A Comprehensive Framework for Semantic Annotation of
Web Content
Manuel Fiorelli
1
, Maria Teresa Pazienza
2
and Armando Stellato
2
1
Department of Civil Engineering and Computer Science, University of Rome, Tor Vergata, Rome, Italy
2
Department of Enterprise Engineering,University of Rome, Tor Vergata, Rome, Italy
Keywords: Semantic Annotation, Semantic Web, User Interface, Software Engineering.
Abstract: Studies on Semantic Annotation reveal how trying to match heterogeneous requirements leads to divergent
methodologies, models and processes for annotation management and exchange. Community efforts
towards the development of shared solutions are important to reduce the “entropy” of the field; nonetheless,
any agreement on the ultimate annotation system is unlikely to be achieved. We propose a solution to this
problem by defining a comprehensive framework, unbound to any specific design/annotation model, and
instantiable into concrete system implementations, to meet different requirements. Towards this goal, we
commit to fairly general assumptions, valid across disparate systems and not excessively constraining.
Firstly, most systems deal with combined management of ontologies and Web content. Secondly, these
systems can be described through a common behavioural model, in terms of an assignment of handlers to
predetermined events. This behavioural model can be then enriched through progressive levels of
specification, thus fostering a convention-over-configuration approach in detailing its characteristics. Then,
recurring design fragments can be identified, in order to provide abstractions and specifications for the
definition of concrete handlers.
1 INTRODUCTION
In the envisioned Semantic Web (Berners-Lee et al.,
2001) the meaning of resources, possibly including
services (Payne and Lassila, 2004), is captured
through annotations with respect to well-defined
ontologies. Formalized knowledge is believed to
allow software agents to better interact with Web
resources and perform intelligent tasks on behalf of
humans, such as buying a vacation package from a
virtual travel agency.
The actual deployment of the Semantic Web
required further investigation on pragmatic aspects
related to the publication and the reuse of disparate
knowledge on the Web. This line of development
eventually flowed into the Linked Open Data
movement which elaborated a collection of best-
practices (Heath and Bizer, 2011) aimed at better
connecting the Semantic Web to the architecture of
the Web. About this topic (Heath, 2009) states that
“Linked Data isn’t about rebranding the Semantic
Web, it’s about clarifying its fundamentals”.
The interest on data publication and integration
is complementary to the idea of annotating
traditional information resources (documents,
images, audio and video material), since the former
provides a sound technological and methodological
framework supporting the latter. In line with this
idea, the W3C defines the SKOS vocabulary (W3C,
2009) as a means to establish a link between the
Linked Open Data cloud and the world of
Knowledge Organization Systems (Hodge, 2000),
historically employed by museums, libraries and
other large organizations to better manage and use
their large body of resources.
So far, systems for annotating information
content with respect to formalized knowledge have
followed different and occasionally contrasting
theories. These theories differentiated in many
aspects: the primary focus of the annotation (e.g. is
the traditional content which needs to be annotated
with respect to a generic category, or are specific
ontological resources to be grounded on existing
documentation?), the granularity of the information
to be reported, and the nature of the annotated
elements. Therefore, even the offer of Semantic
Annotation applications is variegated, and it is often
difficult to see all of the requirements for a particular
245
Fiorelli M., Pazienza M. and Stellato A..
A Comprehensive Framework for Semantic Annotation of Web Content.
DOI: 10.5220/0004545702450252
In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2013), pages 245-252
ISBN: 978-989-8565-81-5
Copyright
c
2013 SCITEPRESS (Science and Technology Publications, Lda.)
usage scenario satisfied by a single system.
We propose here a framework for supporting the
development of systems for combined management
of ontological knowledge and Web content,
including, but not limited to Semantic Annotation
Systems. The framework is a subsystem of Semantic
Turkey (Pazienza et al., 2012), a fully-fledged
environment for knowledge management and
acquisition based on RDF technologies (W3C,
2004), with a user interface deployed as a browser
extension. Such an offer guarantees to end
applications a high level of integration among
browsing capabilities, ontology editing and cross-
boundary features concerning both.
