@inproceedings{danoe-etal-2026-dr,
    title = "Dr-{BERT}-{NL} at {\#}{SMM}4{H}{--}{H}ea{RD} 2026: {DOKTERBERT} {--} Ontology-Grounded Contextual Representations for {D}utch Clinical {NLP}",
    author = "Danoe, Gijs  and
      Voss, Andreas  and
      Hamprecht, Axel  and
      Berends, Matthijs S.",
    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-1.25/",
    pages = "154--159",
    ISBN = "979-8-89176-432-3",
    abstract = "We describe our submission to SMM4H-HeaRD 2026 Task 7, which asks systems tolabel ClinicalImpacts and SocialImpactsspans in Reddit posts about non-medical sub-stance use. We compare four pipeline shapesbuilt on the same DeBERTa-v3-base back-bone: (i) a direct 5-class encoder with a linear-chain CRF head, (ii) a two-stage detect-then-classify pipeline that delegates span typingto an instruction-tuned LLM (Qwen2.5-7Bor Gemma-3-12B, 4-bit NF4), (iii) an auditpipeline in which the same LLM verifies theencoder{'}s predictions, and (iv) a classical-MLvariant that replaces the LLM with an SVMtrained on encoder span embeddings. Across16 configurations, the encoder-only DeBERTa-v3 + CRF configuration is the strongest sin-gle system on the official test split, reaching45.4{\%} strict and 54.2{\%} relaxed F1 {---} +8.6/ +5.3 points above a mental-roberta-basebaseline. LLM audits give a small dev gain thatdoes not transfer to test."
}