Whaleshares Logo

Help to decentralize speech recognition

grider123Posted for Everyone to comment on, 5 years ago2 min read

Smart Assistants and Privacy

Speech recognition has become very popular with the advent of Siri, Alexa and the Google Assistant.
While these smart assistentents work well, they all have one problem: They send the voice data of the users to the companies. This endangers the privacy of users.
Recently there was a small scandal concerning the smart assistant Alexa from Amazon. Amazon sent the recorded voice data to the wrong user! (Read more here).
This incident alone shows that we urgently need other speech recognition software that does not collect personal data!


Amazons smart assistent Alexa.

Difficulty of Open Source Speech Recognition

So-called artificial neural networks are frequently used for speech recognition. These try to imitate human learning and learn using sample data. For a complicated task like speech recognition, an enormous amount of training data (i.e. spoken sentences and their text conversion) is needed.
Without the huge amounts of data that Google and co have, it is difficult to achieve the same quality of speech recognition.

Mozilla's Common Voice Project

This is where Mozilla's Common Voice project comes in. Anyone can join and either speak sentences or validate sentences of other users. The language database is published under the free Creative Commons License and can be used by anyone interested free of charge. With just 5 minutes, anyone can help improve the quality of future speech recognition software. And reduce our dependence on companies like Google. I think this is undoubtedly a great thing and will certainly invest a few minutes every day to help this project!

If you're interested you can start here (without a registration!)

Sign Up to join this conversation, or to start a topic of your own.
Your opinion is celebrated and welcomed, not banned or censored!