A way of using two separate Differentiable Neural Computers (DNCs) to augment the performance of machine comprehension QA.
vanilla.py > access SquadObject that has been produced after running prosquad.py. this is where a vanilla DNC will be run
wordchar_embed.py > GloVe word embedding is concatenated with char-level embedding
prosquad.py > preprocessing of SQuAD documents
char_encode.py > Character-level embedding carried out here
char_embed.py + charmodel.py + train_char.py > Network through which character-level embedding is trained
models > Trained model is saved here
data > where SQuAD dataset is stored