The Machine Learning Behind Google’s Android Recorder


Over the previous two decades, Google has made data widely available through internet search — from basic information, videos and photos, on maps and tasks. However, a lot of the planet’s information is conveyed via speech. Yet people use recording devices to get significant data in conversations, interviews, interviews and much more, it can be exceedingly hard to afterward sifting through hours of records to recognize and extract data of interest. However, what if there was the capability to automatically transcribe and label long records in real-time, helping you to intuitively find the appropriate information you require when you want it?

Every one of these features run completely on-device, with no necessity for an online connection.

Recorder transcribes address in real-time with an on-device automated speech recognition model according to improvements declared earlier this season. Being a crucial component to lots of Recorder’s clever attributes, Google made certain this version can transcribe long sound recordings (a couple of hours) faithfully while browsing conversation by mapping words into timestamps as computed from the speech recognition version. This permits the user to click a note in the transcription and commence playback beginning from there in the recording or to seek out a word and leap into the specific stage in the recording in which it was being mentioned.

Google transcript
Google Transcript

Read more from Google.


Please enter your comment!
Please enter your name here