QAGCN: Answering Multi-Relation Questions via Single-Step Implicit Reasoning over Knowledge Graphs

TL;DR

QAGCN introduces a simplified, end-to-end graph convolutional network approach for multi-relation question answering on knowledge graphs, achieving competitive results with less complex reasoning mechanisms.

Contribution

The paper proposes QAGCN, a novel GCN architecture that performs single-step implicit reasoning for multi-relation QA, simplifying and improving upon existing multi-step explicit reasoning methods.

Findings

QAGCN achieves state-of-the-art or competitive performance.

The method simplifies reasoning process and improves efficiency.

Extensive experiments validate the effectiveness of QAGCN.

Abstract

Multi-relation question answering (QA) is a challenging task, where given questions usually require long reasoning chains in KGs that consist of multiple relations. Recently, methods with explicit multi-step reasoning over KGs have been prominently used in this task and have demonstrated promising performance. Examples include methods that perform stepwise label propagation through KG triples and methods that navigate over KG triples based on reinforcement learning. A main weakness of these methods is that their reasoning mechanisms are usually complex and difficult to implement or train. In this paper, we argue that multi-relation QA can be achieved via end-to-end single-step implicit reasoning, which is simpler, more efficient, and easier to adopt. We propose QAGCN -- a Question-Aware Graph Convolutional Network (GCN)-based method that includes a novel GCN architecture with controlled question-dependent message propagation for the implicit reasoning. Extensive experiments have been conducted, where QAGCN achieved competitive and even superior performance compared to state-of-the-art explicit-reasoning methods. Our code and pre-trained models are available in the repository: https://github.com/ruijie-wang-uzh/QAGCN