Mozilla “collects” voices to make a new speech recognition program

© Provided by: PROTO THEMA S.A. Mozilla “collects” voices to make a new speech recognition program

Everyone can -the last few days- “download” on his computer hundreds of thousands of voice samples already collected by the “Mozilla Foundation” under the new project of “Common Voice”.

This is an open source speech recognition system based on volunteers who recorded their speech. Mozilla’s goal, through the collection of these samples, is to train a new open source application to recognize speech.

The collection of voice samples began a few months ago and to date, “Common Voice” has collected about 400,000 entries from 20,000 different people, a total of about 500 hours of talk time.

Various artificial intelligence systems, like Google’s, are trained in speech recognition using user data. In the case of Mozilla (copyright, among others, of the popular Firefox web browser), the speech recognition engine is open source, which will allow the re-use of this technology.

Until recently, developers, businesses and researchers who wanted to experiment and develop new voice technologies had access to a limited collection of voice data. The collection “Common Voice” is already the second largest known public voice data set, as people around the world gradually add their own voice (initially only in English).

By creating the new public data set, the “Mozilla Foundation” helps overcome barriers and facilitate the development of better speech recognition systems. In the future, it intends to support other languages ​​other than English.

Anyone can download the first edition of the voice collection “Common Voice”  with free license of “common property” at: https://voice.mozilla.org/data.

Source: www.protothema.gr