Joint Biomedical Entity and Relation Extraction with Knowledge-Enhanced Collective Inference

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 6248–6260 August 1–6, 2021. ©2021 Association for Computational Linguistics 6248 Joint Biomedical Entity and Relation Extraction with Knowledge-Enhanced Collective Inference Tuan Lai 1, Heng Ji 1, ChengXiang Zhai 1, Quan Hung Tran 2 1University of Illinois at Urbana-Champaign 2Adobe Research {tuanml2, hengji, czhai}@illinois.edu qtran@adobe.com Abstract Compared to the general news domain, infor- mation extraction (IE) from biomedical text re- quires much broader domain knowledge. How- ever, many previous IE methods do not utilize any external knowledge during inference. Due to the exponential growth of biomedical pub- lications, models that do not go beyond their fixed set of parameters will likely fall behind. Inspired by how humans look up relevant in- formation to comprehend a scientific text, we present a novel framework that utilizes exter- nal knowledge for joint entity and relation ex- traction named KECI (Knowledge-Enhanced Collective Inference). Given an input text, KECI first constructs an initial span graph rep- resenting its initial understanding of the text. It then uses an entity linker to form a knowledge graph containing relevant background knowl- edge for the the entity mentions in the text. To make the final predictions, KECI fuses the ini- tial span graph and the knowledge graph into a more refined graph using an attention mecha- nism. KECI takes a collective approach to link mention spans to entities by integrating global relational information into local representa- tions using graph convolutional networks. Our experimental results show that the framework is highly effective, achieving new state-of-the- art results in two different benchmark datasets: BioRelEx (binding interaction detection) and ADE (adverse drug event extraction). For ex- ample, KECI achieves absolute improvements of 4.59% and 4.91% in F1 scores over the state- of-the-art on the BioRelEx entity and relation extraction tasks 1. 1 Introduction With the accelerating growth of biomedical publi- cations, it has become increasingly challenging to manually keep up with all the latest articles. As 1The code is publicly available at https://github.com/ laituan245/bio_relex Figure 1: An example in the BioRelEx dataset. UIM is an abbreviation of “Ubiquitin-Interacting Motif”. Our baseline SciBERT model incorrectly predicts the men- tion as a “DNA” instead of a “Protein Motif”. a result, developing methods for automatic extrac- tion of biomedical entities and their relations has attracted much research attention recently (Li et al., 2017; Fei et al., 2020; Luo et al., 2020). Many re- lated tasks and datasets have been introduced, rang- ing from binding interaction detection (BioRelEx) (Khachatrian et al., 2019) to adverse drug event extraction (ADE) (Gurulingappa et al., 2012). Many recent joint models for entity and relation extraction rely mainly on distributional represen- tations and do not utilize any external knowledge source (Eberts and Ulges, 2020; Ji et al., 2020; Zhao et al., 2020). However, different from the general news domain, information extraction for the biomedical domain typically requires much broader domain-specific knowledge. Biomedical documents, either formal (e.g., scientific papers) or …

Content truncated to 500 words