Zero Resource Speech Recognition - What's the Difference?

The zero resource community is trying to redefine the stream that you need to follow. We're seeing that a lot of the standard supervised speech processing techniques are just not working. I think there's a big push to try and get these neural models, which are working so well in the supervise case, do al so kind of work very well in the zero resource case. And for that also makes some people use am am, more classical galchian mixture mace models and an unsupervised hiden mark model.