Scalable Multi-Hop Relational Reasoning for Knowledge-Aware Question Answering

TL;DR

This paper introduces MHGRN, a scalable multi-hop relational reasoning module for knowledge-aware question answering, enhancing interpretability and efficiency over external knowledge graphs in pre-trained language models.

Contribution

It presents a novel multi-hop graph relation network that unifies path reasoning and graph neural networks for improved QA performance and transparency.

Findings

Effective on CommonsenseQA and OpenbookQA datasets.

Improves interpretability of reasoning process.

Scalable to large knowledge graphs.

Abstract

Existing work on augmenting question answering (QA) models with external knowledge (e.g., knowledge graphs) either struggle to model multi-hop relations efficiently, or lack transparency into the model's prediction rationale. In this paper, we propose a novel knowledge-aware approach that equips pre-trained language models (PTLMs) with a multi-hop relational reasoning module, named multi-hop graph relation network (MHGRN). It performs multi-hop, multi-relational reasoning over subgraphs extracted from external knowledge graphs. The proposed reasoning module unifies path-based reasoning methods and graph neural networks to achieve better interpretability and scalability. We also empirically show its effectiveness and scalability on CommonsenseQA and OpenbookQA datasets, and interpret its behaviors with case studies.