Use speakers with voice assistant to record CommonVoice sentences :
Why not? A ‘Repeat after me’ application would collect sentences that are heard and not read by volunteers who give their voice.
By proposing, at each start, to choose the number of words of the sentences, this ‘skill’ would allow each volunteer to solve the difficulty of repetition, to avoid mistakes. Speech to Text would even eliminate errors by simply comparing output and input texts before human validation.
Of course, Alexa or Google Home are not free. But doesn’t Firefox work on Windows or Android? So why not connect CommonVoice on existing proprietary voice OS?
In my opinion, the benefits would be many:
- for all, the opportunity to participate in CommonVoice in situations where you do not have your hands free, and greater availability and speed of launching
- for CommonVoice, the ability to record sentences in real situations (what’s more real than a file from an existing speaker)
- for volunteers on the web or smartphone, the possibility to have only the beautiful long sentences, because all the little sentences easy to repeat by memory will have been made on the connected speakers
- for people with motor or visual disabilities, the opportunity to participate also - for people learning a language, a training tool by repetition
- for people with communication disorders (autistic disorders, phobias, …) the possibility of training in social skills.
Undoubtedly, the warnings that the captured files are to be used in CC0 license would require the collection of consent by registering on the web, with the same account as in the connected speaker (eg gmail account for Google Home). But then, a simple application having access to the CommonVoice textual database will be enough to generate a large amount of voice files.
What do the CommonVoice and DeepSpeech communities think about it?