JointLK: Joint Reasoning with Language Models and Knowledge Graphs for Commonsense Question Answering

Abstract

Existing KG-augmented models for question answering primarily focus on designing elaborate Graph Neural Networks (GNNs) to model knowledge graphs (KGs). However, they ignore (i) the effectively fusing and reasoning over question context representations and the KG representations, and (ii) automatically selecting relevant nodes from the noisy KGs during reasoning. In this paper, we propose a novel model, JointLK, which solves the above limitations through the joint reasoning of LMs and GNNs and the dynamic KGs pruning mechanism. Specifically, JointLK performs joint reasoning between the LMs and the GNNs through a novel dense bidirectional attention module, in which each question token attends on KG nodes and each KG node attends on question tokens, and the two modal representations fuse and update mutually by multi-step interactions. Then, the dynamic pruning module uses the attention weights generated by joint reasoning to recursively prune irrelevant KG nodes. Our results on the CommonsenseQA and OpenBookQA datasets demonstrate that our modal fusion and knowledge pruning methods can make better use of relevant knowledge for reasoning.