=Paper=
{{Paper
|id=Vol-218/paper-3
|storemode=property
|title=Probabilistic Ontologies for Efficient Resource Sharing in Semantic Web Services
|pdfUrl=https://ceur-ws.org/Vol-218/paper3.pdf
|volume=Vol-218
|dblpUrl=https://dblp.org/rec/conf/semweb/CostaLL06
}}
==Probabilistic Ontologies for Efficient Resource Sharing in Semantic Web Services==
https://ceur-ws.org/Vol-218/paper3.pdf
Probabilistic Ontologies for Efficient Resource Sharing
in Semantic Web Services
Paulo Cesar G. da Costa, Kathryn B. Laskey Kenneth J. Laskey#
School of Information Technology and Engineering, MITRE Corporation, M/S H305
George Mason University 7515 Colshire Drive
4400 University Drive McLean, VA 22102-7508 USA
Fairfax, VA 22030-4444 USA klaskey@mitre.org
{pcosta, klaskey}@gmu.edu
Abstract. Service Oriented Architecture (SOA) is a key technology to support
interoperability among data and processing resources. Semantic interoperability
requires mapping between vocabularies of independently developed resources,
a task fraught with uncertainty. Probabilistic ontologies enable representation of
knowledge in domains characterized by uncertainty. As such, they promise to
improve the quality of service descriptions, enable more thorough analysis of
service composition opportunities, and provide a theoretically sound methodol-
ogy for semantic mapping under uncertainty. This paper defines probabilistic
ontologies, discusses their application to SOA, and presents a conceptual
scheme for using a federation of ontologies (with both common and probabilis-
tic ontologies) as a semantic mapping tool for service oriented information ex-
change systems with different levels of service descriptions (including legacy
and probabilistic enabled descriptions).
1 Semantic Interoperability in the Semantic Web and Service
Oriented Architecture Frameworks
A fundamental tenet of the Semantic Web (SW) vision is that adding semantics to
web resources can spark a paradigm shift from information-based data exchange to
knowledge-based data-exchange. HTML syntax hard-codes a limited single-purpose
set of semantic categories. In contrast, the Semantic Web envisions resources anno-
tated with well-defined, explicitly represented semantics that provides the basis for
rich description and reasoning. Explicit semantics is essential for appropriate process-
ing of syntactically identical but semantically different terms (e.g., “Washington” the
President, the city, or the baseball team). Ontologies, or shared repositories of pre-
cisely defined concepts expressed in standardized languages, are a vital tool for ena-
bling semantic interoperability among web resources. Thus, ontologies are a means
for transforming the current “Web of shared data” into a “Web of shared knowledge.”
# ·
The author's affiliation with The MITRE Corporation is provided for identification purposes
only, and is not intended to convey or imply MITRE's concurrence with, or support for, the
positions, opinions or viewpoints expressed by the author.
A traditional ontology can at best list multiple possible senses for a word such as
“Washington,” with no ability to grade their relative plausibility in a given context.
This may be sufficient in tightly controlled environments in which a limited set of
allowable usages of each term can be strictly enforced. However, this approach
cannot be used in an open world where incomplete information is the rule and little
control over ontology publishers is feasible. For this reason, traditional ontologies are
inadequate for unambiguous semantic matching of independently defined concepts.
To fill that gap, probabilistic ontologies are proposed as a means to extend common
ontologies, enabling a principled representation of uncertainty and promoting a means
to represent the relative plausibility of a specific meaning of a concept based on the
context in which it is used.
In parallel with the Semantic Web efforts, Service Oriented Architecture (SOA)
has become the leading approach for accessing and using distributed resources devel-
oped by independent entities and working with independently developed vocabularies
and associated semantics. The advent of SOA marks a transformation from a mostly
data-driven Web, with little interaction between requesters and providers of informa-
tion, into an environment in which information and other resources are accessed and
used in a much more dynamic, interactive, and unpredictable fashion.
The supporting technology for the SOA model is composed of XML-based stan-
dards and protocols focused on providing a shared understanding of the available
services. Currently, accepted standards for developing solutions based on Web
Services (the most prevalent implementation of SOA) include SOAP, a message
structure used for exchanging XML serializations of content and message handling
instructions in a decentralized, distributed environment [1], and the Web Services
Description Language (WSDL), which represents messages exchanged when invok-
ing a Web Service [2]. However, these XML-based structures do not have the ability
to explicitly formalize the underlying semantics of a given Web Service description,
rendering them insufficient to ensure a common understanding of the described Web
Service. As pointed out by Paolucci et al. [3], two identical XML descriptions may
have different meanings depending on who uses them and when. Because it is unreal-
istic to expect that all providers and consumers will have equivalent perspectives and
knowledge regarding a given service, a common understanding of a given Web Serv-
ice can be reached only at the semantic level, where the different perspectives and
knowledge can be matched.
Not surprisingly, the need for semantic-aware resource descriptions is widely rec-
ognized, and is being addressed by research focused on enabling Web Service provid-
ers to describe the properties and capabilities of their Web Services in unambiguous,
computer-interpretable form (e.g., OWL-S [4], WSMO [5], SWSL [6], and
SAWSDL [7]).
This paper argues that progress on both SW and SOA is hampered by the lack of
support for uncertainty in common ontology formalisms. We postulate that probabilis-
tic ontologies can fill a key gap in semantic matching technology, thus facilitating
widespread usage of Web Services for efficient resource sharing in open and distrib-
uted environments. We begin with a general definition of a probabilistic ontology.
