=Paper=
{{Paper
|id=None
|storemode=property
|title=Rights declaration in Linked Data
|pdfUrl=https://ceur-ws.org/Vol-1034/RodriguezDoncelEtAl_COLD2013.pdf
|volume=Vol-1034
|dblpUrl=https://dblp.org/rec/conf/semweb/Rodriguez-DoncelGM13
}}
==Rights declaration in Linked Data==
https://ceur-ws.org/Vol-1034/RodriguezDoncelEtAl_COLD2013.pdf
Rights declaration in Linked Data ⋆
Vı́ctor Rodrı́guez-Doncel, Asunción Gómez-Pérez, and Nandana
Mihindukulasooriya
Ontology Engineering Group, Universidad Politécnica de Madrid, Spain
Abstract. Linked Data is not always published with a license. Some-
times a wrong license type is used, like a license for software, or it is not
expressed in a standard, machine readable manner. Yet, Linked Data re-
sources may be subject to intellectual property and database laws, may
contain personal data subject to privacy restrictions or may even contain
important trade secrets. The proper declaration of which rights are held,
waived or licensed is a must for the lawful use of Linked Data at its dif-
ferent granularity levels, from the simple RDF statement to a dataset or
a mapping. After comparing the current practice with the actual needs,
six research questions are posed.
Keywords: Linked Data, licensing, intellectual property rights
1 Introduction
The term Linked Data (LD) is generally defined as a set of best practices for
publishing and connecting structured data on the Web [1], and RDF is the pre-
ferred technology for representing this data. RDF information unit is the triple
(a simple statement with a subject, a property and an object), being a RDF
graph a set of triples, and a RDF dataset a collection of RDF graphs. Triples
provide information about resources (identified by URIs) constituting pieces of
data (or metadata —if the resource is data itself). The resources in a dataset
are often linked to the resources in other datasets, through RDF mappings.
Linked Data is accessible to the public through the HTTP protocol, usually
as RDF dumps in files or in SPARQL endpoints. However, being publicly avail-
able doesn’t entitle the public to do any arbitrary action on the LD resources,
and unless otherwise stated intellectual property (IP) rights and database rights
will be in force if they exist. The most common practice with LD, however, is
waiving some of the rights subject to certain conditions, using public notices
called licenses. If these licenses are generous enough, they are called open li-
censes, and the Linked Data thus licensed is called Linked Open Data (LOD).
The subset of LD that is not licensed as LOD has been termed Linked Closed
Data (LCD) [2], Linking Enterprise Data [3] or simply proprietary data.
⋆
This research is supported by the Spanish Ministry of Science and Innovation through
a Juan de la Cierva postdoctoral fellowship, the LiDER project (FP7-610782) and
the project BabelData (TIN2010-17550).
2 V. Rodrı́guez-Doncel, A. Gómez-Pérez, N. Mihindukulasooriya
The lax and vague terms under which LD has been sometimes published has
sufficed for all purposes, as the data publishers have been usually tolerant with
its improper use not prosecuting it in the courts. However, misusing or disclosing
high value data may suppose a real economic harm for those who have invested
time and money in producing the dataset and it may cause legal trouble to the
breaching user if sued. Further, it may discourage businesses entering the LD
markets as they would fear having similar economic damage themselves. For the
lawful use of Linked Data, a proper rights declaration understandable by humans
and machines alike is a precondition. The following use case illustrates this need,
which is not limited to IP-related laws but also to database laws, privacy laws
and even trade secrecy laws.
Alice, a data engineer, starts working today. She has been given a RDF
dataset with valuable information, with no other indications, and she is
unsure about what she can do. What is the risk of publishing it? Would
she be breaking the IP law or, even worse, disclosing a trade secret? Can
she edit the contents or even change the format of the dataset? What
about distributing or selling the dataset?
The need for rights declaration is not limited to LCD, as lawfully using LOD
also requires the satisfaction of the conditions in the license, nor is a different
problem for either case, as the interplay of LOD and LCD in hybrid business
models is likely to boost them. This paper describes the legal framework for
publishing and consuming Linked Data which must be known by LD engineers
(Section 2 and Section 3). It also makes an overview of the existing vocabularies
for declaring rights and licenses in RDF (Section 4) to follow with an assessment
of the actual use of these licensing terms (Section 5). Finally, the Section 6
opposes the legal requirements for Linked Data-centric business to appear with
the existing of vocabularies and its actual use.
