Common Voice 100 hours sprint

[nukeador (Rubén Martín [❌ taking a break from Mozilla])]

Hi all,

I want to share the campaign that we have just launched today. If you are contributing to Common Voice in English, please sign up and spread the word!

Help free the English speech recognition

Hey Mozillians, Your next opportunity to help Mozilla is coming up on April 22nd - May 5th. Help Mozilla’s Common Voice help teach machines how real people speak English. [mars] As you might already know, Mozilla’s Common Voice project aims to unlock and open the voice ecosystem to all players by releasing a public domain dataset of voices in different languages that can be used freely without compromising users privacy sending data to the big tech giants. In the coming months we want to ad…

Thanks for your contributions!

[Michael_Maggs]

Hi @nukeador, sound good! I have a couple of questions:

1. How can I see how many English sentences there are available to read that haven’t so far been recorded at all?

2. Do you have an urgent need for additional sentences for the sprint? I can easily upload more if that would be helpful, but the bottleneck is getting them through Sentence Collector review. More reviewers are needed there!

[nukeador (Rubén Martín [❌ taking a break from Mozilla])]

Do you have an urgent need for additional sentences for the sprint? I can easily upload more if that would be helpful, but the bottleneck is getting them through Sentence Collector review. More reviewers are needed there!

We have a plan in place to import a lot of new ones by the 22th thanks to a partner work we have been doing, that need will be covered

[Michael_Maggs]

That’s great to hear. Any ideas on question 1 please?

[nukeador (Rubén Martín [❌ taking a break from Mozilla])]

How can I see how many English sentences there are available to read that haven’t so far been recorded at all?

I would have to consult with our dev to check if we can get a number, I’ll see what I can do next week. How do you think this figure can help you?

[Michael_Maggs]

Many thanks. The number of ‘new’ sentences available to read is a measure of the extent to which the uploads and Sentence Collector validations are keeping pace with the readers. It seems important to keep a good number of unread sentences available at all times, as that’s the most efficient way of getting one recording per sentence which is what the ML team want at the moment. If the number drops to zero, I’ll know we need to do a sprint on sentence uploading and validation. Having the number accessible onscreen somewhere would be very useful.

[dabinat]

I can easily upload more if that would be helpful, but the bottleneck is getting them through Sentence Collector review. More reviewers are needed there!

Do you review your own sentences, Michael? Because I try my best to keep my review queue as close to zero as possible, and it might just be you and I doing the most work in English.

[Michael_Maggs]

Do you review your own sentences, Michael?

Yes, I normally do - because otherwise they seem to sit for ever in my list. I uploaded around 4000 sentences yesterday, of which I still have to review around 2700. I don’t see many other people uploading, and the fact that my sentences remain available for me to review for ages if I don’t add a vote myself suggests that it may well be you and I doing the bulk of the English sentence reviewing work.