Clinical named entity recognition in the Portuguese language: a benchmark of modern BERT models and LLMs

2026-03-27

AI Summary

This study addresses the lack of established benchmarks and severe class imbalance in named entity recognition (NER) for Portuguese clinical texts. It presents the first systematic evaluation of multilingual BERT variants—including BioBERTpt, BERTimbau, ModernBERT, and mmBERT—alongside state-of-the-art large language models such as GPT-5 and Gemini-2.5 on this task. To mitigate data imbalance, the authors integrate iterative stratified sampling, weighted loss functions, and oversampling strategies. Experimental results demonstrate that mmBERT-base achieves the best performance under resource-constrained conditions, attaining a micro F1-score of 0.76. The findings validate the effectiveness of the proposed approach and establish the first comprehensive benchmark and practical solution for Portuguese clinical NER.

Abstract

Clinical notes contain valuable unstructured information. Named entity recognition (NER) enables the automatic extraction of medical concepts; however, benchmarks for Portuguese remain scarce. In this study, we aimed to evaluate BERT-based models and large language models (LLMs) for clinical NER in Portuguese and to test strategies for addressing multilabel imbalance. We compared BioBERTpt, BERTimbau, ModernBERT, and mmBERT with LLMs such as GPT-5 and Gemini-2.5, using the public SemClinBr corpus and a private breast cancer dataset. Models were trained under identical conditions and evaluated using precision, recall, and F1-score. Iterative stratification, weighted loss, and oversampling were explored to mitigate class imbalance. The mmBERT-base model achieved the best performance (micro F1 = 0.76), outperforming all other models. Iterative stratification improved class balance and overall performance. Multilingual BERT models, particularly mmBERT, perform strongly for Portuguese clinical NER and can run locally with limited computational resources. Balanced data-splitting strategies further enhance performance.

Problem

• Clinical Named Entity Recognition
• Portuguese language
• class imbalance
• benchmark
• multilabel NER

Innovation

• clinical NER
• Portuguese language
• mmBERT
• iterative stratification
• class imbalance