2 Legal Framework
Linked Data can become a high-value asset worthy to be protected. This protec-
tion can be achieved by means of secrecy –disclosing the data only to selected
parties who have possibly paid, for example, as suggested in [4]. But law also
offers a certain protection for data producers. Following the example before, Al-
ice perhaps had received any of the datasets exemplified in Figure 1. The first
contains literary works, the second a trade secret, the third personal data and
the fourth mere data not subject to IP rights. Each case is different, existing
rights associated to different parties (works‘ creator, the company holding the
secret, the persons whose data is in a file), and all of the databases possibly
subject to database rights. This section examines the different cases.
First, if data is not generally known to the public, if data confers some kind
of economic benefit on its holder (and specifically by the fact that it is not
generally known) and if it is subject of reasonable efforts, then it is object of
protection by trade secrecy laws (in some jurisdictions named as confidential
Rights declaration in Linked Data 3
22ºC
760mm
CocaCola
17ºC
Formula -5ºC
Fig. 1. Examples of databases for which IP, trade secret and personal data law apply
information laws). Trade secrecy was included in the TRIPS agreements1 , and
disclosing a trade secret is a prosecuted action everywhere, whose punishment
is even harsher if the secret is of a military nature –if we are to have Military
Linked Data. Disclosing a secret is an act which may be punishable under the
criminal law. Other laws may preclude the communication of datasets if they
express defamations, libels or other forbidden contents. This may have little
interest for the data engineer, who yet should know that some datasets may not
be spread out of a certain circle or even not communicated at all.
Second, data may be qualified to become object of protection by the intel-
lectual property laws. Data, as a representation of a fact or an idea, is not
necessarily the expression of an intellectual endeavour and in principle it does
not get protection per se. But data can also represent an IP work (image, text...),
in which case the IP law applies.
But also, if the selection and arrangement of others’ literary and artistic
works under the form of anthologies, databases etc. is the result of an intellec-
tual creation, the mere collection is also under the umbrella of the intellectual
property law (without prejudice of the rights of the original works’ authors).
This is universally acknowledged and can it be found in the Berne Convention2 .
Intellectual property rights comprise moral rights and exploitation rights.
Moral rights are untransferable and unwaivable in some jurisdictions, and they
include rights as the author being attributed, the work being respected or staying
anonymous. Exploitation rights can be waived, licensed to the public, or traded
in an economic exchange. Thus, the rightsholder of each of the exploitation rights
may change along the time. These rights traditionally include the reproduction
of the work (e.g., making copies), the distribution of copies (e.g., selling, renting
etc.), the public performance, broadcasting or communication to the public and
the transformation (including translation, adaptation etc.). Additionally, the so
called related rights or neighbouring rights concern other categories of owners of
rights different from the authors, namely, performers, producers, broadcasting
organizations etc. A data curator or a dataset translator may also acquire related
rights on the result of their work.
Third, in some jurisdictions, specific database right laws have been de-
clared for the protection of databases which do not qualify to be intellectual
property objects. This is the case of Europe, but not the United States, where
no database right exists. This sui generis rights, as defined in Europe, protects
1
Art. 39 in WTO Agreement on Trade Related Aspects of Int. Property Rights (1994)
2
Art. 2.5 in Berne Convention for the Protection of Literary and Artistic Works (1886)
4 V. Rodrı́guez-Doncel, A. Gómez-Pérez, N. Mihindukulasooriya
the ”qualitatively and/or quantitatively substantial investment in either the ob-
taining, verification or presentation of the contents”3 . Extracting or re-utilizing
the whole or a substantial part of the contents is prohibited unless permission is
given. Naturally, exceptions usually exist for the case of educational purposes,
injunction, public security etc., and in any case, after 15 years the database
enters permanently into the public domain.
The combination of intellectual property rights and database rights (where
applicable), generates a set of possible scenarios depending on (a) if the dataset
contents are IP protectable assets or not (b) if the dataset creator has IP rights,
database rights or none of them. Determining which of the scenarios corresponds
to an actual case is not an evident task, as pointed out in [5]. For example, in
the USA, Canada, Australia or Japan, a dataset with “the best of” a musician
would be regarded as intellectual property object, for it would be a compilation
involving an aesthetic judgment. But a “complete collection of works” of the
same musician wouldn’t, for it would be an automatable task. This collection
would be protected, however, in Europe, if a verification work or any other
similar effort was carried out.
