Skip to Main Content
The worldwide adoption of the HL7 Clinical Document Architecture (CDA) is promoting the availability of coded data (CDA entries) within sections of clinical documents. At the moment, an increasing number of studies are investigating ways to transform the narratives of CDA documents into machine process able CDA entries. This paper addresses the reverse problem, i.e. obtaining linguistic representations (sentences) from CDA entries. The approach presented employs Natural Language Generation (NLG) techniques and deals with two major tasks: content selection and content expression. The current research proposes a formal semantic representation of CDA entries and investigates how expressive domain ontologies in OWL and SPARQL SELECT queries can contribute to NLG. To validate the proposal, the study has focused on CDA entries from the History of Present Illness sections of CDA consultation notes. The results obtained are encouraging, as the clinical narratives automatically generated from these CDA entries fulfil the clinicians' expectations.