Structuring Clinical Guidelines through the Recognition of Deontic Operators.

In this paper, we present a novel approach to structure Clinical Guidelines through the automatic recognition of syntactic expressions called deontic operators. We defined a grammar and a set of Finite-State Transition Networks (FSTN) to automatically recognize deontic operators in Clinical Guidelines. We then implemented a dedicated FSTN parser that identifies deontic operators and marks up their occurrences in the document, thus producing a structured version of the Guideline. We evaluated our approach on a corpus (not used to define the grammar) of 5 Clinical Guidelines. As a result, 95.5% of the occurrences of deontic expressions are correctly marked up. The automatic detection of deontic operators can be a useful step to support Clinical Guidelines encoding.