Incorporating Domain Knowledge into Medical NLI using Knowledge Graphs

TL;DR

This paper enhances medical natural language inference by integrating structured domain knowledge from UMLS knowledge graphs and sentiment information into existing embedding-based models, leading to improved performance.

Contribution

It introduces a method to incorporate knowledge graph embeddings and sentiment data into NLI models, advancing the integration of structured domain knowledge in medical NLP.

Findings

Knowledge graph fusion improves NLI accuracy

Sentiment information further enhances performance

Method outperforms baseline BioELMo models

Abstract

Recently, biomedical version of embeddings obtained from language models such as BioELMo have shown state-of-the-art results for the textual inference task in the medical domain. In this paper, we explore how to incorporate structured domain knowledge, available in the form of a knowledge graph (UMLS), for the Medical NLI task. Specifically, we experiment with fusing embeddings obtained from knowledge graph with the state-of-the-art approaches for NLI task (ESIM model). We also experiment with fusing the domain-specific sentiment information for the task. Experiments conducted on MedNLI dataset clearly show that this strategy improves the baseline BioELMo architecture for the Medical NLI task.