Outer Analysis of Quality
ARMY RESEARCH LAB ADELPHI MD COMPUTATIONAL AND INFORMATION SCIENCES DIRECTORATE
Pagination or Media Count:
We use type signatures to discuss several specific notions of quality in order to showcase how the simple discipline can aid in the inspection and definition of vague and intuitive notions. We model qualities as functions, thereby shifting attention from what they are to what they may be based upon. The ultimate aim being the design of systems that can deliver high quality data for arbitrary definitions of quality. The report includes a short description of algebraic data types. In this report, we analyze some common concepts of data quality. Indirectly, our aim is to remove ambiguities associated with these concepts, with the ultimate goal of designing systems that deliver high quality data. Our immediate aim, however, is to provide an alternate approach to thinking about and communicating notions of information and quality. If we want to specify general, universally applicable definitions of various qualities, we must avoid specifying how to calculate any particular quality relating to specific data and applications. One approach to this is to propose an ontology that captures the essential structure of quality and then somehow decide how any specific metric fits with the ontology. This approach is not undertaken here. We leave this structure open and instead focus on giving form to individual notions without regard to whether or not they are necessarily parts of a specific collection, the qualities. In this report, we present an outer analysis of qualities, a circumscription of each particular notion. Outer analysis is a term associated with category theory. Its connotation is an investigation of mathematical objects by looking at their relationships with other objects, as opposed to inspecting the internal, element-wise structure. In this vein, we describe qualities such as accuracy, etc., as functions and analyze them in terms of their inputs and outputs.
- Theoretical Mathematics