COGNITIVE OBJECT FORMAT
H. Castro and A. P. Alves
INESC Porto, Fac. de Eng., Universidade Porto Campus da FEUP, Rua Dr. Roberto Frias, 378 4200 - 465 Porto, Portugal
Keywords: MPEG-21, MPEG-7, DID, Semantic Web, Internet, Metadata, Information Object, Symbol, Concept,
Sensation, Perception, Comprehension, Cognition, Brain.
Abstract: The amount of on-line information content is growing without apparent limits. The lack of a coherent and
consistent structure for its expression leads to increasing problems in terms of desired information retrieval
and rendering. Multiple initiatives have been undertaken to bring forth such a global coherence.
Nonetheless, it is still unattained. The informational landscape is highly fragmented in terms of the formats
of the information object (IO) and their semantic interconnection, which is still incipient. This work exploits
a loose and common sense based analogy between the Internet and the brain for the development of a new
and versatile, MPEG 21 based data structuring format (termed Cognitive Object Format), for the description
of information objects, equating them to cerebral memories. The objective is to enable an easier and more
pervasive human (machine aided) or automatic interpretation and access to IO and their meanings in order
to contribute to the development of a coherent base for their declaration and structuring.
1 INTRODUCTION
The part of living organisms that handles sensing,
interpretation and decision on actions upon reality is
the nervous system. In the more complex organisms,
it includes a brain that is a composed and
heterogeneous structure, whose development was
performed gradually, in accordance with the dictates
of natural evolution. The first nervous systems
began as mere decentralized agglomerates of
sensory and nervous cells. The present state was
only reached with time (Sanes et all, 2006).
In spite of its heterogeneity and concurrent nature,
the brain’s global operation is coherent and
integrated. It is a machine that senses and operates
upon the surrounding world, based on an extensive
processing and internal exchange of information.
The Internet is a greatly distributed and
concurrent system as well. Sensing, interpretations
and actions upon reality also take place within the
Internet. Still, its integration and coherence level is
largely inferior.
The Internet has also endured an evolution.
Initially, it was a decentralized set of structures for
the exchange of information between peer machines
(ARPANET (DARPA), X.25 (ITU-T, 2009),
Fidonet (FidoNet Web Site, 2009) and UUCP (Proj.
Web site 2009)). It then evolved to attain
interoperability between networks (use of TCP/IP)
(O'Regan, 2008). This process is the equivalent to
the integration of different nervous cell
agglomerates.
In its next phase, the DNS system was added in
order to provide a scalable way of finding and
organizing on-line resources. Later, HTML, a
network-based hypertext tool was developed, along
with its transfer protocol (HTTP). These progresses
may be equated to a joint evolution of a nervous
tissue that growingly develops centralized
coordinating capabilities.
The continuous increase of computers’ and data
transmission capabilities are also among the
evolutions that the Internet has undergone. They
have allowed it to sustain growingly complex
interactions and the manipulation and exchange of
ever richer information objects (volumes of
contextually coherent and interpretable digital
information). Thus, just like the nervous system, it
also has evolved in the direction of a growing
complexity and an increasing capacity of integration
and interpretation of reality
For all this, the Internet is comparable to a
loosely coupled, distributed cerebral tissue, where
each computer is a coherent fragment of it. The data
transfer technology is a more voluble equivalent to
351
Castro H. and P. Alves A. (2009).
COGNITIVE OBJECT FORMAT.
In Proceedings of the International Conference on Knowledge Engineering and Ontology Development, pages 351-358
Copyright
c
 SciTePress

the axons (information delivering slender projection
of neurons). The applications running on top of that
structure, implement intra or inter-computer
interaction patterns, which are equivalent to the
different interaction patterns supported by neuronal
assemblies (intimately related sets of neurons). The
hardware and software provisions interfacing with
human users, are equivalent to Internet’s “sensory
organs”.
Under the present analogy, the information
objects (IO) exchanged between the different parts
of a cerebral tissue may be equated to
representations of sensations, or of signifying
cerebral connections. The first are IOs that result
from the primary storage of sensory stimulae
received from the world (e.g. a real (non-synthetic)
video file is a record of a sensing event of a specific
aspect of reality). The later are IOs that store
information with intrinsic meaning within the
cerebral system (symbols connecting sensations to
concepts of the brain’s conceptual “tissue”). They
thus endow sensations with meaning.
