Triple Classification for Scholarly Knowledge Graph Completion

TL;DR

exBERT leverages pre-trained transformer models for triple classification to improve scholarly knowledge graph completion, outperforming baselines on multiple datasets and providing new scholarly datasets for research.

Contribution

The paper introduces exBERT, a novel method using transformer models for triple classification in scholarly KGs, and provides new datasets for the research community.

Findings

exBERT outperforms baseline methods on scholarly KG datasets.

The method improves triple classification, link prediction, and relation prediction.

Two new scholarly datasets are introduced for KG research.

Abstract

Scholarly Knowledge Graphs (KGs) provide a rich source of structured information representing knowledge encoded in scientific publications. With the sheer volume of published scientific literature comprising a plethora of inhomogeneous entities and relations to describe scientific concepts, these KGs are inherently incomplete. We present exBERT, a method for leveraging pre-trained transformer language models to perform scholarly knowledge graph completion. We model triples of a knowledge graph as text and perform triple classification (i.e., belongs to KG or not). The evaluation shows that exBERT outperforms other baselines on three scholarly KG completion datasets in the tasks of triple classification, link prediction, and relation prediction. Furthermore, we present two scholarly datasets as resources for the research community, collected from public KGs and online resources.