The Web of Data is fed by \tripli ers", tools able to transform content (often databases) to linked data. Tripli ers implement di erent methods and typically are based on bulk recipes which allow for no or limited customization of the process. Furthermore, their consumption or refactoring is often di cult due to mismatches between the semantics embedded in original structures, and the RDF or OWL semantics obtained thorugh the recipes. Semion is a method and a tool for customizing and expliciting the semantics of data reengineering and refactoring.
Commonly accepted solutions for tranforming non-RDF data
sources into RDF are based on ad hoc semantics-driven
approaches, that make implicit assumptions on the domain
semantics of the non-RDF data source (e.g. a relational
database is trasformed mapping a table into an rdfs:Class, a
table column into an rdf:Property and a table record into an
instance of the speci c RDF table class). The tool described
here, Semion, implements a method that rstly makes no
semantic assumption at the domain level, and just transforms
the data source into RDF triples driven by an OWL
description of the data source structure (a source meta-model),
which can be de ned and customized by the user. Secondly,
the RDF triples can be modelled by aligning them to any
additional RDF or OWL ontology, which acts as either a
metalevel \mediator" to the required semantics (e.g. SKOS [
The method implemented in Semion is based on an
approach that substantially divides the reengineering process
from the modelling one. The reengineering process performs
the semantic lifting just extracting RDF triples driven by the
OWL description of the structure of the datasource provided
as input. On the other hand, the modelling process allows
to introduce semantics in the extracted data set, by using
a semiotic-cognitive approach based on the Linguistic-Meta
Model (LMM) [
The method described in the previous section is implemented in Semion. Currently the tool is still a prototype and has been tested only for transforming relational databases, but it was designed to perform the transormation of any kind of data source. The gure 2 shows the reengineering interface of the Semion tool. It helps the user to de ne the schema of the database structure that is described by using the meta-model provided for the structure of the database itself. Both because the database could be large and because the user could not know exactly how the database was designed, the tool provides a wizard interface that automatically extracts the RDF of a database's schema. Once the RDF of the database's schema is available, the interface allows the user to transorm the data from the database to the RDF format. Before performing data extraction from the database it is also possible to correct possible issues derived from a bad design or a bad mantainance of the database. In fact, the tool provides functionalities to set in the RDF of the database's schema primary and foreing keys and eventually related relations. The refactoring interface allows the user to align the dataset to speci c ontologies for adding semantics to data. The alignment ontologies can by chosen following the method that Semion implements i.e. rst the dataset is aligned to the LMM framework, then to an ontology that contains the distinctions of the formal semantics and nally to an ontology that contains the logics. Semion performs ontology alignments through SPARQL CONSTRUCT, that are obtained from the rules written in a human-readable syntax (see gure 3), that are based on the form:
antecedent ! consequent
Using this syntax, a rule e.g. asserting that being an
instance of class Table in dataset meta-model implies to be a
Concept of DOLCE [
dbs : T able(?x) ! DU L : Concept(?x) This rule will be interpreted as the SPARQL query: CONSTRUCT f ?x rdf:type DUL:Concept. g WHERE f ?x rdf:type dbs:Table. g With the same syntax can be written, through the Semion tool, rules for transforming LMM to the FormalSemantics vocabulary. The rules could be IOLite:FormalExpression(?x) ! FormalSemantics:Query(?x) DUL:Relation(?x) ! FormalSemantics:Class(?x) Rules for aligning the FormalSemantics vocabulary to OWL can be written as the following FormalSemantics:isSubsumedBy(?x, ?y) ! rdfs:subClassOf(?x, ?y) FormalSemantics:Class(?x) ! owl:class(?x) The Semion tool can be downloaded from the following URL http://stlab.istc.cnr.it/software/semion/tool