bayesian-dependency-parsing

Trying to implement a Bayesian neural network to perform graph-based semantic dependency parsing! General problem setup and inspiration is from A Fast and Accurate Dependency Parser using Neural Networks

setup

Required libraries:

tensorflow - 1.15
tensorflow probability - 1.15

running

Edit the Config class in the respective file and run python {classifier/bayesian_classifier}.py. Training and inference should take about 30 minutes without a GPU, within 5 when using one.

results:

Short paper