talnarchives

Une archive numérique francophone des articles de recherche en Traitement Automatique de la Langue.

Adaptation Datasheets for Plain Language in Health: A Pragmatics-Centered Traceability Schema

Rémi Cardon, Lourdes Moreno, Paloma Martínez

Abstract : Plain Language (PL) adaptation of healthcare texts involves pragmatic decisions (linked to relevance, ordering, prominence, actionability) that affect comprehension and safety. Traceability is often limited to before-and-after text and global readability metrics, making these decisions hard to audit. We propose a lightweight schema that records pragmatic decisions as structured, span-linked fields, intended as an annotation layer for existing PL corpora and as a documentation artifact for LLM-assisted workflows. We illustrate the schema on a Spanish post-consultation clinical note adapted for patients/caregivers, report descriptive statistics, and position it with respect to existing operation-based annotation frameworks.

Keywords : simplification, pragmatics, traceability, healthcare NLP