The rest of the paper is organized as follows. In
section 2, we review the state-of-the-art in the field.
In section 3, we motivate our work. In section 4, we
elicit requirements for our framework, while we
discuss its architecture in section 5. Finally, in
section 6, we conclude and outline future works.
2 BACKGROUND
We can shortly state that an annotation establishes a
link between two resources, asserting that one is
“somewhat” about the other. The nature of this
association is heavily domain and application
dependent. For instance, informal free-text
annotations are usually found as comments in a
document to drive its edition, while structured
annotations are the output of numerous NLP tasks,
including named entity recognition and relation
extraction. These scenarios depend on different
assumptions regarding the nature of the annotations,
their granularity, their level of formality and the use,
if any, of formal ontologies.
Early works on the annotation of Web resources
include Annotea (Kahan and Koivunen, 2001),
which aimed at establishing a framework for the
collaborative annotation of Web resources. Initially
thought for supporting the collaborative
development of specifications within the W3C, the
project aimed at establishing standards for textual
annotations of marked-up documents.
Later initiatives within the bioinformatics
community, Annotation Ontology (Ciccarese et al.,
2011) and Open Annotation Model (Sanderson and
Van de Sompel, 2010), had a wider breath, aimed at
the annotation of any media type possibly with
respect to a supplied ontology. Those projects
flowed into the Open Annotation W3C community
project, whose mission is to develop an RDF based
model for the annotation of digital artefacts. The
Domeo annotation system developed by (Ciccarese,
et al., 2012) supports the Annotation Ontology and it
is expected to adopt the results of the novel W3C
Community Group. With respect to early attempts, it
is worth of notice that a shared data model is
deemed sufficient, whereas dedicated protocols for
querying and manipulating the annotations are no
longer considered necessary, thanks to the
availability of standards for performing such tasks
developed meanwhile (e.g. SPARQL
(Prud'hommeaux and Seaborne, 2008)).
In the context of these RDF models an
annotation is established though the assertion of at
least a statement relating a resource (the target) to
another (the body) which represents the desired
attachment. In case of Semantic Annotation the
latter is found within a formally defined ontology.
The choice of a domain/application ontology should
reflect the particular point of view behind the
annotation process. (Ma et al., 2011) introduced a
higher order semantics for capturing the meaning of
semantic annotations with respect to the ontological
nature of the attached resource and how it is related
to the target. They also show how different levels of
analysis (i.e. linguistic and semantic) can cooperate,
for example to suggest annotations or highlight
possible errors.
Beyond the problems inherent to the
representation of annotations, there is need for a
clear process to create and maintain them.
According to (Staab et al., 2000), this process should
cope with the evolution of the domain ontology and
the presence of mirrors or altered version of the
annotated resources.
The production of annotations by human users is
often regarded as the bottleneck limiting the scale of
the annotation process. (Kiryakov et al., 2004)
discussed the design issues related to an holistic
system integrating semantic annotations, indexing
and (semantically powered) retrieval. They propose
the reengineering of state-of-the-art NLP tools for
automatically producing semantic annotations with
respect to a lightweight upper ontology, called
KIMO. The existence of a reference upper ontology,
possibly extensible to address domain and
application specific needs, is a distinctive feature,
since most works assume that semantic annotations
are taken against any arbitrary domain ontology.
This idea was implemented in the platform KIM
(Popov et al., 2003), which was heavily tested for
the automatic annotation of news stories.
Finally, (Uren et al., 2006) provided an overview
of Semantic Annotation systems by comparing them
on the basis of a set of requirements that the authors
KEOD2013-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment
246
consider key-features for the annotation task.
3 MOTIVATION
In the previous section we outlined the main
research lines in the field of Semantic Annotation,
showing how conflicting requirements demand
diverging design decisions, making the definitive
annotation system unlikely to appear.
Even a strong agreement on a universal data
model for annotations is difficult to achieve: recent
proposals focus on widely accepted usage scenarios,
therefore failing to satisfy very specific needs.
Divergent methodologies have been proposed to
support manual annotation rather than automatic
generation of annotations. The latter can benefit, as
shown by KIM, of the reuse of state-of-the-art IE
tools; this entailing complex integration challenges.
Incompatible design decisions tend to cumulate,
leading to very different system architectures and
implementations. Therefore, pursuing the goal of
realizing the ultimate annotation system appears to
be fruitless, while it appears reasonable to aim at the
definition of a comprehensive framework supporting
alternative designs.
Analogously, designing a framework unbound to
any prior assumption makes no sense as well,
because an architecture is always based on some
grounding which characterizes its offer to the user.
Therefore, our contribution narrows its scope to
Semantic Web annotation systems and, in general,
any application combining ontological knowledge
with Web content. This is a fairly general model
which avoids any commitment to specific goals,
interaction patterns, methodology (e.g. human
labour vs machine learning) or presentation
mechanisms.
For what concerns the scope of our architecture
(RDF and Web Documents), RDF is by no means
the only formalism to capture semantics, though it is
now widely spread and there are different W3C
recommended vocabularies supporting different
modelling needs. The choice for supporting Web
documents is mostly a starting point (which does not
contradict the generality of the approach), and future
evolutions may foresee extensions for other kind of
sources, different in format or media type.
4 SYNTHESIS OF
REQUIREMENTS
In order to design the architecture of a
comprehensive framework for Semantic Annotation,
we have both analysed state-of-the-art systems, and
taken into consideration principles for their design
acquired from literature.
While we take into account the results of the
discussed standardization efforts (see section 2), we
decided not to commit to a specific model, and have
instead an agnostic approach, which starts from the
mere annotation acts and allows for the adoption of
arbitrary models.
We have thus adopted and incremented the
feature classification provided by (Uren et al., 2006),
and positioned the class of systems that can be
realized with our framework, with respect to those
requirements:
Standard Formats: RDF(S), OWL and SKOS for
the representation of semantic descriptors;
pluggable models for Semantic Annotation (most
notable models provided by default as libraries);
concrete implementations for different ranges
should be provided as component libraries (e.g.
offset or XPointer (DeRose et al., 2002) based
ranges).
User Centred/collaborative Design: the UI for
ontology editing/annotation should be deployed
as a web browser extension, while the browser
itself hosts the web content. This approach
exploits an environment the user is well
acquainted with (the browser), while providing
new functionalities.
Ontology Support: the framework should support
the editing of arbitrary ontologies to be used as
domain for annotations;
Support of Heterogeneous Document Formats: it
is indeed a desirable feature, though currently
our framework is tailored to Web documents;
however, this is a technological limitation of the
current implementation and not a theoretical
choice.
Document Evolution: different choices in the
annotation format and in data preservation may
be more or less prone to degradation with respect
to the evolution of the annotated content; the
framework should permit to retain metadata
about the target document to be able to detect
changes. Option for XPointers guarantees better
resilience to changes than plain offsets;
AComprehensiveFrameworkforSemanticAnnotationofWebContent
247
Annotation Storage: as noted in the (Uren et al.,
2006), there is no universally winning choice for
storing the annotation content: the framework
should thus allow annotations to be stored
separately from the annotated resources (offline
annotations), or to be embedded into them.
Automation: hosting of components for
automatic annotation of content should be
supported, as well as productive exploitation of
their results and suitable interaction with the user
for validating and refining these results.
Granularity: both coarse grain and fragment
level;
5 ARCHITECTURE
This section is organized as follows: we introduce
by first the concepts that have driven the synthesis of
the architecture; we then detail specific design
choices; finally, we describe the end-user
customizability.