Finally, if personal data is conveyed in the database, data protection laws
have to be considered too4 ). These laws give no rights to the database creators,
but rather impose certain obligations which have to be respected. These obli-
gations include implementing security measures to be taken to physically and
digitally protect the information, generating periodic security reports or keeping
data access logs. In some jurisdictions, different levels of protection exist as a
function of how sensitive the information is. As an example, the law in Spain
defines three levels of confidentiality in a personal data file, ranging from the
most trivial information but attributable to a person, to the most sensitive in-
formation like the sexual or religious preferences. Persons whose information is
contained in a file have the right to access and rectify their records, which in
any case can only be gathered for a declared purposed and can only live for a
limited period of time.
To sum up, these laws ultimately protect the rights of (a) the authors who
have created contents collected in a dataset (b) the dataset creators who have
selected, curated and arranged the registers (c) the individuals whose personal
information is in the dataset (d) third parties damaged if data is disclosed.
3 Rights Declaration for Linked Data
To precise the rights declaration for Linked Data, the following levels should
be independently considered: (a) a single RDF triple, as the simplest unit of
information, (b) RDF graphs or RDF datasets, as collections of data, (c) the
3
Art. 7(1) in Directive 96/9/EC on the legal protection of databases (1996)
4
For example, see the corresponding European Directive 95/46/EC on the protec-
tion of individuals with regard to the processing of personal data and on the free
movement of such data (1995)
Rights declaration in Linked Data 5
RDF links, as mappings play the key role in the added value of Linked Data and
(d) external resources referred by RDF.
Single RDF triples (or a reduced group of them) are not protected by
intellectual property or database laws, which explicitly exclude individual data
from the protection scope –unless a work or a full data collection is contained
in a literal. However, they may be protected as trade secrets or its access by
restricted for other reasons. As with copyrighted material, stamping a top secret
or similar notice is merely informative and no additional protection is conferred
by its addition.
In general, a RDF dataset matches the legal concept of database5 . Its
creator may claim database rights in certain countries, plus intellectual property
rights if the dataset contains works creatively selected and arranged –a claim
difficult to be justified in most of the cases. Database rights do not exist if,
for example, the RDF dataset was only an effortless syntax transformation of
what was included in another database. Other rights may exist over datasets if
they contain trade secrets or personal data whose handling is subject to further
restrictions.
RDF datasets aggregating data from different RDF sources require the spe-
cific authorization from the different dataset owners or the existence of a public
license allowing to do so –in which case, possibly some conditions will have to be
respected. RDF mappings are collections of triples relating resources in two
or more different RDF datasets. Excepting for the case of automatic mappings,
linking vocabularies or resources is a costly effort which almost immediately
qualifies the work as a protectable asset: RDF mappings are a first class citizen
in the Linked Data ecosystem.
Referring an external entity is an action always allowed: even if the resource is
not yours, you can freely comment on it –or link it to your concept. But declar-
ing a mapping (either an added-value mapping or merely re-using mappings
already in the public domain) leads to opening the door to using information
from different RDF resources with possibly a different legal character. The use
of data obtained by following links in RDF mappings may be subject to rights
whose declaration would ease the lawful use of Linked Data. This also applies
to resources referred by RDF subject to protection, although most of users
are possibly aware of this.
Finally, declaring if a Linked Data resource contains personal data or con-
fidential information is merely informative, but it can ease its handling and
strengthen the rightsholder in case of litigation. To sum up, for the lawful use of
Linked Data, which may be created, acquired, transformed and published in a
value chain where several parties intervene, a proper holistic rights declaration
is a must.
5
A database is: a collection of independent works, data or other materials arranged
in a systematic or methodical way and individually accessible, in European Directive
96/9/EC
6 V. Rodrı́guez-Doncel, A. Gómez-Pérez, N. Mihindukulasooriya
4 Linked Data for rights declaration
If the declaration of rights for RDF data is needed, RDF itself can be the vehicle
for its expression. The basic information to be given informs that a Linked Data
resource (a triple, a graph, a dataset, a mapping,. . . ) is subject to certain rights
and if they are kept (e.g. a copyright statement), unconditionally released (e.g.
a waiver notice) or given subject to certain conditions (e.g. licensed).
