I really want real-world speech data. I think speech-to-text could open the doors to lots of creative tech to help hard of hearing folks, but the dataset barrier is so massive :(
The data is derived from read audiobooks from the LibriVox project and has been carefully segmented and aligned.
If you mean real world as in "with realistic noise", Baidu's DeepSpeech showed that adding noise to clean audio can make a resilient model. Indeed, having clean audio is better as you can add different types of noise to clean data to, in effect, create regularization.
You can correct many of it's errors with simple language models. For instance the phrase "wreck a nice beach" has much less probability than "recognize speech". So if the model isn't sure which one it is, you can assume it's the more probable one. Then train on that, and it will get even better at recognizing those words.
Why not just build that heuristic into your initial supervised "weak" model? Training on data labeled by a model introduces no new information, so you're not gaining anything there.
The way I'm describing it is probably not the optimal way to do it. I don't know if there is a better way. But the point is it must be possible to take advantage of the vast quantities of unlabelled data we have. Humans brains somehow do something similar.
Semi supervised learning is really cool. I saw one successful example where they labelled just a few emails as spam and not spam. Then they trained used their weak classifier to label thousands of unclassified emails, and then used those as training data for an even stronger model. It actually works: http://matpalm.com/semi_supervised_naive_bayes/semi_supervis... https://en.wikipedia.org/wiki/Semi-supervised_learning
It's better to go with movies and audiobooks because the text 'transcription' is human generated. Aligning known text to audio is much more accurate than speech recognition.