Document Structure aware Relational Graph Convolutional Networks for Ontology Population

TL;DR

This paper introduces a document structure-aware relational graph convolutional network that improves ontology population accuracy by leveraging document structure, outperforming standard R-GCN models by approximately 15 percentage points.

Contribution

It proposes a novel method that incorporates document structure into relational graph convolutional networks for more effective ontology population.

Findings

Achieves about 15 points higher accuracy than baseline R-GCN models.

Utilizes document structure to enhance the learning of ontological relationships.

Demonstrates improved performance in ontology population tasks.

Abstract

Ontologies comprising of concepts, their attributes, and relationships are used in many knowledge based AI systems. While there have been efforts towards populating domain specific ontologies, we examine the role of document structure in learning ontological relationships between concepts in any document corpus. Inspired by ideas from hypernym discovery and explainability, our method performs about 15 points more accurate than a stand-alone R-GCN model for this task.