Authoring case based training by document data extraction

Background: Modeling is the bottleneck to successful implementation of knowledge management systems. In this paper, we propose an evolutionary approach to modeling based upon word processing documents and we describe the tool Phoenix providing the technical infrastructure. Methods: We applied our approach and software system to authoring of medical case based training systems. So far, authors needed to either hand-code the content (usually as HTML) or to use highly sophisticated authoring systems which require instructions and experience to master the complex systems. With our approach we carry further the ideas Felciano and Dev put into practice in their system Short Rounds [4]. They only presented pre-existing documents as an electronic patient record. Following our approach of evolutionary modeling, authors annotate documents to build fully flavored diagnostic training cases [5]. Results: For our training environment d3web.Train [6, 7], we developed a tool to extract case knowledge from existing documents, usually dismissal records, extending Phoenix to d3web.CaseImporter [8]. Independent authors used this tool to develop training systems e.g. in rheumatology, gastroenterology, and cytology, observing a significant decrease of time for setteling-in (from several month down to 1 hour) and a decrease of time necessary for developing a case (down to 4-6 hours) [9]. Conclusions: This paper describes the general approach and provides an in-depth analysis of the document parsing engine (Phoenix) . To generalize the success of d3web.CaseImporter, we conclude by sketching further http://www.d3webtrain.de Phoenix is available under LGPL open source license from https://sourceforge.net/projects/phoenix-ie/.