An important feature of deep learning is that deep learning will extract features automatically that are predictive. However, what do humans need to do in order to set this up? What comes to mind here is part of speech tagging and sentence parsing. This can't be done with only the text corpora in front of us, but relies on models trained from huge amounts of tagged text. So, how is deep learning able to incorporate POS tagging automatically? Do we include a parser as one of the 'models' we stack?
An important feature of deep learning is that deep learning will extract features automatically that are predictive. However, what do humans need to do in order to set this up? What comes to mind here is part of speech tagging and sentence parsing. This can't be done with only the text corpora in front of us, but relies on models trained from huge amounts of tagged text. So, how is deep learning able to incorporate POS tagging automatically? Do we include a parser as one of the 'models' we stack?