Inspecting reference structures
In the project we develop several mechanisms to check the integrity of data after conversion from legacy systems to RDF formats.
Topics we work on in this product group (see also below):
- Import validation
- Introspection
- Wayback
Import validation
During the import of external structure, first it should be checked if the import meets the minimal technical requirements. In order to be able to convert the import to RDF format, at least all entities and their interrelations must have uniform descriptions.
An entity, or object, can be described semantically by a namespace (a type of web address) and a unique identifier. An RDF triplet, RDF's building block, consists of two objects and their interrelations. Thus, such an RDF triplet can be described using five entities: twice by a namespace and an identifier for both objects, and a description of the relationship.
In practice, other data will be included, resulting in a total of at least six mandatory entities. These can be regarded as six spreadsheet columns. A validation application must then check whether all these columns have been entered correctly.
Introspection
After the technical check, the semantic integrity of the important structure will be validated. For instance, the taxonomic thesaurus used by the natural-heritage sector must be a pure tree structure with just one top node, the root node. In addition, each concept (node) must have a scientific name.
This can be checked through introspection. Introspection involves checking whether the necessary object-property pairs (RDF generally uses the term property for relationship) are present.
Version comparison
Often people working on reference structures will want to create a snapshot of a paticular status of a reference structure. This can be necessary when there is a legal demand for preserving a particular state, but it can also prove a useful tool for evaluating the effect of mutations.
The ability to create and view such snapshots is called wayback in the web industry.
Monday, April 2, 2007