Ensuring inclusion for voice contributions of impaired people


I’m posting that in English on behalf of @Gernelf. The point is about ensuring that people who have speech disabilities such as dyslexia or stuttering can contribute and have their sentences properly validated. This would ensure we have coverage of such way of speaking in the dataset, which is one obvious goal of the project.

First, I think that having validation guidelines displayed during validation, with examples if possible, would help document what we should accept. Incuding some impaired speech into those guidelines should be required.

A second element was that it would be nice to have a way, through the accents selection, to document that. This way, the dataset could be tagged better, and this might also help people working on speech solution for impaired people. A rough accent « impaired speech » could be a good solution, I’m not sure getting finer-grain would be useful here.

I guess it would be useful if an “impaired speech” tag was visible to validators reviewing the clip, so that they know to make accommodations and not class it as misspoken.