LSI_UNED at #SMM4H–HeaRD 2026: Grid-Based Biomedical Named Entity Recognition Across Languages and Entity Types

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, quick

Summary

The LSI_UNED team participated in the MultiClinAI sub-task at the #SMM4H-HeaRD 2026 Workshop, focusing on multilingual clinical named entity recognition (NER) across seven languages. Their task involved identifying diseases, procedures, and symptoms in clinical case reports. The team proposed W2NER architecture-based systems, training a separate model for each language and entity type. For Spanish, a RoBERTa-based model with data augmentation was used, while English and Italian systems employed different biomedical BERT variants. Results showed consistent performance, with Spanish achieving the best overall scores. Data augmentation notably improved recall and F1 for Spanish. English and Italian models were competitive but scored slightly lower, and symptom recognition proved the most challenging entity type across all languages.

Key takeaway

For NLP engineers developing multilingual clinical named entity recognition systems, consider implementing language- and entity-specific models based on architectures like W2NER. You should prioritize data augmentation strategies, especially for languages like Spanish, to boost recall and F1 scores. Be prepared for symptom recognition to be the most challenging entity type, requiring focused effort or specialized techniques.

Key insights

Separate W2NER models per language and entity type effectively address multilingual clinical NER challenges.

Principles

• Data augmentation enhances NER recall and F1 scores.
• Symptom recognition is consistently challenging across languages.
• Language-specific models improve multilingual NER performance.

Method

Train W2NER-based models separately for each language and entity type, utilizing RoBERTa with data augmentation for Spanish and biomedical BERT variants for English/Italian.

In practice

• Apply W2NER for fine-grained clinical NER tasks.
• Prioritize data augmentation for Spanish clinical NER.
• Select specific BERT variants for target languages.

Topics

• Biomedical Named Entity Recognition
• Multilingual NLP
• Clinical Text Mining
• W2NER Architecture
• RoBERTa
• Data Augmentation
• BERT Variants

Best for: AI Scientist, NLP Engineer, Research Scientist

Related on AIssential

• Data Augmentation for Historical NER: A Systematic Comparison of Lexical and LLM-based Approaches — This paper systematically compares lexical and LLM-based data augmentation for Historical Named Entity Recognition.
• Predicting relations between SOAP note sections: The value of incorporating a clinical information model — Fine-tuning a RoBERTa-base model on manually annotated clinical data improves NER for SOAP note sections.
• How to Leverage NER and Advanced NLP Techniques for Life Sciences — NER and advanced NLP are essential for structuring and interpreting the vast unstructured text data in Life Sciences.
• Cross-Lingual Empirical Evaluation of Large Language Models for Arabic Medical Tasks — English-centric LLMs exhibit significant performance gaps and reliability issues in Arabic medical tasks.
• Data Augmentation for Named Entity Recognition in Domain-Specific Scenarios in Portuguese — Data augmentation significantly improves NER performance in low-resource, domain-specific Portuguese datasets.
• EDEN: A Large-Scale Corpus of Clinical Notes for Italian — EDEN provides the largest Italian clinical note corpus and a CRF-filling benchmark for LLM-based information extraction.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.