5.1 Concept
The proposed framework has to support applications
interacting with Web content. (Kahan and Koivunen,
2001) distinguish two strategies to meet this
requirement: whether dedicated capabilities are
injected into the browser, or into the content
provided by a proxy. Our research effort focuses on
the first approach, by relying on the extensibility of
modern Web browsers to develop the additional
capabilities. The user experience with the browser
does not change in traditional web navigation, and is
only minimally affected when users explicitly
trigger one of the extended annotation capabilities.
Despite being tightly coupled with the Web, a
browser extension is under all aspects a desktop
application, with all the advantages deriving in terms
of robustness, integration with the local system, and
customizability.
In our usage scenario (see ), the traditional
browser frame for visualizing the web content is
complemented with a dedicated panel showing the
reference domain model (e.g. an OWL ontology or a
SKOS concept scheme).
Possible interactions fall into three main
categories, with respect to the resources they affect.
The first category comprises the interactions devoted
to the navigation of the Web, for instance, activating
a hyperlink to reach another Web page. As discussed
in section 3, those interactions are completely
managed by the hosting browser. The user might as
well modify the domain model through interactions
falling into the second category. Finally, there are
interactions that encompass both realms: for
instance, when the user drags a selection of text
from a Web page and drops it onto a resource, as
common in most annotation systems.
Our work develops from this scenario, by
identifying a framework for realizing applications
tied to both ontological resources and Web content,
and not necessarily limited to semantic annotation.
In our setting, we envision unlimited binding
possibilities between annotated content fragments,
their originating sources and the resources belonging
to the domain model. This should allow, for
instance, to generate new ontology individuals while
annotating their occurrences within web pages, to
create and annotate relationships between
individuals, etc..
The framework abstracts a collection of events
out of gestures involving concrete user interface
elements. These events are, to an extent,
independent from the underlying presentation
mechanism and the supporting technology. The
framework dispatches events to suitable handlers,
which implement application dependent logic. Event
handlers must implement a given signature, whereas
there is no prescription on their internal structure.
Within this framework, collections of event
handlers define concrete applications, which might
be characterized through a variety of (possibly
orthogonal) dimensions, including, but not limited
to, the following:
annotation model;
presentation mechanism;
relevant ontological resources.
While two applications might differ along a few
of those dimensions, they could be very close to
each other along others. Therefore, applications are
rarely completely orthogonal and in most cases share
part of their user interface, behaviour and data
management.
The paradigm based on the assignment of
handlers to events meets the requirement of
minimum commitment to the application goals.
Nonetheless, the fact that most applications have
overlapping designs would force the developers to
implement the same user interfaces and behaviours
multiple times. Therefore, a collection of ready-to-
use components for common design fragments is
required.
KEOD2013-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment
248
5.2 Design
In the forthcoming we refer to a combination of an
annotation model, events and related handlers as an
annotation family, and by a slight abuse of language
we will identify possible applications with distinct
annotation families. We discuss here three different
levels for characterizing a family.
Currently, the framework (see ) declares the
following events:
selectionOverResource: when a selection from a
Web page is dropped onto an ontological
resource
resourceOverContent: upon gestures for the
association of Web content with an ontological
resource regardless of their occurrence in the text
contentLoaded: triggered when Web content is
loaded, in order to execute presentation related
activities, e.g. highlighting the annotated
fragments
So far, this basic set of events provides a core
specification, which is sufficient to implement the
entire machinery for an annotation system: handlers
for the first two events encapsulate the logic for the
creation of new annotations, whilst a handler for the
third event is in charge of retrieving and properly
visualizing annotations for a Web content (and for
injecting the code to manage them). For instance,
operations such as the deletion of annotations can
actually be invoked by code which is injected into
the content by handlers intercepting the
contentLoaded event, thus leaving the specification
of these functions opaque to the framework.
The framework treats different genres of RDF
resources (e.g. classes, individuals, and properties)
in a uniform manner, by declaring events concerning
only generic resources. The uniform treatment of
resources entails that the same event might be
handled differently on the basis of the target
resource. Moreover, applications might foresee the
binding of multiple distinct handlers (see ) to an
event related to a single resource, each handler
implementing a distinct way for consuming that
event. A mechanism based on preconditions allows
guarding the execution of handlers on the basis of
contextual information. Standard preconditions are
also defined and provided with the framework. The
basic preconditions include filters based on the role
of the resource (e.g. a class, individual etc..), so that
a given handler may be activated only for certain
resources. The discussion above might be more
accessible through an example concerning the event
selectionOverResource. As stated previously this
event is fired when a selection from a Web page is
dropped onto a resource, regardless of its type.
Actually, this event might be processed in several
ways. By first, a handler may simply annotate an
occurrence of that resource within the Web page.
Other handling strategies include more complex
activities, which are valid only on a subset of the
events. For instance, when the target is a class, a
handler might create and annotate a new instance for
that class, basing on the selected content; otherwise,
if the target is a SKOS concept, another handler
might create and annotate a narrower concept.
Figure 1: Overview of the Annotation Framework. The Web page is annotated with concepts (insects, plants and pesticides)
and relations (isPestOf) from the thesaurus AGROVOC.
AComprehensiveFrameworkforSemanticAnnotationofWebContent
249
Figure 2: Event Based Architecture.
By following a convention-over-configuration
approach to design, we provided a further level of
specification, consisting in a set of interfaces which,
if implemented, can be exploited by the framework
on the basis of the previously defined events. The
following abstract services can thus be implemented
for each family:
checkAnnotationsForContent(contentID)
This function checks whether a given content
source has been annotated. By default, this
function is invoked by a framework
predefined handler, upon triggering of the
contentLoaded event
getAnnotationsForContent(contentID)
This function returns the annotations taken
over a specific content source. Actually, it
returns proxies for the annotations (which
depend on the model) exposing some
framework mandatory fields, such as the id
and range of the annotations. The
implementation/serialization of these
annotation elements is left to the specific
family, and must be consistent with the other
services implemented in the family.
This function is automatically invoked by the
framework after a positive (returned value =
true) check performed by the previous
function (in the context of a contentLoaded
event).
getAnnotationsForResource(RDFResource)
Analogous to the previous one, this function
retrieves all annotations associated to a given
RDF resource.
When constructing a description for a RDF
resource in the UI, the framework may exploit
this function to produce a list of actionable
links to annotated content sources.
decorateContent(annotations)
This is a client function for injecting elements
inside the content, usually to show the
annotations which have been previously taken
over it.
A standard text highlighting mechanism for
web documents is provided by the system and
invoked on the result of a
getAnnotationsForContent(), in the context of
a contentLoaded event. This mechanism can
be overridden by implementing this function
with custom content decorators.
deleteAnnotation(annotID)
This function takes care of removing all the
information related to a given annotation. The
standard highlighter injects calls to this
function for each annotation shown on the
web document.
The system thus, in line with the convention-
over-configuration paradigm, allows for high
flexibility, while reducing effort and need for
detailed specification through massive availability of
conventions (and in some cases, implementations).
5.3 Implementation
The annotation framework we presented is
embedded in the Knowledge Management and
Acquisition Platform Semantic Turkey (Pazienza et
Handlers
Families
bookmarking
open annoation
(coarse grain)
open annoation
(fine grain)
Events
selectionOverResource
resourceOverContent
contentLoaded
selectionOverResource
resourceOverContent
contentLoaded
selectionOverResource
resourceOverContent
contentLoaded
KEOD2013-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment
250
Figure 3: End-user Customization: handlers are enabled for a given event and can then be filtered – by editing their
preconditions – when that event is fired.
al., 2012), and comes out-the-box with a few
annotation families which differ in the underlying
annotation model and, notably, in the tasks they
support. The default handlers take into consideration
the annotation of atomic ontological resources, and
complex activities that are provided as macros, e.g.
the creation of new instances, the definition of new
subclasses in OWL, or of narrower concepts in
SKOS. Semantic Turkey works on a per-project
basis, and by default, annotations are stored as
further RDF metadata inside the RDF repository of
the managed project.
Semantic Turkey allows the deployment of third-
party annotation families, additional preconditions,
or the enrichment of existing ones by the addition of
further handlers. The hosting platform offers to the
implementers a wide choice of reusable capabilities.
The browser provides technologies for the definition
of user interfaces, the manipulation of information
resources and the interaction with the Web. An
annotation family might exploit them to support
inline annotations (included in the document itself),
which can then be saved in a updated copy of the
web page. An annotation family may depend on core
services provided by Semantic Turkey as well as
define new ones for dealing with the specifics of its
annotation mechanism. There is however no limit to
the features that can be provided by adding new
services, e.g. dedicated export mechanisms and
ontology evolution management.
5.4 End-user Customizability
For any given family of annotations, even final users
(i.e. human annotators) may customize their
experience to some extent, with no need of coding
intervention nor of performing complex
configuration on the system.
Concretely, a user can customize a family (see )
by enabling only a portion of the annotation
functionalities associated to each event, or by
refining the preconditions of its associated handlers.
Most usage scenarios in fact, only concern with a
subset of the possible interactions which a given
family may offer, and users may want to enable only
those actions which they are using in their setting.
Users are normally prompted with the list of suitable
handlers (obviously, well presented through
appropriated descriptors) after they trigger an event
as a consequence of performing an action; as an
automatic shortcut, when such a list reduces to a
single handler, it is executed without prompting the
user.
6 CONCLUSIONS AND FUTURE
WORKS
The proposed framework has been experimented in
its evolution, through the development of several
concrete applications for semantic annotation
(Fallucchi et al., 2008; Pazienza et al., 2009;
Pazienza et al., 2012). These experiences have
helped us in understanding the features which a core
framework for semantic annotation should exhibit,
and the right trade-off in flexibility which should be
granted to system developers, while still benefiting
them with concrete support from the software.
Evaluation of frameworks in general is difficult
to perform and is based on non-standard
AComprehensiveFrameworkforSemanticAnnotationofWebContent
251
considerations (e.g. the set of features must be
decided arbitrarily), which are inherently highly
biased by the aspects being put into examination.
However, we plan for the future to offer an overall
view of the features offered by most notable
annotation systems at the current state of the art, and
observe if these can be enabled in our framework.
By emphasizing the amount of development effort
necessary when developing a system with specific
features, and the effort that is required to master our
framework and build those same features over it, we
can obtain a fair map of the improvements and
benefits in adopting it. Regarding further evolutions,
while the framework seems to us general enough in
its basic assumptions, we want to improve it in terms
of concrete support to developers. We will thus
increment the set of available conventions and create
template libraries for recurring annotation patterns.
These libraries will provide partial implementations,
which can be bound to specific needs through
dedicated extension points. Our interest in semi-
supervised processes for knowledge acquisition
(Fiorelli et al., 2010) motivates our attention to
integrating automatic extraction engines and to
combining them with proper human interaction, into
more virtuous acquisition workflows. We have
already explored this approach in (Pazienza et al.,
2012), with the development of a text analytics
system for the discovery of new semantic relations
among concepts belonging to the AGROVOC
thesaurus (Caracciolo et al., 2012). We plan to
integrate this system to the proposed framework and,
in the meanwhile, extend its scope to the projection
of arbitrary information onto an ontology.
REFERENCES
Berners-Lee, T., Hendler, J. A. & Lassila, O., 2001. The
Semantic Web: A new form of Web content that is
meaningful to computers will unleash a revolution of
new possibilities. Scientific American, 279(5), pp.34-43.
Caracciolo, C. et al., 2012. Thesaurus Maintenance,
Alignment and Publication as Linked Data - The
AGROVOC Use Case. International Journal of
Metadata, Semantics and Ontologies (IJMSO), 14
August, 7(1), pp. 65-75.
