Integrating Graph Contextualized Knowledge into Pre-trained Language Models

TL;DR

This paper introduces a transformer-based model that integrates contextualized knowledge from subgraphs in knowledge graphs into pre-trained language models, enhancing performance on medical NLP tasks.

Contribution

It generalizes knowledge representation learning to incorporate subgraph contexts, improving knowledge embedding quality and downstream NLP task performance.

Findings

Achieves state-of-the-art results on medical NLP tasks.

Outperforms TransE, indicating better knowledge capture.

Effectively models graph contextualized information.

Abstract

Complex node interactions are common in knowledge graphs, and these interactions also contain rich knowledge information. However, traditional methods usually treat a triple as a training unit during the knowledge representation learning (KRL) procedure, neglecting contextualized information of the nodes in knowledge graphs (KGs). We generalize the modeling object to a very general form, which theoretically supports any subgraph extracted from the knowledge graph, and these subgraphs are fed into a novel transformer-based model to learn the knowledge embeddings. To broaden usage scenarios of knowledge, pre-trained language models are utilized to build a model that incorporates the learned knowledge representations. Experimental results demonstrate that our model achieves the state-of-the-art performance on several medical NLP tasks, and improvement above TransE indicates that our KRL method captures the graph contextualized information effectively.