This allows us identifying three questions: which subjects can be attributed
with rights expressions? Which predicates can be used for rights declaration?
And which licenses can be used in the rights declaration? The rest of the section
describes the existing choices to express these pieces of information.
4.1 Properties for Linked Data rights declaration
The predicate for rights declaration can be taken from Dublin Core (DC), per-
haps the most used vocabulary in Linked Data after the language constructs
(RDF, OWL, etc.): rights is one of the fifteen core properties defined in the
Dublin Core Metadata Element Set6 for use in resource description. Defined as
Information about rights held in and over the resource, it has been generally
used to include descriptions of the copyright information or references to rights
management services. It is present in two different namespaces, usually prefixed
as dc7 and dcterms8 .
This predicate is generic enough as to be used to refer to any of the rights
described in Section 2. Dublin Core specifies two properties refining the rights
property: accessRights (information about who can access the resource or an
indication of its security status) and license (a legal document giving official
permission to do something with the resource). The former may be used to de-
clare that a resource contains personal data (like a phone number), while the
latter has been extensively used to declare the intellectual property license of a re-
source. The Creative Commons property cc:license9 , derived from dc:license,
has also been used to point at a well-known license.
4.2 Subjects of Linked Data rights declaration
The subject of a rights declaration is the piece of information object of the rights,
either a referred resource, an RDF triple, a dataset, or a mapping.
To declare rights of a referred resource, a simple property can be stated
about the resource. The following example attributes a Creative Commons CC-
BY license to an external resource.
@prefix dcterms: .
@prefix ex: .
ex:externalResource1
dcterms:license .
6
http://dublincore.org/documents/dces/
7
http://purl.org/dc/elements/1.1/
8
http://purl.org/dc/terms/
9
http://creativecommons.org/ns#
Rights declaration in Linked Data 7
To declare access restrictions to a RDF triple, a reificated statement can be
attributed with rights declaration. The following example attributes a privacy
statement to a phone number.
@prefix rdf: .
_:x rdf:type rdf:Statement ;
rdf:subject ex:Alice ;
rdf:predicate foaf:phone ;
rdf:object "654321987" ;
dcterms:accessRights "PersonalData".}
Note the use of dcterms:accessRights, which according to Dublin Core can
also be used to give information regarding access or restrictions based on pri-
vacy, security, or other policies. However, on despite of existing a vocabulary for
privacy preferences ontology10 , no common term has been accepted to tag that
a piece of information contains personal data and so a simple literal “Personal-
Data” has been given: an acknowledged URI is missing here.
A complete RDF dataset may be attributed within or outside the dataset
it-self. Within the dataset, the most common practice is to attribute the URI of
the dataset as in the example below.
dcterms:license .
The dataset can also be described in a separate RDF graph, possibly based on
the VoID11 or DCAT12 vocabularies. In this case, the instance of void:Dataset,
dcat:Dataset or even of its parent class dctype:Dataset would be attributed the
corresponding rights declaration.
Finally, RDF mappings can receive the same treatment as RDF datasets,
save that in VoID a dataset subclass is defined: void:LinkSet (a collection of
RDF links between two datasets). This linkset can specify the referred dataset
through the void:target property, which in turn can receive a rights declaration
–for example a public domain license.
http://URI-OF-A-LINKSET> a void:Linkset ; # mapping
void:target .
a void:Dataset . #external dataset
dcterms:license .
4.3 Rights declaration for Linked Data
The rights declaration should convey the information of which rights are held,
waived or licensed. For data licensing, specific data licenses exist and can be
identified by known URIs. It is the case of the Open Data Commons13 (ODC)
licenses, the Creative Commons license CC0 and licenses defined by some gov-
ernments. This makes possible and easy the assignment of a license to a RDF
dataset. The most common data licenses are:
10
http://vocab.deri.ie/ppo#
11
http://www.w3.org/TR/void/
12
http://www.w3.org/TR/vocab-dcat/
13
ttp://opendatacommons.org
8 V. Rodrı́guez-Doncel, A. Gómez-Pérez, N. Mihindukulasooriya
– Public Domain Licenses. They waive all the possible intellectual property
and neighbouring rights (database rights) of the dataset and its contents.
