Aligning representations of anatomy using lexical and structural methods.

Abstract

The objective of this experiment is to develop methods for aligning two representations of anatomy (the Foundational Model of Anatomy and GALEN) at the lexical and structural level. The alignment consists of the following four steps: 1)acquiring terms, 2) identifying anchors (i.e., shared concepts) lexically, 3) acquiring explicit and implicit semantic relations, and 4) identifying anchors structurally. 2,353 anchors were identified by lexical methods, of which 91% were supported by structural evidence. No evidence was found for 7.5%of the anchors and 1.5% received negative evidence. The importance of taking advantage of implicit domain knowledge acquired through complementation,augmentation, and inference is discussed.