Navigating Virtual Information Sources with Know-ME

Abstract

In many application domains such as biological sciences, information integration faces a challenge usually not observed in simpler applications. Here, the tobe- integrated information sources come from very different sub-specialties (e.g., anatomy and behavioral neuroscience) and have widely diverse schema, often with little or no overlap in attributes. Yet, they can be conceptually integrated because they refer to different aspects of the same physical objects or phenomena. We have proposed model-based mediation (MBM) as an information integration paradigm where information sources with hard-to-correlate schemas may be integrated using auxiliary expert knowledge to hold together widely different data schemas. The expert knowledge is captured in a graph structure called the Knowledge Map. In MBM, we extend the global-as-view architecture by lifting exported source data to conceptual models (CMs) that represent more source specific knowledge than a logical schema. The mediator’s IVDs are defined in terms of source CMs and make use of a semantically richer model involving class hierarchies, complex object structure, and rule-defined semantic integrity constraints. Additionally, sources specify object contexts, i.e., formulas that relate a source’s conceptual schema with the global domain knowledge maintained at the mediator. In this paper, we introduce a tool called Knowledge Map Explorer (Know-ME) for a user to explore both the domain knowledge, and all data sources that have been integrated using it.