There are two equivalent choices, the ODC-PDDL (Public Domain Dedica-
tion and License) and the CC0 public domain waiver.
– Attribution Licenses. They waive all the possible rights, requiring only the
mere attribution. Example: ODC-By, attribution for data/databases.
– Share-alike Licenses. The rights are also waived requiring that derived or
adapted databases keep the same license. Examples: ODC-ODBL (Open
Database License), or the UK-OGL (UK Open Government License).
Some other licenses famous have been used, like the general Creative Commons
licenses. These pre-defined licenses are also identifiable by URIs, but they are
intended for general works and do not mention the database rights which might
apply in places like Europe. These Creative Commons licenses always require
attribution (BY), and they may require the share-alike (SA) condition, a non-
commercial flag (NC) or the non-derivatives (ND) restriction. “Non-commercial”
means that the work (nor derived versions thereof) can be use for profit, non-
derivatives means that no transformations of the original work can be published.
The combination of these conditions leads to having licenses known as CC-BY
(only attribution), CC-BY-SA (with share alike), etc.
The imprecise use of licenses for datatasets is even more evident when licenses
like the GFDL (GNU Free Documentation License) conceived for documents or
even software licenses are used. Attending to their degree of restrictiveness, a
categorization is shown in Table 1.
Kind of license Licenses Comment
Not specified not specified No license has been specified
Public Domain cc-zero, odc-pddl All the rights have been waived
Attribution cc-by, odc-by Attribution is required
Share alike cc-sa, odc-odbl, uk-ogl, gfdl Copyleft licenses
With restrictions cc-nc, cc-nd, cc-nc-nd, etc. More severe restrictions
Closed all rights reserved, etc. Closed licenses
Other unknown, etc. Not catalogued
Table 1. Classification of licenses by their restrictiveness
More complex ad-hoc licenses can be defined with one of the digital Rights
Expression Languages like ODRL (Open Digital Rights Language) [6] or MPEG-
21 REL [7], although they are XML based and do not intend to imbricate with
the rest of the web of data as proposed in [8]. A new breed of vocabularies,
interconnected and not intended for its use in specific Digital Rights Management
systems is now appearing: vocabularies like LiMO14 , L4LOD15 or ODRS16 , but
14
http://data.opendataday.it/LiMo
15
http://ns.inria.fr/l4lod/v2/l4lod v2.html
16
http://schema.theodi.org/odrs/
Rights declaration in Linked Data 9
so far only the Creative Commons RDF ccREL [9] has been used by the Linked
Data community.
5 Current practice in rights declaration in Linked Data
Quantitatively observing the current practice about rights declaration in Linked
Data is a difficult task as RDF sources are multiple and embracing every piece
of Linked Data in the web is not possible. Yet relevant or extensive parts of it
can be analyzed.
For example, the LOD cloud is important for being a reference of high quality
data, accounting 338 datasets, although biased regarding licensing: they are sup-
posed to be openly licensed. A broader collection of datasets, easily accessible,
is that listed in the CKAN archive (Comprehensive Knowledge Archive Net-
work17 ), a registry of open data and content packages provided by the OKFN,
excelling for its completeness at cataloguing existing datasets. A more selected
collection of sources, periodically compiled and analyzed is that of the DyLDO18 ,
in a framework to monitor Linked Data over the time. It includes datasets from
the LOD cloud and the Billion Triple Dataset challenge19 . Finally, another source
of study may be Sindice20 , a lookup index over resources crawled on the Semantic
Web, which ingests RDF, RDFa and microformats.
Again recalling the study of Section 4, the questions to be answered by the
experimental work can be formulated as Which subjects are actually attributed
with rights expressions? Which predicates are actually used for rights declara-
tion? And which licenses are actually used in the rights declaration?
5.1 Rights declaration for Linked Data in practice
In order to assess the use of licenses, a double test was made: determining which
licenses were in use in the official LOD datasets, and which licenses were in use
in the broader set of the LD datasets in CKAN. The set of LOD datasets could
be obtained by using the REST API of CKAN (the LOD cloud diagram was for-
mally managed through the CKAN repository). CKAN also records information
about the license of each dataset, as declared at registering time. In a similar
manner, the license in general LD datasets in CKAN was queried.
