Skip to content
/ lambda Public

Two networks, one location-blind and another content-blind, work together to solve augmented SQuAD

Notifications You must be signed in to change notification settings

jooner/lambda

Repository files navigation

PROJECT LAMBDA

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

About

Two networks, one location-blind and another content-blind, work together to solve augmented SQuAD

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages