Google dedicates a good portion of its time to developing an app to make artificial intelligence that it uses in many of its products. A clear example is Google's assistant installed on Android, Nest products and others, however there are many services that use this machine learning And one of them is his record request.
Recently, Google announced a new functionality and it was a fact that real-time audio distribution was added. Get a note on a piece of paper about that recording, whether it's in another language (hello, translation mode). In addition to this text, Google allows it search within audio file
Divide, analyze, label, label
A Google Account, as expected, after a quick translation and quick write-up is your help. OK Google for everything. The power of processing audios is also completely offline. No uploads to the cloud yet everything is processed on the device itself
The sound is divided into words and all refer to specific points in the text. This way it's really easy to get back to any point of recording especially and do a search. Everything is based, as we have said, in the text itself. Each word leads to a specific time which you can go to later to start listening from there.
But more than that, Google is committed to distinguishing the various audio recordings of the time, all of which analyze 50-mile blocks that maintain one color or another. So, Google's intelligent machine when spoken, when to play music and you can see what's playing. All of this with a lot of different processes running at the same sound file.
Audio is analyzed and recorded in blocks of 50ms, thus creating audio and voice tags
Google has also been able to recognize the various sounds that are collecting at the same time and the label itself is outstanding. All of this, remember, in real time. But all that Google tells us is in terms of the recording process itself, and it leaves something to the end. Once the recording is complete, Google is able to raise headlines to save sound
And this process entails artificial intelligence because the frequency of words and their significance in context are analyzed. As such words that are considered "empty" are excluded from the significance level, as bad names, and a series of advanced labels are produced.
This is how intelligence works, or processes of machine learning, what causes real-time text of audio from a Google recorder. Translation and labeling of audio files during recording. And of course, by intervening here with the AI produced by Google, it will be more efficient and sustainable over time.
More details | Google