As of May 2013, 1,836 Linked Data datasets21 were registered in the CKAN,
belonging 338 of them to the LOD group. Each of the datasets had one or
more resources (i.e. different data files, SPARQL endpoints etc.) but each of
the datasets was homogenously licensed through the resources. The results of
this observation are shown in Table 2, which has grouped the licenses with the
17
http://datahub.io/
18
http://swse.deri.org/dyldo
19
http://km.aifb.kit.edu/projects/btc-2011/
20
http://www.sindice.com/
21
A dataset was considered to be LD if it had one resource marked with a type con-
taining the following strings: rdf, rdfs, owl, ttl, turtle, nquads, ntriples, nt or sparql.
10 V. Rodrı́guez-Doncel, A. Gómez-Pérez, N. Mihindukulasooriya
criteria of Table 1. This grouping hides, nonetheless, the fact that a 29% of the
licenses were intended for works and not specifically for data.
- All Linked Data datasets in CKAN LOD datasets in CKAN
Kind of license Num. (%) Num. (%)
Not specified 469 (26%) 132 (39%)
Public Domain 291 (16%) 69 (21%)
Attribution 440 (24%) 66 (20%)
Share alike 322 (18%) 35 (10%)
With restrictions 143 (8%) 5 (2%)
Closed 43 (2%) 16 (5%)
Other 128 (7%) 3 (1%)
Total 1,836 (100%) 338 (100%)
Table 2. Licensing of Linked Data datasets
Disregarding the object where a license has been applied (a RDF dataset,
external resources, etc.), an SPARQL query can be made to observe which kind
of licenses are used in extensive pieces of Linked Data. Having made this query
in Sindice, public domain and attribution licenses again gathered the largest
percent of all the licenses: 63% against the 53% used for CKAN datasets. Share-
alike licenses accounted for a 27%, against the 24% in CKAN datasets and
licenses with restrictions (non-commercial, no derivatives) were 6% in Sindice
against the 11% in CKAN datasets.
5.2 Properties for Linked Data rights declaration in practice
The goal of this observation is to assess which RDF elements are most used
to specify a license. To achieve this, different SPARQL queries were made on
Sindice, inquiring for each of the most common elements used for licensing.
The results, shown in Table 3 shows the dc:rights as the champion. Yet, this
element is used about one order of magnitude less than the dc:title element.
The queries for Dublin Core included both namespaces as described in Section
4.1.
Vocabulary Element Usage Usage (%)
Dublin Core rights 5,905,519 59%
XHTML license 3,825,939 38%
CreativeCommons license 263,805 3%
Dublin Core license 32,922 neglectable
Table 3. Relative use of licensing terms in Linked Data
Rights declaration in Linked Data 11
More licensing elements proposed in other vocabularies were also tested, but
their presence in Sindice was neglectable if not zero. These vocabularies included
properties as the DOAP22 doap:license, the PREMIS23 premis:licenseTerms,
the OMV24 omv:hasLicense, the Music Ontology25 mo:License, the VAEM26
vaem:hasLicenseType, or more sofisticated classes in Dublin Core as dcterms:
RightsStatement or dcterms:LicenseDocument.
5.3 Subjects of Linked Data rights declaration in practice
In the previous sections, the RDF triple, the RDF dataset and the RDF map-
ping had been identified as the key ingredients of Linked Data. In the following
experiment, Sindice was queried to learn how often a licensing property had been
applied to rdf:Statements, void:Datasets and void:Linksets.
The experiment revealed that rights declaration had been expressed un-
evenly for these levels. Sindice included 48,968 reificated statemenets, of which
13,505 had rights information, but coming from exclusively a single dataset.
RDF Datasets declared with void:Dataset accounted a total number of 4,549,
of which 92 used a Dublin Core rights and 26 a Dublin Core license. Finally,
none of the 1,163 mappings declared with void:Linkset and found by Sindice
had rights information.
6 Conclusions
An ecosystem of entities (public bodies, academic institutions, enterprises, etc.)
producing, transforming and consuming Linked (Open) Data in a marketplace
is now starting to bloom [11], and it will presumable flourish more healthy if
enough guarantees exist for all the parties in the value chain and their rights.
