Describing RDF Resource Links with the Vocabulary of Links (VoL)

30 July 2013

Latest version:

Authors:


Abstract

The Vocabulary of Links (VoL) is an RDF Schema vocabulary for expressing metadata about links between resources & entities within and across RDF datasets. While there do exist limited ways to express links, for instance with RDFS/OWL expressivity such as owl:sameAs or rdfs:seeAlso or SKOS, up to now, there is no vocabulary at hand to further describe the semantics of a link. This is of particular importance for links obtained with automated entity linking techniques, where links are usually associated with some form of probability (such as a connectivity score) instead of simple links where the decision of a connection is binary. In addition, often some of ranking has to be expressed to, for instance, rank the relevance of a particular link for a given entity or dataset (e.g. the association of a category with a particulary RDF resource or dataset).  VoL provides a vocabulary to describe a link with respect to obtained scores and utilised methods for obtaining the links.While we specifically designed it to be used together with VoID dataset descriptions, as part of VoID linksets, it is intended for general-purpose description of links between resources within and across distinct RDF datasets.

Status of this Document

This section describes the status of this document at the time of its publication.

A first version of this document was developed and published by the authors, starting in 2013. This is an extended and improved version, based on community feedback received since the original publication.

Table of Contents

     


1. Introduction

While there do exist limited ways to express links, for instance with RDFS/OWL expressivity such as owl:sameAs or rdfs:seeAlso or SKOS, up to now, there is no vocabulary at hand to further describe the semantics of a link. This is of particular importance for links obtained with automated entity linking techniques, where links are usually associated with some form of probability (such as a connectivity score) instead of simple links where the decision of a connection is binary. In addition, often some of ranking has to be expressed to, for instance, rank the relevance of a particular link for a given entity or dataset (e.g. the association of a category with a particulary RDF resource or dataset).

The Vocabulary of Links (VoL) is an RDF Schema vocabulary for expressing metadata about links between resources & entities within and across RDF datasets. VoL provides a vocabulary to describe a link with respect to obtained scores and utilised methods for obtaining the links.While we specifically designed it to be used together with VoID dataset descriptions, i.e. as part of VoID linksets, it is intended for general-purpose description of links between resources within and across distinct RDF datasets.

VoL covers the following main concepts:

As such, we deem the vocabulary particularly useful for the wealth of libraries, tools, services and approaches related to entity linking - where up to now, no vocabulary existed for sharing and expressing the resulting data. In addition, VoL can be applied to express relationships between RDF subgraphs or otherwise clustered sets of RDF statements or resources.

Deployment and discovery of VoID descriptions is discussed as well.

1.1 Scope

This document is aimed at dataset providers who plan to further describe automatically obtained links, their semantics as well as the obtained connectivity measures and used methods.

Readers of this document should be familiar with the core concepts of RDF [RDF-PRIMER] and RDF Schema [RDFS]. Some knowledge of widely-used vocabularies (Dublin Core [DC], Friend of a Friend [FOAF], Vocabulary of Interlinked Datasets [VoID]) is also assumed.

2. Overview of Classes and Properties

The Vocabulary of Links (VoL) is concerned with metadata about RDF Links and Relations. It is an RDF Schema vocabulary that provides terms and patterns for describing links between RDF resources (statements, datasets, entities). It is intended to complement existing RDF expressivity for describing links with additional semantics and less strict notions of links associated with some form of probability or connectivity. 


