection and on-the-fly language model biasing” is used to accomplish a personalised model of your talking habits.

2000 hours of anonymized voice commands – each having 20 different variations to simulate indoor and outdoor conditions – from the Google voice search traffic is fed to the voice assistant. The training sessions for the artificially intelligent voice iteration of Google Now are designed to resemble daily life scenarios which are attained by introducing background voices from recordings and YouTube videos along with voice commands.

Google plans to present their new research at the 41st International Conference on Acoustics, Speech and Signal Processing which will be held between 20 and 25 March 2016 in Shanghai, China.

You’ll find these interesting:

SHARE
Aditya Tiwari
Aditya likes to cover topics related to Microsoft, Windows 10, and interesting gadgets. But when he is not working, you can find him binge-watching random videos on YouTube (after he has wasted an hour on Netflix trying to find a good show). Reach out at [email protected]