However, the mismatch between the needs described in Sections 2 and 3 and the
practices observed in Section 5 lead to formulating a series of pending challenges:
1. Vocabularies for declaring rights information exist, but are not complete.
Terms like the Dublin Core license and rights have gained popularity (as
shown in Section 4.1), but they fail to be precise. While it is vaguely assumed
that the rights or licenses are IP related, other legal concerns as the privacy
statements or confidentiality stamps (Section 2) are ignored.
2. Vocabularies for licensing terms exist, but they need further development.
Some existing licenses are now well known and widely accepted. But specific
terms of use are still referenced in natural text. The new vocabularies for
licensing LD which are now sprouting should become more mature, better
documented and accompanied of easy tools for producing rights expressions.
22
http://usefulinc.com/ns/doap
23
http://multimedialab.elis.ugent.be/users/samcoppe/ontologies/Premis/premis.owl
24
http://omv.ontoware.org/2005/05/ontology
25
http://musicontology.com/
26
http://www.linkedmodel.org/schema/vaem
12 V. Rodrı́guez-Doncel, A. Gómez-Pérez, N. Mihindukulasooriya
3. Licensing information for Linked Data should be done at different granularity
levels. The description of RDF Datasets and mappings with VoID is adequate
(see Section 4.2) but its use should be more spread to have a unified way
of expressing rights declaration at different granularity levels, not forgetting
RDF statements and RDF graphs.
4. Many pieces of existing Linked Data lack a proper rights declaration or are
incorrectly licensed. A high percentage of the datasets in the LOD cloud
(39%), champions of the LD, have no rights declaration specified, and 30% of
the licensed datasets are using licenses for IP works unaccurate for database
rights (see Section 5.1).
5. Mappings have not received proper attention. The fact that no mapping was
licensed (Section 5.3) in the linksets with a VoID description reveals that LD
creators may not have understood the value of a well-done mapping, and the
importance of properly attributing rights for its use by third parties.
6. There are no tools granting trust when aggregating data form different sources.
In order to easily build up richer datasets, reliable provenance information
is needed, along with the precise knowledge of which licenses are possible
when aggregating data from differently licensed sources, as described in [10].
As these leaks are not intrinsic of Linked Data, and they are technically
solvable with appropriate vocabularies, standards and tools, it can be expected
that the development of new LD business models will gradually bridge the gap.
References
1. Klyne G. and Carroll, J. J., eds. Resource Description Framework (RDF): Concepts
and Abstract Syntax. W3C Recommendation. (2004)
2. Cobden, M. et al.: A research agenda for linked closed data, in Proc. of the 2nd Int.
Workshop on Consuming Linked Data (2011)
3. Servant, F.P.: Linking enterprise data, in Proc. of WWW Workshop Linked Data
on the Web, vol. 369 (2008)
4. Villata, S., Delaforge, N., Gandon, F., and Gyrard, A.: An Access Control Model
for Linked Data. In On the Move to Meaningful Internet Sys., pp. 454—463 (2011)
5. Aliprandi, S.: Open licensing and databases. Int. Free and Open Source Software
Law Review, North America, 4(1), pp. 5–18 (2012)
6. Ianella, R. and Guth, S.: ODRL Version 2.0 Common Vocabulary, W3C Community
Group Final Specification (2012)
7. ISO/IEC 21000-5:2004, Information Technology– Multimedia Framework (MPEG-
21)– Part 5: Rights Expression Language (2004)
8. Rodrı́guez-Doncel, V., Delgado, J.: Towards an Expression Language for Licensing
Content in the Connected Semantic Web, in: Proc. of the 9th Int. Workshop on
Virtual Goods (2011)
9. Abelson, H., Adida, B., Linksvayer, M., and Yergler, N.: ccREL: The Creative Com-
mons Rights Expression Language. Technical report. (2008)
10. Villata, S. and Gandon, F.: Licenses Compatibility and Composition in the Web
of Data. in Proc.of the 2nd Int. Workshop on Consuming Linked Data. (2012)
11. Harris K.: Selling and Building Linked Data: Drive Value and Gain Momentum,
in Linking Enterprise Data, pp. 65–76, ed. Springer (2010)