Article Preview
Top1. Introduction
Software domain modeling is a process where individuals seek to conceptualize an application in order to arrive at a solution that meets all of the application’s domain needs and requirements. Conceptualization can be defined as a structure {D, R}, where D is the domain (scope of the application) and R is the set of relevant relations (functionality and interactions of the application) in the D (Gruber, 2005). Currently, one popular domain conceptualization approach is ontologies where the representation of knowledge occurs via the definition of concepts that can be related to one another for sharing and reuse of that knowledge. In practice, W3C supports two major ontology development frameworks, the Resource Description Framework (RDF) (Powers, 2003) and the Web Ontology Language (OWL) (Allemang & Hendler, 2011), where the latter is built on top of the former. Additionally, a wide range of knowledge representational frameworks (e.g., KIF (Genesereth, 1991), KL-ONE (Brachman & Schmolze, 1985), DAML+OIL (Horrocks, 2002), etc.) are available. When reviewing the usage of these frameworks in real applications, the overriding theme is realizing a specific instance-level solution for a particular system, rather than designing a solution that can be reused by multiple similar applications within the same general domain. For example, in health information technology (HIT), multiple systems such as AllScripts (AllScripts, 2012), Centricity (GE Centricity, 2012), MS Health Vault (Microsoft Health Vault., 2013), etc., are developed with their own specific (and different) medical ontologies despite the fact that each one is storing and managing patient data. Any attempt to integrate data from across multiple EHRs, referred to as health information exchange (HIE) requires integrating their different medical ontologies which at best is a semi-automated and arduous task. While many of these HIT systems leverage existing standard medical terminologies (e.g., LONIC (LOINC, 2001), UMLS (Bodenreider, 2004), ICD (ICD, 2009), etc.), there is no uniform attempt to design an ontology domain model for medical knowledge that is general purpose and reusable in multiple contexts. For example, one HIT system may organize medical knowledge in an ontology by disease, symptoms (for each disease), diagnoses (for each disease), and treatments (for each disease), while another HIT system might invert and target this information from a symptom to diagnosis to disease to treatment basis. Attempting to reconcile this medical knowledge in two or more different ontologies is an arborous, time-consuming, and at best semi-automated task, making HIE difficult to achieve.