Input features of DNNs

[Chiara]

I am not sure I get how the linguistic features that provide the input to the NN are obtained. In other words, how are the vectors of “answers” to binary questions about linguistic context/prosoy obtained: is it decision trees, similarly to HMM based synthesis? At what point is this done, and by what model?
Thank you

[Simon]

I’ll add more detail in the next lecture about how the linguistic specification from the front end is converted to a vector of numerical (mostly binary) features, ready for input to the DNN.

[Author]

Posts