Directed Graph Networks for Logical Entailment

EasyChair Preprint 2185, version 1

Abstract

We introduce a neural model for approximating propositional entailment, a benchmark task for logical reasoning, based upon learned graph convolutions on directed graphs. The model dispenses with some of the inflexible inductive biases applied in previous work on this domain, while still producing competitive results on the dataset. In particular, model performance on larger problems surpasses all previous work.

Keyphrases: Graph Neural Network, Logical Entailment, automated reasoning, directed acyclic graph

@booklet{EasyChair:2185,
  author    = {Michael Rawson and Giles Reger},
  title     = {Directed Graph Networks for Logical Entailment},
  howpublished = {EasyChair Preprint 2185},
  year      = {EasyChair, 2019}}