Information extraction from radiology reports mingled two languages

This study presents overall of information extraction (IE) for SNUH (Seoul National University Hospital) radiology reports coexisted Korean and English using Concept Node (CN) definition which is a case frame as extraction rule. The following steps are performed: design conceptual model by terminology exploration, create a CN definition based on syntactic relationship between two major semantic entities analyzed with sentences by a domain expert and implement automatic IE system using CN definition. Main purposes is to investigate whether syntactic and semantic analysis technique using extraction rule (CN) is effective for typical Korean medical text in mixed two different languages.