Every time there's a voice recognition post here someone comments about acquiring data. Why is this method better than having access to all of the video and podcasts sites on the internet?
You can get people to utter the same sentence.
Righto. Everyone is saying the same thing so it's the cleanest data set you can get.