Biomedical Entity Linking for Dutch: Fine-tuning a Self-alignment BERT Model on an Automatically Generated Wikipedia Corpus

TL;DR

This paper introduces the first Dutch biomedical entity linking model, fine-tuned on Wikipedia-derived data, showing promising results but highlighting challenges in non-English biomedical NLP.

Contribution

It presents a novel Dutch biomedical entity linking approach using self-alignment BERT and Wikipedia data, advancing non-English biomedical NLP capabilities.

Findings

Achieved 54.7% classification accuracy on Dutch biomedical data

Attained 69.8% 1-distance accuracy in evaluation

Manual review shows 65% correct concept linking for extracted entities

Abstract

Biomedical entity linking, a main component in automatic information extraction from health-related texts, plays a pivotal role in connecting textual entities (such as diseases, drugs and body parts mentioned by patients) to their corresponding concepts in a structured biomedical knowledge base. The task remains challenging despite recent developments in natural language processing. This paper presents the first evaluated biomedical entity linking model for the Dutch language. We use MedRoBERTa.nl as base model and perform second-phase pretraining through self-alignment on a Dutch biomedical ontology extracted from the UMLS and Dutch SNOMED. We derive a corpus from Wikipedia of ontology-linked Dutch biomedical entities in context and fine-tune our model on this dataset. We evaluate our model on the Dutch portion of the Mantra GSC-corpus and achieve 54.7% classification accuracy and 69.8% 1-distance accuracy. We then perform a case study on a collection of unlabeled, patient-support forum data and show that our model is hampered by the limited quality of the preceding entity recognition step. Manual evaluation of small sample indicates that of the correctly extracted entities, around 65% is linked to the correct concept in the ontology. Our results indicate that biomedical entity linking in a language other than English remains challenging, but our Dutch model can be used to for high-level analysis of patient-generated text.