In face of what has thus far been exposed, the
ongoing evolution of the Internet appears to be
suggestive of the development of a distributed
nervous tissue that progressively acquires superior
global capacities for the processing, storage and
coherent internal exchange of information, as well as
it develops a greater level of integration, central
coordination, and sensation/interaction with the
outer world. At the present moment, that tissue may
be considered to be at a development level
comparable to the primordial stages of its biological
counterparts. These growing parallels indicate that
an approach based on the cerebral-cognitive
operation for the description of information objects
is advantageous and well prepared to deal with the
predictable evolution of the web, paving the way for
greater future developments. The fact that the
principal interacting agents on the Internet are
human beings is also something that is
advantageously handled by this approach.
We do not claim that this approach will result
in an immediately simpler or faster Internet
operation. Oppositely, an immediate extra-burden,
on all its entities is to be expected. The advantages
are visible only in the greater picture of Internet’s
overall operation. The continuous increase of its data
transport and processing capacities will render this
brain-oriented migration possible and even probable,
and make negligible the burden of the greater
technical responsibilities deposited on the Internet’s
constituting provisions by our proposed approach.
To contribute to this evolution we develop an
analysis of the process through which the brain goes
from the sensing of materiality to the detection of
patterns and shapes in them, to the interpretation and
valuation of the later, to the development of
concepts and signification relationships that are
transversal to multiple sensations and intertwined in
a global sensorial-conceptual tissue. Based on this
analysis, a format was created to structure
informational objects - Cognitive Object Format
(COF).
2 RELATED WORK SCENE
Plenty of work has already been developed on
content description and annotation. The initiatives
undertaken in this area are generally divided
between those oriented towards the semantic web,
employing OWL (W3 Org, 2009) or other RDF (W3
Org, 2009) based ontologies, those devoted to the
annotation and description (especially low-level) of
multimedia content, employing XML based tools
(MPEG 7, MPEG 21, etc), and those that attempt, to
conciliate the first two.
The main focus of the work described in
(Athanasiadis et all, 2005) is the knowledge-based
automatic extraction of semantic information from
multimedia content. Still, an ontology based
structure is used for the expression of content
describing metadata. DOLCE (Gangemi et all, 2002)
is employed as the core ontology. An MPEG 7 based
ontology was used to describe the low-level aspects
of media content. Higher level semantics of the
content were described using DOLCE based domain
specific ontologies.
The works described in (García et all, 2008)
and (Vembu et all, 2006) also merge high-level and
low-level descriptions, where low-level
characteristics of the media record are described
employing an MPEG 7 based ontology, and high-
level semantic aspects about the content are
expressed in an RDF compliant way.
In (Bloehdorn et all, 2005) a work is presented,
which was developed in close proximity to that of
(Athanasiadis et all, 2005), yet with a greater focus
on the formalization of the interrelationship of high
and low-level multimedia concept descriptions.
In (Arndt et all, 2007) the COMM tool is
defined. It reengineers the most important parts of
MPEG 7, (for describing the structure and content of
media items) employing the DOLCE foundational
ontology.
KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development
352

All these works thus attempt, in an RDF
oriented manner, to conciliate the two tendencies in
content meta-description, by converting the
audiovisual feature describing tools (namely MPEG
7) to RDF based ontologies, and inscribing the entire
descriptive metadata (feature and semantic) in a
global RDF based ontology.
As argued in (Stamou et all, 2006), the
information contained in a multimedia document
may be divided into separate layers: the sub-
symbolic, symbolic and logical layers. The first
represents the raw multimedia information. The
second provides a structural layer on top of the
binary media stream so that it is possible to further
process the information, to what the third is devoted.
The mostly used standards for media
descriptions (e.g. Dublin Core, MPEG-7/21, etc)
generally operate on the symbolic level. This
approach presents a problem as the semantics of the
information expressed in such standards are implicit
(to its structure and terminology), and only valid
within its framework, thus impeding interoperability.
This may be handled by replacing the symbolic layer
with one composed of formal, machine-processable
semantics, typically expressed in the RDF language
(Stamou et all, 2006). Broadly, this is the approach
taken by most works in the field, including those
mentioned above. However, it fails to take
advantage of existing XML-based metadata, and
ignores the advantages of an XML-based structural
layer. A purely RDF based semantic description is
very general, open or variable. A tailored RDF based
ontology for low-level technical media characteristic
description presents an overhead when compared to
the existing XML based standards. Furthermore,
logically, the structural layer is not at a cognitively
semantic level, but more at a perceptual one. An
XML based and implicit semantic language, for the
structuring of media items (e.g. MPEG 21) and their
technical description (e.g. MPEG 7) is therefore
more practically and logically appropriate for the
symbolic level.
An alternative solution to the implicitness of
the structural layer’s semantics is thus to add a third
layer (the logical abstraction level) that maps the
structured information sources to the domain’s
formal and explicit knowledge representation, thus
providing the semantics for the symbolic level.
The work presented here is in line with this
later approach. For the middle layer, we employ
MPEG 21 for the overall structuring and relating
of information objects and MPEG 7 (structurally
Figure 1: Revised Semantic Web Stack.
contained within the MPEG 21 body), to describe
substructures within the media objects and their low-
level perceptual characteristics. The logical
abstraction level may employ any number of explicit
semantics annotation tools (RDF based).
This proposal therefore implies a change in the
semantic web stack, (depicted in Figure 11). Its base
would effectively become MPEG 21+MPEG 7+
RDF, instead of only RDF. It is a radical but useful
change. The relative rigidity and implicit semantics
of the symbolic layer tools are an advantage as they
provide simplicity and efficiency. All necessary
semantic explicitness is added by the third layer. An
optimal trade-off is thus achieved between
simplicity, logical correctness and accuracy. The
approach is taken even further by basing it on a
broad view of the human perception structure.
3 COF LOGIC AND STRUCTURE
To develop a logic and a structure for informational
objects that is inspired on the cerebral operation, it is
necessary to elaborate on the manner in which this
structure apprehends reality and coherently stores
valuating information about it. This process is still
relatively unknown, and fairly beyond the skills of
the authors. This analogy is thus based on a present,
common sense view of that process.
For the context of this study, the authors
considered that the apprehension of reality is divided
into three main parts: sensation (sampling of reality
by the sensory organs), perception (processing of
sensory samples and further structural interpretative
COGNITIVE OBJECT FORMAT
353

elaboration upon them) and comprehension
(valuation of reality).
3.1 Sensation Level
The stimulae resulting from the sensing of
materiality are passed, in specific formats, to the
appropriate cortex, submitted to processing, and
storage. Thus, the basic registers are created.
Equivalently, the sensory structures of our
reality sampling devices (e.g. camera) also perform
an initial capture of aspects of reality, which are
subjected to pre-processing, specific encoding and
storage. Thus, the base level of the COF information
structure is that of the basic and non-signifying
sensation registers, the “Sensation Objects” (SO).
Obviously, a video, a sound recording, etc.,
may not be devoid of symbolic value. Still, in a
brain, that signification relationship only exits as an
association that follows sensation.
3.2 Perception Level
Perception consists of the cerebral processing that is
performed over the sensorial information occurring
just above sensation, but not yet at a meaningful
level. It includes:
 the basic perception of the functional aspects, of
the space-time structuring of sensations, and of
the space-time relations between sensations;
 the apprehension of basic features of the sensed
materialities;
 the laying of the perceptual foundations for the
construction of concepts.
For simplicity, in the context of the COF, it is
considered that those activities are functionally
isolable from sensations and procedurally posterior
to them. Therefore, the Perception Objects (PO) are
above the SOs. They are the equivalent to a
crystallization of the abovementioned phase of
apprehending reality into a static description.
POs consist of one ore more SOs and also one
interpretative information carrying object for each of
the types of perceptual processing mentioned above.
The Functional and Space-Time
Characterization Object (FSTCO) carries the
metadata describing the functional aspects of the
capture and register of the sensory data and the
space-time relations between the SOs.
The Conceptualization Root Object (CRO)
carries the metadata that divides the sensation
according to the most relevant shapes and patterns
(see Figure 2). These delineations of objective
Figure 2: Perception of a Visual Sensation.
bodies are the roots of concepts.
The Base Characteristics Object (BCO) carries
metadata identifying the global sensation’s relevant
characteristics and those of each of its sub-portions
(CRO defined). This information is thus bound to
the CRO (presented by the linking of tags 1, 2, and 3
to divisions A, B and C in Figure 22), or to its
corresponding sub-objects.
The POs are of two different types: simple POs
(SPO) containing SOs, FSTCO, CRO and BCO; and
composed POs (CPO) containing other PO, FSTCO
and CRO.
Given that visual or auditory perception, are
different processes in the brain, different types of
sensations must be contained in different SPOs.
3.3 Comprehension Level
Comprehension corresponds to the valuation and
conceptual-symbolic interpretation of “inferior”
sensorial-perceptual constructions. In the COF, the
Comprehension Objects (CO) are above the POs.
They are the crystallization of the semantic
comprehension process into a static description. This
description, which is pure meaning in the context of
the COF, is contained in the “Semantic Objects”
(SmO). The SmO thus expresses relationships
between sensation records and concepts through the
use of (written) symbols.
Each CO may carry one or multiple POs or sub-
Cos, exclusively. They will also contain a CRO and
SmOs. The CRO defines the global concept roots,
based on those of all PO or CO children. The SmOs
play different roles in the COs:
 an SmO may be associated to the CRO (or one
of its segments), performing the connection of
the POs (whose CROs are pointed to by the
global CRO) to the conceptual-symbolic fabric,
KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development
354

within a specific context, endowing them with
meaning;
 an SmO may be associated to a CO, expressing
the “positioning” of the CO in the global
conceptual structure. It also expresses the
semantic relations between the CO and its sub-
objects. This is comparable to a cerebral process
of reflection on other “mental objects”;
 an SmO may also be associated to another
SmO. It performs the contextualization of the
concepts expressed in the target SmO.
Figure 3: COF Structure Example Overview.
The CO may also contain solely SmOs. These COs
correspond to “cerebral IOs” devoid of immediately
associated sensation-perceptions. They may be
viewed as a “thought” over other IOs (CO).
Information on intellectual rights over info
objects may be viewed as such a “thought”. For this,
in the COF structure, the expression of intellectual
rights over IOs will be made with COs carrying only
SmOs that contain rights expression metadata, in
accordance with a precise standard.
Figure 33 presents an overview of some of the
possible structures of COF objects.
4 COF’S STANDARD DATA
FORMAT
4.1 Standards and Tools
The selected standards for the construction of the
COF structure are MPEG-21 (Chiariglione, 2002)
(parts 2 (ISO/IEC FDIS 21000-2, 2005) (DID), 3
(ISO/IEC FDIS 21000-3, 2005) (DII), 5 (ISO/IEC
FDIS 21000-5,2006) (REL), 15 (ISO/IEC FDIS
21000-15, 2006) (ER) and 17 (ISO/IEC FDIS
21000-17, 2006) (Fragment ID)) and MPEG-7
(Chiariglione, 2004). MPEG-21 is used in the
overall structure of COF objects and for other varied
purposes. MPEG-7 tools are used to segment,
characterize and provide meaning to the sensation-
perception objects.
4.2 Data Format
4.2.1 Sensation Objects
The SOs consist only of MPEG-21 DID and DII
metadata and, possibly, of the raw inline media
content. Each SO, is represented by a did:Item
element within a superior structure (the PO). That
did:Item contains a did:Descriptor where a system
wide identifier (wrapped in a DII structure) is
present, as well as a did:Component, that encloses
the media content itself or a URL referencing the
location of the “sensation data” within its child
did:Resource element.
4.2.2 Perception Objects
POs encapsulate the SOs. Each PO is represented by
a did:Item element and contains one or more SO.
The PO also carries an FSTCO, a CRO and a BCO.
The FSTCO is represented by a did:Item
element carrying a series of did:Annotation
elements. The did:Annotations specify (within a
did:Descriptor), functional information regarding
the capturing and recording of the sensory stimulae,
as well as information describing the space-time
relations between the different sensations carried in
the PO. Ergo, in each FSTCO, there is:
 a did:Annotation bound to each SO, specifying
its encoding format. MPEG-7 part 5
(Multimedia Description Schemes) tools are
used for this (MediaFormat D);
 a did:Annotation bound to each segment of each
SO, which specifies the sensation capture
device and any relevant setting related to the
capture episode. MPEG-7 part 5 tools are
employed for this (Creation DS);
 a did:Annotation bound to each fragment of
each SO, specifying the positioning of its
capture point within an xzy relative (to an
arbitrated centre of the overall sensational
event) axis system as a function of time.
Custom metadata were developed for this
purpose.
