- From: Chimezie Ogbuji <ogbujic@bio.ri.ccf.org>
- Date: Thu, 17 Aug 2006 15:12:53 -0400 (EDT)
- To: Dan Connolly <connolly@w3.org>
- cc: public-grddl-wg <public-grddl-wg@w3.org>
On Thu, 17 Aug 2006, Dan Connolly wrote: > > Chime, > > You mentioned clinical research data in our 1st meeting. > > I suspect that's an interesting XML Schema (or plain XML?) use case. > > Do you have any details you can share? Yes, absolutely. Actually this is the primary motivation for joining the GRDDL WG. We have been developing a clinical research data management system which uses XML as the 'main' representation format (organized around a patient record), edits the XML remotely (on a variety of devices) via XForms, submits the XML document (via HTTP PUT to a unique URI for each such record) to a server which (as part of the content management services) transforms the XML to an RDF equivalent graph for persistence. Ofcouse, the expense of dual representation is space, but the primary value is being able to query both as XML and as RDF (the latter being more amenable for investigative question that rely on alot more interpretation than a structured format such as XML will provide). A GRDDL approach could eleviate this expense by allowing a patient record (or any XML-based collection of clinical research data) to be queried semantically (via SPARQL) 'on demand' by associating a GRDDL profile to the specific patient record XML vocabulary. Imagine a fellow assigned to determine a search criteria to identify a patient population for a particular study. He might have a set of classifications specific to the study he could express as logical rules (N3 rules). Then, he could write a client (that understood GRDDL) that speculatively picked a few patient records at random from a remote server (as XML documents) each of which would be associated (by GRDDL profile) to a transform to extract the clinical data as RDF (expressed in a universally supported vocabulary for CPR - such as the HL7 OWL ontology that Helen Chen from Agfa has been working on) and ask his speculative questions of the resulting RDF graph. Or (to take the scenario a step further), apply the study specific rules on the resulting RDF to classify the patient data according to his domain of interest (specific diagnoses, pathological observations, etc..) Chimezie Ogbuji Lead Systems Analyst Thoracic and Cardiovascular Surgery Cleveland Clinic Foundation 9500 Euclid Avenue/ W26 Cleveland, Ohio 44195 Office: (216)444-8593 ogbujic@ccf.org
Received on Thursday, 17 August 2006 19:13:02 UTC