G-DEE incorporates text processing functions to support encoding through a first level of automatic structuring. It also supports various document analysis features, some dealing with specific text display, and others triggering text analysis functions to extract knowledge or information (e.g. in a rule-based format, GEM). The analyzed expressions are structured using specific mark-ups, which supports hypermedia presentation of the recommendations.
Central to our approach is the fact that automatic content processing should support document structuring by generating mark-ups each time specific linguistic markers are recognized (for instance linguistic markers signaling recommendations). Recommendations are the essence knowledge of the clinical guidelines and are taken into account to elaborate knowledge bases and decision support systems. Clinical guidelines are naturally structured through the occurrence of specific linguistic expressions, known as “deontic operators”. These operators manifest themselves through such verbs as “pouvoir” (“to be allowed to or may”), “devoir” (“should or ought to”), “interdire” (“to forbid”) and correspond to traditional deontic modalities: permission, obligation and prohibition. We have developed an ad hoc parsing technology based on Finite-State Automata which parses the document and generates mark-ups corresponding to deontic operators and their scopes. The marked-up document can subsequently be the object of various XSL-based transformations.

The interface supports the selective processing of text fragments, which are analyzed for deontic operators (interface button 1) and marked-up accordingly (window B). The resulting marking-up can be validated interactively by the user (button 1 of the interface). In addition, G-DEE enables to automatically display contents of specific GEM elements, as well as deontic operators in window C. Window D displays decision rules automatically derived from the marked-up text, which can be used for knowledge extraction or analysis of text coherence.

The automatic processing component, which underlies the marking-up process, has been evaluated using two complete clinical guidelines (corresponding to over 300 recommendations). As a result, precision of marker identification varied between 88 and 98% and recall between 81 and 99%.

G-DEE was originally conceived as a research environment but has since been adopted as an experimental tool for assisting the guideline authoring process in the French National Authority for Health (HAS), and has been used to provide an independent analysis of guidelines structure.