COGNITIVE OBJECT FORMAT
355

For the time segmentation of the SO resource,
MPEG-7 tools are employed for the definition of
media object time segments (VideoSegment DS and
AudioSegment DS).
The CRO is represented by a did:Item element
that carries a series of did:Annotation elements.
These perform the logical role of attributing
information to some target information body. The
target body’s (an xml element or portion) ID is
specified in the did:Annotation target attribute. The
did:Annotations are logically divided into two
levels:
 the base did:Annotations bind (using a
did:Anchor and did:Fragment element
sequence) an identifying did:Descriptor to a
specific segment of the sensorial record (data
resource). This did:Descriptor is logically
attributed to the did:Resource element of the
targeted did:Component of a specific did:Item
element. They are thus bound to perceived
“objects” within the SO (represented by
did:Item elements);
 the top level did:Annotations are bound to
several base level ones (via the target attribute),
so as to logically unite different “object”
perceptions in different SOs into one single
multi-SO “object” perception. The bound
information is again identifying data.
Figure 44 exemplifies the employment of the
mentioned sequence for the delimitation and
identification of a perceived “object” in an image
fragment (StillRegion DS tools used). Each
did:Annotation thus delimits a perceived “object”
that plays the role of a concept root.
MPEG-7 part 3 (Visual) tools are used to shape
and localize descriptions, if the SO stores a visual
resource (StillRegion DS, StillRegion3D DS,
MovingRegion DS VideoSegment DS). For audio
resources the audio segment defining functionalities
of MPEG-7 are used (AudioSegment DS).
The BCO is represented by a did:Item element that
carries a series of did:Annotation elements. Each
did:Annotation binds low-level feature describing
information to a did:Annotation contained in the
CRO. MPEG-7 part 3 (Visual) tools are used to
describe colour texture and motion if the targeted
fragment is of visual type. MPEG-7 part 4 (Audio)
tools are employed if the targeted fragment is of an
audio type.
......
<did:Annotation
id="item:perception:conceptroot:objectID_1_ro
otID_1"
target="#item:sensation:objectID_1_componentI
D_1">
<did:Descriptor>
<did:Statement mimeType=" text/uri-
list">
item:sensation:objectID_1_ componentID_1_xyz
</did:Statement>
</did:Descriptor>
<did:Anchor>
<did:Fragment>
<mpeg7:Mpeg7>
<mpeg7:Description
xsi:type="mpeg7:ContentEntityType">
<mpeg7:MultimediaContent
xsi:type="mpeg7:ImageType">
<mpeg7:Image>
<mpeg7:SpatialLocator>
<mpeg7:Polygon>
<mpeg7:Coords
mpeg7:dim="2 5">
5 25 10 20 15 15 10 10 5 15
</mpeg7:Coords>
</mpeg7:Polygon>
</mpeg7:SpatialLocator>
</mpeg7:Image>
</mpeg7:MultimediaContent>
</mpeg7:Description>
</mpeg7:Mpeg7>
</did:Fragment>
</did:Anchor>
</did:Annotation>
......
Figure 4: XML Snippet with Annotation, Anchor and
Fragment Sequence.
A number of differences occur if the sensed content
is of a textual nature, and already stored as such (not
as an image for instance). Such an information
object is not as closely related to materiality as its
visual or audio counterparts. No low-level features
are extracted and thus there is no BCO. The textual
content must be marked up according to some XML
mark-up language. The CRO specifies the different
concept roots similarly to what has been explained
above. However, the XML Pointer Language is
employed in the did:Annotations’ target attribute
instead of MPEG-7 tools.
The FSTCO only specifies the coding format
employed in the record of the textual object. In this
case, the fragment specification is carried out with
the XML Pointer language within the value of the
did:Annotations’ target attribute.
A PO may also contain sub-POs instead of SOs.
In this case, the PO’s structure is the same as the
structure described above, except that it has no
BCO. Its FSTCO only specifies the space-time
relation between the sub-POs, by binding the
previously mentioned data to the appropriate time
fragments of the sensorial resources inside the SO of
KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development
356

the PO. The CRO will add further conceptual root
defining did:Annotations to the inferior CROs.
4.2.3 Comprehension and Semantic Objects
The COs encapsulate the POs. Each CO is
represented by a did:Item element and contains one
or more PO and a global CRO object. The CO also
carries a series of SmO with different purposes. The
CRO object binds concept roots from different PO
into joint concept roots.
