Let's do something good together. Let's help Mozilla train SI recognizing Polish speech

Share:
Have you ever thought about it completely disinterestedly contribute to the development of an interesting project in your free time? Such on the Internet can be found in bunches, and examples do not need to go far. For example, Wikipedia, the world's most popular free online encyclopedia, is created by thousands of people from around the world who do not charge any remuneration for their contributions. Similar projects can be multiplied. Today I wanted to offer you an offer to participate in an interesting one projekcie Common Voice, which Mozilla is responsible for. All you have to do is put in a minimum of effort and spend at least a few minutes a day to help teach a human speech machine.

Mozilla Common Voice, which is an open voice recognition system

The Mozilla Foundation rightly notes that speech is natural and … human for us all. That is why the goal of teaching machines to recognize our speech in different languages ​​is an extremely ambitious task. For this purpose, a powerful database of voice samples is required, which must also be verified for correctness.

The company draws attention to the fact that the databases currently used by large companies are closed and unavailable to smaller players, which effectively inhibits technological development and pushes up the prices of implementing interesting innovations. Hence the idea of ​​creating a project Common Voice, i.e. an open speech recognition system available to everyone.

You enter the stage here

Using the site Voice.Mozilla.org You can register voice samples in the language of your choice (preferably your native language), as well as evaluate already sent voice samples from other Internet users.

common voice mozilla 1

These voices contribute to huge databases that anyone can download so that they can be used to train speech-enabled applications. For example, the English language database contains 38 GB of data under the CC-0 license, which consists of 51 072 votes and a total of 1118 checked recording hours from among 1488 submitted. The voices are male and female and have different accents.

common voice mozilla 4

Currently, intensive work is underway to create an appropriate database of voices from Poland, which application developers could then download and improve their solutions – this is currently relatively small. It is worth helping with this, devoting at least 10-15 minutes of your time. Nobody tells you to record samples – just evaluate the existing ones. It is also an appropriate form of support.

Don't you want to "work" for free? Don't be selfish

With amusement and some horror, I follow the voices of project critics on the web, who say that they do not lift a finger on the matter for free. Every day, the same people use free web browsers (including Firefox) from Mozilla, free open source applications and read many free news sites with free ad blockers enabled. The comment is probably superfluous.

I encourage you to offer your support. Let's show that the Polish Internet can unite in a just cause. You can help by developing the initiative by passing to this address.