Call for Participation MultiClinAI Shared Task (SMM4H-HeaRD in the ACL 2026)

Multilingual Clinical Entity Annotation Projection and Extraction

https://temu.bsc.es/MultiClinAI/


MultiClinAI is the first shared task focused on (1) the automatic creation of comparable multilingual corpora and (2) the automatic detection of key clinical concepts (diseases, symptoms, and procedures) in seven languages: Spanish, English, Italian, Dutch, Romanian, and Czech. MultiClinAI will be held as part of the #SMM4H-HeaRD Workshop at the ACL 2026 conference (online).

INFO
Web: https://temu.bsc.es/MultiClinAI/
Data: https://doi.org/10.5281/zenodo.18508039
Annotation guidelines: https://zenodo.org/records/13151040
Registration: https://temu.bsc.es/MultiClinAI/registration/

MOTIVATION:

There has been considerable progress in clinical language technology solutions, resulting in a variety of highly relevant practical applications to cope with the growing amount of biomedical and healthcare-related data and unstructured content sources. In particular, the automatic extraction of key clinical entities such as diseases, symptoms, and procedures is extremely valuable for most clinical data analytics and predictive modeling use cases. Nevertheless, there are very few high-quality annotated corpora, datasets, and annotation guidelines available for the training or robust evaluation of advanced NLP- or LLM-based clinical entity recognition systems, which are typically limited to monolingual scenarios covering only a single language.

Thus, there is a clear need to foster more efficient strategies to generate not only annotated datasets in multiple languages, but also to ensure that they align in terms of the underlying annotation criteria, in order to generate comparable labeled datasets across languages and promote comparable entity extraction systems across languages. Multilinguality is clinically important because healthcare systems document patient information in local languages. However, most clinical NLP tools are developed primarily for English, limiting their applicability in non-English-speaking contexts. Developing multilingual models helps reduce linguistic bias and improves the global applicability of clinical language technologies. Such models enable more equitable AI deployment across different regions and healthcare systems.

Multilingual clinical NLP has numerous important use cases including:

- In international clinical trials, it can be used to extract structured data from trial sites across different countries and to ensure consistent outcome definitions across languages.
- For cohort identification, it enables the identification of eligible patients from unstructured electronic health records (EHRs) and the extraction of phenotypes for observational studies.
-In disease surveillance, multilingual systems can help detect rare diseases or emerging health trends and identify post-marketing drug safety signals.
- For epidemic monitoring, they support the early detection of infectious disease patterns and the analysis of multilingual emergency department notes.
-In cardiovascular and chronic disease monitoring, such systems can track symptom progression across large multilingual datasets and study treatment adherence patterns.
- They also contribute to data standardization by converting free text into structured, interoperable datasets and enabling the secondary use of EHR data.
- Finally, multilingual knowledge graphs can link extracted entities across languages and support federated learning across institutions.

In this context, the MultiClinAI (Multilingual Clinical Entity Annotation Projection and Extraction) shared task addresses the creation and evaluation of comparable multilingual clinical resources across seven languages, focusing on three key entity types: diseases, symptoms, and procedures.

1) MultiClinNER subtask: multilingual clinical named entity recognition across expert-annotated gold-standard datasets.

2) MultiClinCorpus subtask: automatic generation of comparable multilingual clinical corpora through annotation projection techniques.

This setup will enable a robust benchmarking scenario for multilingual clinical NLP approaches.

SCHEDULE:
MultiClinAI Shared Task – training set release (February 6, 2026)
MultiClinNER test set release (March 18, 2026)
MultiClinCorpus test set release (March 25, 2026)
MultiClinNER test set prediction submissions (March 27, 2026)
MultiClinCorpus test set prediction submissions (April 9, 2026)
Result / evaluation returned to teams (April 14, 2026)
Participant proceedings due (April 24, 2026)
Notification of acceptance (May 15, 2026)
Camera-ready papers due (May 25, 2026)
ACL Proceedings due (hard deadline) (June 1, 2026)
Workshop (online) (July 2–3, 2026)

Publications and SMM4H-HeaRD in the ACL 2026 workshop:

Teams participating in MultiClinAI will be invited to contribute a systems description paper for the ACL 2026 Working Notes proceedings and a short presentation of their approach at the ACL 2026 workshop (online).

Main Organizers:
Salvador Lima-López, Barcelona Supercomputing Center (BSC), Spain.
Fernando Gallego-Donoso, Barcelona Supercomputing Center (BSC), Spain.
Jan Rodríguez-Miret, Barcelona Supercomputing Center (BSC), Spain.
Judith Rosell, Barcelona Supercomputing Center (BSC), Spain.
Martin Krallinger, Barcelona Supercomputing Center (BSC), Spain.