Classes


vol:Link
URI: http://purl.org/vol/ns#Link
Documentation: Section 3. Describing a Link
Label: Link
Disjoint classes: vol:LinkingMethod
Description:
A vol:Link describes the relation between two RDF resources (associated with the Link via the vol:linksResource property). In cases where a Link is associated with a dataset (for instance a void:Dataset) and is part of a void:Linkset, e.g. to describe the relevance of a dbpedia:Category for the entire dataset, only one resource needs to be
vol:LinkingMethod
URI:
http://purl.org/vol/ns#LinkingMethod
Definition:
Section 4. Describing a Linking Method
Label:
LinkingMethod
Disjoint classes:
vol:Link

Properties


vol:linksResource
URI: http://purl.org/vol/ns#linksResource
Documentation: Section 3. Describing a Link
Label: links Resource
Super-properties: -
Description:
This property is used to refer to the RDF resources linked by the vol:Link instance. Should be instantiated maximum 2 times per Link instance. If a link is part of a void:Linkset and is associated with the entire void:Dataset, only one vol:linksResource property should be instantiated.
Domain:
http://purl.org/vol/ns#Link
Range:
http://www.w3.org/2000/01/rdf-schema#Resource
vol:hasLink
URI: http://purl.org/vol/ns#hasLink
Documentation: Section 3. Describing a Link
Label: has Link
Super-properties: -
Description:
The link between two RDF resources. In our use case for VoID, a Link is always part of a void:Linkset and links a specific DBpedia resource to a void:Linkset and in turn, void:Dataset
Domain:
http://www.w3.org/2000/01/rdf-schema#Resource
Range:
http://purl.org/vol/ns#Link
vol:hasType
URI: http://purl.org/vol/ns#hasType
Documentation: Section 3. Describing a Link
Label: has Type
Super-properties: -
Description:
The type of the link, for instance, an owl:sameAs link or other forms of relatedness.
Domain:
http://purl.org/vol/ns#Link
Range:
http://www.w3.org/2001/XMLSchema#string
vol:derivedFrom
URI: http://purl.org/vol/ns#derivedFrom
Documentation: Section 3. Describing a Link
Label: derived From
Super-properties:
Description:
Points to one or more resources which suggested the link. Note: this property is distinct from vol:linksResource as it does not specify the resources actually linked by the link but the related resources which might have suggested the link (for instance, resources/nodes within the RDF graph which link the linked resources).
Domain:
http://purl.org/vol/ns#Link
Range:
http://www.w3.org/2000/01/rdf-schema#Resource
vol:hasScore
URI: http://purl.org/vol/ns#hasScore
Documentation: Section 3. Describing a Link
Label: has Score
Super-properties: -
Description:
This property describes any measure computed for detecting the link. As such it describes the probability and level of connectivity of a given link. While entity linking methods usually provide such a measure (instead of a plain preference), it is crucial to describe the cohesion between two resources and to allow data consumers to select links of their required level of connectivity.
Domain:
http://purl.org/vol/ns#Link
Range:
http://www.w3.org/2001/XMLSchema#double
vol:usedLinkingMethod
URI: http://purl.org/vol/ns#usedLinkingMethod
Documentation: Section 3. Describing a Link
Section 4. Describing a Linking Method
Label: Used Linking Method
Super-properties: -
Description:
A property to link a vol:Link instance to the LinkingMethod instance describing the method used to compute the score.
Domain:
http://purl.org/vol/ns#Link
Range:
http://purl.org/vol/ns#LinkingMethod

3. Describing a Link

The central concept of VoL is vol:Link which describes the actual link between two RDF resoruces. The following example describes a vol:Link and associates it with a specific vol:LinkingMethod instance:
<vol:Link rdf:about="http://data.example.org/link/examplelink123">
<rdfs:label>Link instance 339922012839</rdfs:label>
<rdfs:comment>A vol:Link instance describing a link between a DBpedia entity and a resource in the LAK Dataset (see http://datahub.io/dataset/lak-dataset).</rdfs:comment>
<vol:linksResource rdf:resource="http://data.linkededucation.org/resource/lak/conference/edm2012/paper/52" />
<vol:linksResource rdf:resource="http://dbpedia.org/resource/Data_mining"/>
<vol:hasScore>0.56</vol:hasScore>
<vol:hasType>Semantic connectivity</vol:hasType>
<vol:usedLinkingMethod rdf:resource="http://data.example.org/linkingmethod/SCS" />
</vol:Link>

