Skip to main content
Fig. 5 | BMC Medical Informatics and Decision Making

Fig. 5

From: Fine-grained information extraction from German transthoracic echocardiography reports

Fig. 5

Simplified processing. From top to bottom: input (English, German), subsection annotations, segment annotations, two layers of concept annotations, concept attachments, example terminology. The segmentation algorithm detects subsections and segments. Ambiguity of attributes like stenosis is resolved according to the recognized structure. Postprocessing operations add implicit states for attributes with boolean nature (“Aorteninsuffizienz”, “Stenose”) and remove values that should be suppressed, for instance, at “Insuffizienz”. Suppression activating values are values where the property “suppress” is not true. The expected output contains one value (moderate mitral valve stenosis) that has been imported to the attribute mitral valve stenosis (id=12) from template T1 (Severity). Their ids are composed to 12:1

Back to article page