Relation-Aware Language-Graph Transformer for Question Answering
This addresses the need for better integration of knowledge graphs in question answering systems, though it appears incremental as it builds on existing GNN-based methods.
The paper tackles the problem of limited relational information usage and interaction between language models and knowledge graphs in question answering by proposing QAT, a relation-aware language-graph transformer, which achieves state-of-the-art performance on datasets like CommonsenseQA, OpenBookQA, and MedQA-USMLE.
Question Answering (QA) is a task that entails reasoning over natural language contexts, and many relevant works augment language models (LMs) with graph neural networks (GNNs) to encode the Knowledge Graph (KG) information. However, most existing GNN-based modules for QA do not take advantage of rich relational information of KGs and depend on limited information interaction between the LM and the KG. To address these issues, we propose Question Answering Transformer (QAT), which is designed to jointly reason over language and graphs with respect to entity relations in a unified manner. Specifically, QAT constructs Meta-Path tokens, which learn relation-centric embeddings based on diverse structural and semantic relations. Then, our Relation-Aware Self-Attention module comprehensively integrates different modalities via the Cross-Modal Relative Position Bias, which guides information exchange between relevant entites of different modalities. We validate the effectiveness of QAT on commonsense question answering datasets like CommonsenseQA and OpenBookQA, and on a medical question answering dataset, MedQA-USMLE. On all the datasets, our method achieves state-of-the-art performance. Our code is available at http://github.com/mlvlab/QAT.