Please note that a vol:Link can be associated with any kind of RDF resources and can be associated with a specific linkset (such as a void:Linkset, see Vocabulary of Interlinked Datasets [VoID] for details). In cases where a link is part of a void:Linkset and is associated with an entire void:Dataset, only one resource needs to be specified via the vol:linksResource property, since the resource is linked to the entire dataset. This applies for instance in cases where a  DBpedia category is associated with an entire dataset as in the case of  ..... [LOD-TOPICS]

4. Describing a Linking Method

A vol:LinkingMethod instance is used to describe and identify further the method or tool (such as [SILK]) used to obtain the link. To this end, it can link and refer to any related library, documentation, academic publication or API and is mainly used to unambiguously identify the method used to obtain the link.  For the description, we suggest to use established vocabularies such as (Dublin Core [DC]) or  simple <rdfs:seeAlso> references.
<vol:LinkingMethod rdf:about="http://data.example.org/linkingmethod/SCS">
<rdfs:label>Linking Method instance Semantic Connectivity Score</rdfs:label>
<rdfs:comment>A vol:LinkingMethod instance describing the Semantic Connectivity Score approach introduced by Nunes et al.</rdfs:comment>
<rdfs:seeAlso>http://eswc-conferences.org/sites/default/files/papers2013/nunes.pdf<rdfs:seeAlso>
<dcterms:creator rdf:resource="http://dblp.l3s.de/d2r/resource/authors/Bernardo_Pereira_Nunes"/>
</vol:LinkingMethod>
It is encouraged to provide as much information as possible, to enable the unambiguous identification of the used methods.

5. Deploying and Quering VoL Data - Examples

Tbd.

6. Acknowledgements

We would like to thank everyone who helped to shape the design of VoL, provided use cases and ideas. The work has partly been supported by the following projects:

References

[DBPEDIA]
DBpedia project homepage, http://dbpedia.org/
[DC]
DCMI Metadata Terms, Dublin Core Metadata Initiative, 11 October 2010. http://dublincore.org/documents/2010/10/11/dcmi-terms/
[FOAF]
FOAF Vocabulary Specification, D. Brickley, L. Miller, 9 August 2010. http://xmlns.com/foaf/spec/20100809.html
[LOD-TOPICS]
Geonames project homepage, http://www.geonames.org/
[RDF-CONCEPTS]
Resource Description Framework (RDF): Concepts and Abstract Syntax, G. Klyne, J. Carroll, W3C Recommendation 10 February 2004. http://www.w3.org/TR/2004/REC-rdf-concepts-20040210/
[RDF-PRIMER]
RDF Primer, F. Manola, E. Miller, W3C Recommendation 10 February 2004. http://www.w3.org/TR/2004/REC-rdf-primer-20040210/
[RDFS]
RDF Vocabulary Description Language 1.0: RDF Schema, D. Brickley, R.V. Guha, W3C Recommendation 10 February 2004. http://www.w3.org/TR/2004/REC-rdf-schema-20040210/
[SILK]
Silk - A Link Discovery Framework for the Web of Data, R. Isele, A. Jentzsch, C. Bizer, J. Volz. http://www4.wiwiss.fu-berlin.de/bizer/silk/
[SKOS]
SKOS Simple Knowledge Organization System Primer, A. Isaac, E. Summers, W3C Working Group Note 18 August 2009. http://www.w3.org/TR/2009/NOTE-skos-primer-20090818/
[SPARQL]
SPARQL Query Language for RDF, E. Prud'hommeaux, A. Seaborne, W3C Recommendation 15 January 2008. http://www.w3.org/TR/2008/REC-rdf-sparql-query-20080115/#rdfDataset
[TURTLE]
Turtle - Terse RDF Triple Language, D. Beckett, T. Berners-Lee, W3C Team Submission 14 January 2008. http://www.w3.org/TeamSubmission/2008/SUBM-turtle-20080114/
[VOID-VOC]
VoID Vocabulary, K. Alexander, R. Cyganiak, M. Hausenblas, J. Zhao. http://rdfs.org/ns/void