Jointly Learning Clinical Entities and Relations with Contextual Language Models and Explicit Context

TL;DR

This paper introduces a multi-task learning approach that explicitly incorporates contextual information for joint named entity recognition and relation extraction, achieving near state-of-the-art results and surpassing existing methods in end-to-end performance.

Contribution

It presents a novel method of segmenting entities from context and building contextual representations, enhancing joint NER and RE performance.

Findings

Achieves near state-of-the-art performance on NER and RE tasks.

Outperforms SOTA RE system in end-to-end NER & RE with 49.07 F1.

Demonstrates the effectiveness of explicit contextual integration.

Abstract

We hypothesize that explicit integration of contextual information into an Multi-task Learning framework would emphasize the significance of context for boosting performance in jointly learning Named Entity Recognition (NER) and Relation Extraction (RE). Our work proves this hypothesis by segmenting entities from their surrounding context and by building contextual representations using each independent segment. This relation representation allows for a joint NER/RE system that achieves near state-of-the-art (SOTA) performance on both NER and RE tasks while beating the SOTA RE system at end-to-end NER & RE with a 49.07 F1.