Modeling Global Semantics for Question Answering over Knowledge Bases

TL;DR

This paper introduces gRGCN, a relational graph convolutional network model that captures global and relational semantics for improved question answering over knowledge bases, outperforming existing models.

Contribution

The paper proposes a novel gRGCN model that integrates structure and relational semantics for semantic parsing in KBQA, emphasizing global question understanding.

Findings

gRGCN outperforms existing models on benchmark datasets.

The hierarchical relation attention mechanism effectively captures relational semantics.

Global semantic extraction improves question-to-query graph mapping.

Abstract

Semantic parsing, as an important approach to question answering over knowledge bases (KBQA), transforms a question into the complete query graph for further generating the correct logical query. Existing semantic parsing approaches mainly focus on relations matching with paying less attention to the underlying internal structure of questions (e.g., the dependencies and relations between all entities in a question) to select the query graph. In this paper, we present a relational graph convolutional network (RGCN)-based model gRGCN for semantic parsing in KBQA. gRGCN extracts the global semantics of questions and their corresponding query graphs, including structure semantics via RGCN and relational semantics (label representation of relations between entities) via a hierarchical relation attention mechanism. Experiments evaluated on benchmarks show that our model outperforms off-the-shelf models.