The SmO are represented by a did:Item element
that contains a series of did:Annotations. These
did:Annotations bind semantic data to the concept
roots defined within the global CRO. There are three
types of SmO:
 the first type endows the PO (its concept roots)
with human interpretable meaning within some
context;
 the second type specifies information and
semantic relations about and between different
POs or SOs;
 the third type specifies the context in which the
signification defined by the first two types of
SmOs is valid.
Given the diversity of information that may be
expressed in the SmO and the vastness and
specificity of the possible contexts for that
information, many different standards and protocols
may be used to express it. In this work, we propose
only a few basic characteristics for that information
and some tools to express it.
The first type of SmO should, if contextually
possible, answer questions, such as who, which
object, which action, where, when, why and how,
and add other relevant commentaries. This
information is then bound to a concept root. MPEG-
7 part 5 (Semantic DS or SemanticBase DS) based
ontologies or others may be employed for this
descriptive purpose.
The second type of SmO employs a custom
ontology. It specifies all the POs, or sub Cos, as the
objects of semantic relationships, which are in their
turn described as well. For instance, a textual
content carrying PO may be declared as a caption or
summary for a visual content carrying PO.
The third type of SmO includes the first two
types. The tools employed in the SmO of the first
type may be used here as well in order to specify the
context of the targeted semantic did:Annotations.
As previously explained, there are also COs
dedicated to the expression of intellectual rights
information over other COs. These are also
represented by did:Item elements containing only
SmOs. The SmOs in question are also represented
by a did:Item element that contains did:Annotations.
Specifically, each of these CO will have two SmO:
 the first SmO carries MPEG-21 REL metadata
to express intellectual rights. It binds such data
to a CO via its did:Annotation’s target attribute;
 the other SmO points to the first SmO.
When the CO carries sub-COs instead of POs, there
are no differences in the rest of its internal structure.
Each type of CO will be contained in its own
separate MPEG-21 DID.
5 COF USAGE
The COF provides a decoupling between structural
and low level technical description and high level
semantic valuation of information objects. The
provision of such a comprehensive and malleable
description base allows a simple and efficient
declaration, delineation, base characterization and
logical structuring of information objects, with
MPEG 21 and MPEG 7. The description of higher
level features with RDF oriented tools enables a
very versatile and precise way of attributing
meaning to information objects.
For all this the COF structure may be employed
in the declaration, structuring and semantic
describing and interrelating of a myriad of different
types of information constructs (audiovisual info
objects, combined textual and audio visual, etc),
which are to be delivered in an integral manner, as
opposed to a fragmented or streamed manner.
6 CONCLUSIONS
This work defines a format to structure and describe
information that is inspired, in a broad sense, by the
way that the brain apprehends reality.
Other developments have been undertaken in
the field of content structuring and annotation,
semantic web (Berners-Lee et all, 2001) and web
integration and interoperability. Still, the present
work merges the content structuring and technically
describing information with the semantically
describing information, within an integrated
approach upon the nature of reality cognition. It
builds on an analogy between the Internet and the
brain, structuring info objects in a way that, makes it
easier for both (computer aided) human and
automatic means to understand them, and opens the
COGNITIVE OBJECT FORMAT
357

way for the expression of increasingly complex
information objects in a “semantically enabled”
manner. That understanding is eased by the
structuring, standardization and uniformization
effects of the employment of the MPEG 21 protocol
at the base of the semantic web stack, as well as by
the use of the comprehensive and precise MPEG 7
protocol to describe “pre-semantic” aspects of media
objects.
Creating, maintaining and using COF content
descriptions may lead to a considerable overhead,
when compared to the sole maintenance of raw data
objects. Still, if compared to the majority of
existing- or under development - description
structures, the presented costs are reasonably the
same. Furthermore, having the definition of COF
being based on two powerful open standards, such as
MPEG-21 and MPEG-7, facilitates its employment,
expansion and conversion into other protocols.
Some possible future work is the replacement
by RDF or OLW based domain specific ontologies,
of the custom metadata tools developed for tasks not
covered by MPEG 7 (space-time relative positioning
of SO objects within a PO, structuring textual SO,
expressing semantic functional relations between
informational objects (PO and SO) as informational
objects, expressing high-level semantic
information).
REFERENCES
DARPA Home Page, http://www.darpa.mil, retrieved
05/03/2009.
