The largest (and often most valuable) sources of information frequently have their own schema. The schema may be as simple as the attribute names in key-value pairs or row or heading columns in a spreadsheet, to formally written specifications and definitions. Sometimes they may just be organized headings in a table of contents.
Whatever the formalism of the schema, the semantics of these external forms need to be mapped to the existing structures within KBpedia in order to interoperate with the new source data.
Proper mapping means that ambiguous matches are reconciled and the predicates used for the asserted mappings capture the semantics of the external schema. Sometimes concepts in the external schema may be an exact match to what already exists in KBpedia; sometimes they are concepts that are broader or narrower than existing concepts in KBpedia; and sometimes they are only somewhat related and not part of a true subsumption relationship.
Cognonto applies a variety of tools and techniques in its mapping services in order to generate initial candidates. Ultimately, all candidate mappings are manually inspected for accuracy and subjected to various logic, consistency and coherency tests. Bad mappings are as deleterious as bad data. Fortunately, the tooling and high degree of automation in Cognonto's mappings suite means expert manual attention is directed at the truly problematic mappings, leading to acceptable quality costs. Clients may also inspect the mappings and their rationales to conduct their own vettings.
Cognonto's mapping services are often pursued in conjunction with our data preparation and staging services.