Towards a theoretical foundation for the harmonization of linked data

Enrico Daga Knowledge Media Institute (KMi) - The Open University; Semantic Technology Laboratory (STLab), Institute of Cognitive Sciences and Technologies (ISTC) - Italian National Research Council (CNR)

Doctoral Symposium, International Semantic Web Conference, 2012 (Submitted)

[edit] List of types

List of types in hyerarchy:

Dataset : the container of a triple collection
- TempDataset : a temporary container
- SourceDataset : a triple collection accessible remotely (read-only)
- OutputDataset : the container of the output (the harmonized dataset)
rdfs:Resource
- IRI
- rdfs:Literal
Graph : an RDF triple collection
- Slice : a dataset portion, described as a graph template
  - StatementTemplate : must define the predicate with the 'rdf:predicate' property and may specify also rdf:subject and rdf:object or object datatype or lang.
    - rdf:Statement : a statement template with specified subject, predicate and object
  - Path : an rdf:List of statement templates such that each object is the subject of the subsequent. A Path have a length (the number of steps as StatementTemplates) and a size (the amount of instances of this path).
    - Chain : A Chain is a Path which define subject, predicate and object on any of its StatementTemplate, which are then rdf:Statements. Its size is always 1.
  - Frame : A Frame is composed of a Class and a set of Predicates.

All instances of Class in the dataset MUST have all the predicates of the Frame

[edit] More about graph Slices

A slice is a graph template which selects a subgraph.

Example: a Statement Template:

[ ] predicate [ ]

evaluates to the subgraph matching all triples with the specified predicate. A statement template may also indicate the subject, the object or the object datatype or the object lang.

Other statement templates:

subject predicate [ ]

[ ] predicate object

[ ] predicate [ ]@en

[ ] predicate [ ]^^xsd:string

Example: a Path:

[ ] predicate1 [ ] / [ ] predicate2 [ ] / [ ] predicate3 [ ]

evaluates in the triple collection having all the three predicates in chain.

An Statement Template can include conditions (within [ ] ), for example:

[rdf:type = owl:Class] predicate [ ]

would select all triples with subjects with rdf:type owl:Class

[edit] Conditions

Conditions are evaluated to be TRUE or FALSE with respect to the node to which they apply.

[ condition ] property [ condition ]

Conditions are admitted on subject or object of a statement template.
When conditions are put before the property, they apply to the subject, when after, to the object or value.
Conditions can be put in any level of a path.
Conditions can nest conditions.
Conditions admit some functions.

[edit] Functions in conditions

Aggregation test functions

Aggregation test functions can be used in any CONDITION, and have as input the value(s) of the selected property.

Operator	Parameters	Returns	Description
MAX	predicate IRI [datatype IRI]	BOOLEAN	Evaluates as TRUE the node with the maximum value from the set of values.
MIN	predicate IRI [datatype IRI]	BOOLEAN	Evaluates as TRUE the node with the minimum value from the set of values.

Node test functions

Node test functions can be used in any CONDITION, and have as input the currently selected node.

Operator	Parameters	Returns	Description
EQUALS	predicate IRI to [ IRI \| LITERAL ] policy STRICT \| LAZY	BOOLEAN	If the value of the property is the given IRI or LITERAL. With policy STRICT, also datatype or lang must match. Default policy is LAZY.
RANGE	predicate IRI min LITERAL max LITERAL	BOOLEAN	If the selected literal is in the range given. The min and max parameters are CASTed to the datatype of value. If value does not define a datatype or the datatype is not numeric^[1], then xsd:integer is used.
LANG	predicate IRI literal LITERAL lang LANG	BOOLEAN	If the given literal is tagged with lang lang

[edit] Sanitized Statement Template

A sanitized Statement Template results by deleting all conditions from a Statement Template. A Statement Template without conditions is then equal to its sanitized version.