Date of Award
Doctor of Philosophy (PhD)
Harry S. Delugach
Ramazan S. Aygun
Dan M. Rochowiak
Mary E. Weisskopf
Personal construct theory., Knowledge acquisition (Expert systems), Knowledge representation (Information theory), Conceptual structures (Information theory)
This dissertation examines the role of context in data integration. Traditionally, data integration solutions attempt to find semantic correlations between data sources and determine which elements match. These traditional solutions do not ask in what context the candidate data elements match, nor do they make an attempt capture it. Instead, they assume a single implicit "correct" context. This is odd considering how strong an effect context has in how people perceive, process, and make decisions based on the information available to them. The knowledge space framework introduced in this work remedies this shortcoming by introducing a formal way to capture multiple simultaneous, and possibly contradictory, contexts for use in data integration. The work provides a three-layer architecture based on dataspace pay-as-you-go integration philosophy. The work identifies two important types of context that hold sway during data integration: the context of intent that captures information pertaining to the creation of each data source, and the context of use that captures information pertaining to the combining of data sources. To capture the semantics that define a context and the relations between data elements, the knowledge space framework uses a conceptual graph concept system. A sample implementation of the framework is presented and discussed. Knowledge spaces provide a scalable pay-as-you-go solution where understanding can grow and be revised overtime and data sources can be added early and explored later when interest and resources allow.
Sabados, William T., "Knowledge spaces : a context-driven concept system framework for data integration" (2015). Dissertations. 70.