- Spotify now has support for Siri on Apple Watch
Spotify has added support for voice commands Siri on any device Apple Watch that has watchOS 6. With the latest version ...
Google is devoting a good part of his time to development of services the development of artificial intelligence that employs quite a few of their products. The most obvious example is the Wizard of Google includes in Android, in the Nest products and others, but there are more services that make use of this machine learning and one of them is his application of recorder.
Not long ago, Google announced a new functionality and it was the fact that there is a further transcription of the audio in real-time. Get a text on the fly about what is recording, even in another language (hello, mode Interpreter). In addition to this transcript, Google allows you to search within a file audio, and now the team from Mountain View has been explained a bit about how it works, without going into very technical.
to Divide, analyze, transcribe, tag
Account Google, as was to be expected, after the instant translation and transcription instant is his Assistant. Ok Google for everything. The processing power of the audio is, in addition, completely offline. There is No uploaded to the cloud, but that everything is processed in the device itself and there is where all the audio is despieza and categorized. But more important, it is tagged with a code easily identifiable by the user.
The audio is split by words, and all of them are referenced to particular points in the text that has been transcribed. In this way it is very easy to go back to any point of the recording, in particular, and to perform searches. All this based, as we have said, in your own transcription. is Each word leads to an exact time frame to go to the post to start listening from there.
But besides this, Google is dedicated to separate the different types of audio that you are recording at that time, all of this analyzing blocks of 50 milliseconds that will coloring in one and another color. Thus, the machine of artificial intelligence Google know when you are talking to, when music is playing, and is also able to recognize what’s playing. All of this through a multitude of separate processes that operate at the same time on the same audio file.
Is parsed, and label the audio in blocks of 50ms, thus, forming markers for audio and voice
Google gets also to recognize different sounds that are being collected simultaneously, and label the dominant. All this, remember, in real-time. But all of this that Google tells us has to do with the process of the recording itself, and leaves something to the end. Once the recording has been completed, Google is able to suggest titles to save the audio in function of what it has been doing.
And in this process also enters the artificial intelligence for analyzes word frequency and the importance of these in the context. So, subtract the words that are considered “empty” at the level of importance, such as “swear words”, and generate a series of tags main.
this is how it works artificial intelligence, or the procedures of machine learning, behind the transcription in real-time of audio to the recorder is from Google. Interpretation and labeling of sound files at the time of the recording. And, of course, is to intervene here, the AI developed by Google, will be more and more efficient with the passage of time.
More information | Google
The news Google explains the operation of the artificial intelligence after its transcription instant audio was originally published in Xataka Android by Samuel Fernandez .
December 19, 2019
- ← Urban innovations: Banks 100% recycled
- Epic Games Store announces the changes and the new features of your store →