@inproceedings{rey-blanes-etal-2026-icb,
    title = "{ICB}-{UMA} at {\#}{SMM}4{H}{--}{H}ea{RD} 2026: Hybrid Clinical Entity Projection for {M}ulti{C}lin{AI}: Adaptive Candidate Windows, {XGB}oost, and {LLM} Refinement",
    author = "Rey-Blanes, Alvaro  and
      Gim{\'e}nez-G{\'o}mez, Sara  and
      Veredas, Francisco J.  and
      Moreno-Barea, Francisco J.",
    editor = "Lopez-Garcia, Guillermo  and
      Gonzalez-Hernandez, Graciela",
    booktitle = "Proceedings of the 11th Social Media Mining for Health Research and Applications ({SMM}4{H}-{H}ea{RD} 2026) Workshop and Shared Tasks",
    month = jul,
    year = "2026",
    address = "San Diego, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl-workshops/2026.smm4h-1.21/",
    pages = "127--132",
    ISBN = "979-8-89176-432-3",
    abstract = "This paper presents our submission to the MultiClinAI Shared Task (Gallego-Donoso et al., 2026) on cross-lingual clinical entity annotation projection from Spanish to English. Our system transfers expert annotations for $\textit{Diseases}$, $\textit{Symptoms}$ and $\textit{Procedures}$ entities. The approach integrates three core components: adaptive candidate window generation, an XGBoost classifier leveraging surface and semantic features, and an LLM-based post-processing stage to resolve complex misalignments. Our highest-performing run ranked 3rd on the official leaderboard, achieving strict F1 scores of 0.737, 0.549, and 0.538 for $\textit{Diseases}$, $\textit{Symptoms}$ and $\textit{Procedures}$, respectively. These results show that combining supervised candidate scoring with targeted LLM refinement provides a robust strategy for clinical entity projection."
}