Sanes, D. H., Reh, T. A., Harris, W. A., 2006.
“Development of the Nervous System”. Elsevier
Academic Press.
FidoNet Official Web Site, http://www.fidonet.org/,
retrieved 05/03/2009.
O'Regan, G., 2008. “A Brief History of Computing”.
Springer.
ITU-T’s page on X.25, http://www.itu.int/rec/T-REC-
X.25-199610-I/en, retrieved 05/03/2009.
Mosaic Web Browser’s page at NCSA,
http://www.ncsa.uiuc.edu/Projects/mosaic.html,
retrieved 05/03/2009.
MPEG-7 Web Page at Leonardo Chiariglione’s Web Site,
“MPEG-7 Overview”,
http://www.chiariglione.org/mpeg/standards/mpeg-
7/mpeg-7.htm, 2004, retrieved 08/03/2009.
MPEG-21, 2005. “ISO/IEC FDIS 21000-2:2005(E)
MPEG-21 - Part 2: Digital Item Declaration”.
MPEG-21, 2005. “ISO/IEC FDIS 21000-3:2005(E)
MPEG-21 - Part 3: Digital Item Identification”.
MPEG-21, 2006, “ISO/IEC FDIS 21000-5:2006(E)
MPEG-21 - Part 5: Rights Expression Language”.
MPEG-21, 2006. “ISO/IEC FDIS 21000-15:2006(E)
MPEG-21 - Part 15: Event Reporting”.
MPEG-21, 2006. “ISO/IEC FDIS 21000-17:2006(E)
MPEG-21 - Part 17: Fragment Identification of MPEG
Resources”.
MPEG-21 Web Page at Leonardo Chiariglione’s Web
Site, “MPEG-21 Overview v.5”,
http://www.chiariglione.org/mpeg/standards/mpeg-
21/mpeg-21.htm, 2002, retrieved 08/03/2009.
Berners-Lee, T., Hendler, J., and Lassila, O., 2001. “The
semantic web: A new form of web that is meaningful
to computers will unleash a revolution of new
possibilities”. Scientific American.
UUCP Project Web Site, http://www.uucp.org/info.shtml,
retrieved 05/03/2009.
W3C’s OWL specification, http://www.w3.org/TR/owl-
features.
W3C’s RDF specification, http://www.w3.org/RDF.
Athanasiadis, T., Tzouvaras, V., Petridis, K., Precioso, F.,
Avrithis, Y. and Kompatsiaris, Y., 2005. “Using a
Multimedia Ontology Infrastructure for Semantic
Annotation of Multimedia Content”. Proc. of 5th
International Workshop on Knowledge Markup and
Semantic Annotation.
Gangemi, A., Guarino, N., Masolo, C., Oltramari, A., and
Schneider, L., 2002. “Sweetening Ontologies with
DOLCE”. Knowledge Engineering and Knowledge
Management. Ontologies and the Semantic Web.
Proceedings of the 13th International Conference on
Knowledge Acquisition, Modeling and Management.
García, R., Tsinaraki, C., Celma, O., Christodoulakis, S.,
2008. "Multimedia Content Description using
Semantic Web Languages". In Y. Kompatsiaris, P.
Hobson, (Eds.) Semantic Multimedia and Ontologies:
Theory and Applications, pp. 17-54. Springer.
Bloehdorn, S., Petridis, K., Saathoff, C., Simou, N.,
Tzouvaras, V., Avrithis, Y., Handschuh, S.,
Kompatsiaris, I., Staab, S., and Strintzis, M., 2005.
“Semantic annotation of images and videos for
multimedia analysis”. Proc. of European Semantic
Web Conference.
Vembu, S., Kiesel, M., Sintek, M., and Bauman, S., 2006.
"Towards bridging the semantic gap in multimedia
annotation and retrieval". Proc. of First International
Workshop on Semantic Web Annotations for
Multimedia.
Stamou, G., Ossenbruggen, J., Pan, J. Z., Schreiber, G.,
2006. ”Multimedia Annotations on the Semantic
Web”. IEEE MultiMedia, v.13 n.1, p.86-90.
Arndt, R., Troncy, R., Staab, S., Hardman, L., and Vacura,
M., 2007. “COMM: Designing a Well-Founded
Multimedia Ontology for the Web”. In 6th
International Semantic Web Conference.
KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development
358