CSV To RDF Converter v3

Change log

What was done

This version is essentially the same than the v2 version except that:

• Only the relevant cases are managed
  • The column name is now always the predicate (because other situations make no sense).
  • The column entry in the configuration file is still there and will have to be removed in v4.
• The object and suject types are all subClassOf rdfs:Class and the predicates types are all subPropertyOf rdf:Property.
• The rdfs:domain and rdfs:range were implemented for all the predicates which enables SPARQL inference.

The v4 will change some stuff:

• Removal of the column 'role' in the main configuration file, because the column type should always be the predicate.
• Timestamp of the objects and subjects with the date of processing of the file, understood as the date the data were included into a semantic database (like Jena or Allegro).
• The predicates will be versioned with a super type. The versioning system will use URI suffixes.
• The documentation will be complete and the reasons of the design choices will be explained in more details.

Principles

Vocabulary:

• A basic triple is noted domain1:subject domain2:predicate domain3:object .
• We will name "data" any subject of object in a hierarchy that leads to rdfs:Class and "relation" all entities leading to rdf:Property
• A cell value is an instance of a class, noted rdf:type or a.
  • It can be preprocessed before the triple is generated.
• A class type is instance of refs:Class and often located in a type hierarchy.
• A property type is instance of rdf:Property and often located in a type hierarchy.

Principle #1

When data exist, they exist forever, independently from their version or applicability.

Principle #2 (not implemented in v3)

Data are stamped with their creation date, when they enter the semantic database (or with another significant date) like domain1:data global:created domain3:"2023-12-08"^^xsd:date.

Principle #3 (not implemented in v3)

Versions are reserved for a specific namespace of relations, in relation with an absolute namespace for long term relations.

Let's take an example:

• domain1:Part_12 appdomain_v01:applicable_to domain1:Variant_G .
• domain1:Part_12 appdomain_v02:applicable_to domain1:Variant_G .
• domain1:Part_12 appdomain_v02:applicable_to domain1:Variant_H .
• appdomain_v01:applicable_to rdfs:subPropertyOf applicability:applicable_to.
• appdomain_v02:applicable_to rdfs:subPropertyOf applicability:applicable_to.

We can track the versions at each new version of data, and we can make SPARQL reasoning work because their father property is absolute.

See also

The README file of the version 2 of this code in the csv2rdf2 folder.