Ciccarese, P., Ocana, M. & Clark, T., 2012. Open
semantic annotation of scientific publications using
DOMEO. Journal of Biomedical Semantics, Volume
3, pp. 1-14.
Ciccarese, P. et al., 2011. An open annotation ontology for
science on web 3.0. Journal of Biomedical Semantics,
Volume 2, pp. 1-24.
DeRose, S. et al., 2002. XML Pointer Language
(XPointer). [Online] Available at: http://
www.w3.org/TR/xptr/
Fallucchi, F. et al., 2008. Semantic Bookmarking and
Search in the Earth Observation. s.l., Springer, pp.
260-268.
Fiorelli, M. et al., 2010. Computer-aided Ontology
Development: an integrated environment. La Valletta,
Malta, s.n.
Heath, T., 2009. Linked Data? Web of Data? Semantic
Web? WTF?, s.l.: s.n.
Heath, T. & Bizer, C., 2011. Linked data: Evolving the
web into a global data space. Synthesis Lectures on the
Semantic Web: Theory and Technology, 1(1), pp. 1-
136.
Hodge, G., 2000. Systems of Knowledge Organization for
Digital Libraries: Beyond Traditional Authority Files.
Washington, DC: Council on Library and Information
Resources.
Kahan, J. & Koivunen, M.-R., 2001. Annotea: an open
RDF infrastructure for shared Web annotations. Hong
Kong, Hong Kong, ACM, pp. 623-632.
Kiryakov, A. et al., 2004. Semantic annotation, indexing,
and retrieval. Web Semant., #dec#, 2(1), pp. 49-79.
Ma, Y., Lévy, F. & Ghimire, S., 2011. Reasoning with
Annotations of Texts. s.l., s.n.
Payne, T. R. & Lassila, O., 2004. Semantic web services.
IEEE Intelligent Systems, 19(1), pp. 14-15.
Pazienza, M. T., Scarpato, N. & Stellato, A., 2009. STIA:
Experience of Semantic Annotation in Jurisprudence
Domain. s.l., IOS Press, pp. 156-161.
Pazienza, M. T., Scarpato, N., Stellato, A. & Turbati, A.,
2012. Semantic Turkey: A Browser-Integrated
Environment for Knowledge Acquisition and Mana-
gement. Semantic Web Journal, 3(3), pp. 279-292.
Pazienza, M. T. et al., 2012. An Architecture for Data and
Knowledge Acquisition for the Semantic Web: The
AGROVOC Use Case. In: P. Herrero, H. Panetto, R.
Meersman & T. Dillon, eds. On the Move to
Meaningful Internet Systems: OTM 2012 Workshops.
s.l.:Springer Berlin Heidelberg, pp. 426-433.
Popov, B. et al., 2003. KIM Semantic Annotation
Platform. Florida, USA, Springer-Verlag Berlin
Heidelberg, pp. 834-849.
Prud'hommeaux, E. & Seaborne, A., 2008. SPARQL
Query Language for RDF. [Online]
Available at: http://www.w3.org/TR/rdf-sparql-query/
Sanderson, R. & Van de Sompel, H., 2010. Making web
annotations persistent over time. New York, NY,
USA, ACM, pp. 1-10.
Staab, S., Maedche, A. & Handschuh, S., 2000. Creating
Metadata for the Semantic Web—An Annotation
Environment and the Human Factor, s.l.: s.n.
Uren, V. et al., 2006. Semantic annotation for knowledge
management: requirements and a survey of the state of
the art. Journal of Web Semantics, 4(1), pp. 14-28.
W3C, 2004. Resource Description Framework (RDF).
[Online] Available at: http://www.w3.org/RDF/
W3C, 2009. SKOS Simple Knowledge Organization
System Reference. [Online] Available at: http://www.
w3.org/TR/skos-reference/[Accessed 22 March 2011].
KEOD2013-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment
252