Menu
Home
Log in / Register
 
Home arrow Computer Science arrow Linked Open Data
< Prev   CONTENTS   Next >

2 Vocabulary Mapping

Linked Data sources use different vocabularies to describe the same type of objects. It is also common practice to mix terms from different widely used vocabularies[1] with proprietary terms. In contrast, Linked Data applications usually expect data to be represented using a consistent target vocabulary. Thus these applications need to translate Web data to their local schema before doing any sophisticated data processing.

To overcome these problems, we have developed the R2R Framework [2][3].

This open source framework consists of a mapping language for expressing term correspondences, best practices on how to publish mappings on the Web, and a Java API for transforming data to a given target vocabulary according to the mappings. We also provide the R2R Graphical User Interface, a web application that allows loading, editing and executing R2R mappings on data sources that are either located in a triple store or in RDF dumps.

As the R2R mapping language is designed for publishing mappings as Linked Data on the Web, mappings are represented as RDF and each mapping is assigned its own dereferenceable URI. The language defines the link type r2r:has Mapping to interlink mappings with RDFS or OWL term definitions and voiD dataset descriptions. The syntax of the R2R mapping language[3] is very similar to the SPARQL query language, which eases the learning curve.

The mapping language covers value transformation for use cases where RDF datasets use different units of measurement, and can handle one-to-many and many-to-one correspondences between vocabulary elements. R2R also offers modifiers to be used for assigning data types and language tags or converting a literal into a URI reference using a pattern. The language provides a set of common string functions, such as concat or split, arithmetic and list functions. See Listing 1 for a mapping example (prefix definition omitted), in which the firstName and lastName properties are concatenated into the name property.

Listing 1. R2R mapping example

The R2R Mapping Engine applies a mapping composition method for selecting and chaining partial mappings from different sources based on a mapping quality assessment heuristic. The assumptions are that mappings provided by vocabulary maintainers and data publishers themselves are likely to be of a higher quality, and that the quality of data translations decreases with the length of the mapping chains.

We evaluated the R2R Mapping Language by formulating mappings between DBpedia and 11 data sources that are interlinked with DBpedia, see [3] for further details. The language proved to be expressive enough in this experiment to represent all mappings that were required. The experiment also showed that far more expressivity is required to properly translate data to a target schema than currently provided by standard terms such as owl:equivalentClass, owl:equivalentProperty or rdfs:subClassOf, rdfs:subPropertyOf.

  • [1] E.g FOAF for representing data about people – foaf-project.org/
  • [2] wifo5-03.informatik.uni-mannheim.de/bizer/r2r/
  • [3] Full specification at wifo5-03.informatik.uni-mannheim.de/bizer/r2r/spec/
 
Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
 
Subjects
Accounting
Business & Finance
Communication
Computer Science
Economics
Education
Engineering
Environment
Geography
Health
History
Language & Literature
Law
Management
Marketing
Philosophy
Political science
Psychology
Religion
Sociology
Travel