Speak up for #opensource!
Have you heard about the Common Voice
project by Mozilla? They are currently putting together a #voice
databases for various languages and make them accessible with a public domain license to allow developers to work with them. They also plan to create their own voice recognition software.
While I think it's foolish to use one of the contemporary commercial smart speakers that transfer your data to some company without your control, I can think of various good uses of voice recognition, especially for disabled or elderly people. That's why I decided to check out the project.
When I did, I noticed there were very few female #German
voices among the recordings to be verified, so I added some clips myself. Then I learned that women are underrepresented overall and there is a big backlog of #English
soundbites to be verified. They started recording English before they started recording other languages, so there is much more material and it's of worse quality, because they had to figure out how to create a good recording process along the way. By now recording has become quite easy and - be warned - quite addictive. I have had a lot of fun listening to English clips by speakers from various regions. My favourite was someone cheerfully singing "It's a long way from Tipperary, it's a long way from home" with perfect intonation. Yes, voice recognition should get that.
Verifying German clips I liked the recordings of a male speaker whose first #language
is probably French. His #accent
is faint, but noticeable. I realized a voice recognition software needs to be able to deal with all kinds of accents. So I decided to record some English clips as well. Among the German clips I checked few people spoke a recognizable #dialect
. There is one Bavarian woman (while I have lived in Bavaria for over 20 years, people still mistake me for a tourist because of my northern accent), but that was it.
As for other languages, there is very little material yet. Recordings for #French
have begun. Collecting #Spanish
sentences is still work in progress.
Take a look and maybe contribute, if you like the project: https://voice.mozilla.org