Next, we present PR-OWL, a language for defining probabilistic ontologies. Finally,
we explore possible use cases for applying probabilistic ontologies in a Service
Oriented Architecture.
2 Common vs. Probabilistic Ontologies
Since its adoption in the field of Information Systems, the term ontology has been
given many different definitions. A common underlying assumption is that the formal
foundation for knowledge representation and reasoning would be classical logic. We
argue that first-order probabilistic logic is a more appropriate foundation.
The de facto standard for developing ontologies geared to the Semantic Web is
OWL, a W3C recommendation [8]. OWL has its roots in its own web language
predecessors (i.e. XML, RDF), and in traditional knowledge representation formal-
isms that have historically not considered uncertainty. Examples of these formalisms
include Frame systems [9] and Description Logics, which evolved from the so-called
“Structured Inheritance Networks” [10]. This historical background somewhat ex-
plains the lack of support for uncertainty in OWL, a serious limitation for a language
expected to support applications in uncertainty-laden domains such as biogenetics or
medicine. In fact, virtually all current ontology formalisms are based on classical
logic, and SW languages such as OWL provide no consistent support for uncertainty
representation or plausible reasoning.
This lack of support for uncertainty can be justified in closed systems designed to
perform well-defined tasks, for which clear and unambiguous vocabularies can be
constructed. But the Semantic Web vision requires heterogeneous systems to interop-
erate in an open world. Inevitably, vocabularies that are adequate for a single stand-
alone application break down when required to interoperate with systems employing
different vocabularies originally tailored to different tasks. Inevitably, there is in-
complete and partial overlap of terminology and concepts. Even when concepts are
clearly defined, in an open-world system available inputs may be insufficient to de-
termine which meaning is most appropriate. For example, a standard ontology might
enumerate different senses for the word “Washington,” such as the United States as
an agent, the first President of the United States, a state in the Pacific Northwest, or a
baseball team. Semantic Web applications employing the ontology must identify
which of these senses is most appropriate in a given context, e.g., when the word is
embedded in the sentence, “Washington voiced strong objections to the proposed
policy,” extracted from a book about the American revolution, or alternatively, from a
newspaper story about a recent United Nations debate. As another example, the de-
velopers of an ontology for military planning [11] identified over a dozen different
doctrinal uses of the word “clear” within the United States Department of Defense
[12]. In complex open-world problems, legislating unambiguous usage is often infea-
sible. Several items of evidence in combination may be required to disambiguate
among different meanings of a given term. Evidential reasoners require information
about the strength of association between items of evidence and the conclusions to
which they point, as well as contextual factors that affect the strength of evidence.
We argue that the ontology layer is the appropriate place in the Semantic Web ar-
chitecture for representing declarative knowledge about likelihood. That is, in envi-
ronments in which noisy and incomplete information is the rule, likelihood informa-
tion is a key aspect of domain knowledge, and should be included in formal domain
ontologies. A counter-argument has been made that probability (with the possible
exception of microscopic quantum phenomena) is epistemic, but formal ontology
should represent phenomena and relationships as they exist in the world (e.g. [13]).
Carried to its extreme, however, this philosophical stance would preclude the use of
virtually every ontology that has yet been developed. To explore this idea further, we
note that if computational ontologies had existed in the 17th century, Becher and his
followers might well have developed an ontology of phlogiston 1. We may chuckle
now at their naïveté, but who among our 17th century predecessors had the foresight to
judge which of the many scientific theories then in circulation would stand the test of
time? Researchers in medicine, biology, defense, astronomy, and other communities
have developed a plethora of domain ontologies. It is virtually certain that at least
some aspects of some of these ontologies will turn out in retrospect to be as well
founded as the theory of phlogiston. Shall we outlaw use of all these ontologies until
the day we can prove they contain only that which is ontological, and nothing that is
mere epistemology? We take the pragmatic stance that although our ultimate objec-
tive is to seek the truth about Reality as it is, full knowledge is unattainable in the
lifetime of any human. Nevertheless, it is necessary and desirable to do the best we
can with the knowledge we have. To pretend certainty when we are uncertain is not
doing the best we can. Formal ontology provides a useful means of communicating
domain knowledge in a precise and interoperable manner, and of extending and revis-
ing our descriptions as human knowledge accrues. To do this in a sound and princi-
pled manner requires a sound and principled way to represent, communicate, and
reason with uncertainty. Probabilistic ontologies provide a means of doing so.
Not surprisingly, as ontology engineering research has achieved a greater level of
maturity, the need for representing uncertainty in ontologies in a principled way has
become more and more clear. There is increasing interest in extending traditional
ontology formalisms to include sound mechanisms for representing and reasoning
with uncertainty.
Although interest in probabilistic ontologies has been growing, there is as yet no
commonly accepted formal definition of the term. Over the past several decades,
semantically rich and computationally efficient formalisms have emerged for repre-
senting and reasoning with probabilistic knowledge (e.g., [14]-[15]). Building upon
the advances in this area, we have adopted a formal definition based on the core no-
tion that a probabilistic ontology formalism should provide the means to express all
relevant uncertainties about the entities and relationships that exist in a domain in a
logically coherent manner. This would not only provide a consistent representation of
uncertain knowledge that can be reused by different probabilistic systems, but would
also allow applications to perform plausible reasoning with that knowledge.
Definition 1 (from [16]): A probabilistic ontology (PO) is an explicit, formal knowl-
edge representation that expresses knowledge about a domain of application. This
includes:
• Types of entities that exist in the domain;
• Properties of those entities;
• Relationships among entities;
• Processes and events that happen with those entities;
1
Phlogisten is a hypothetical element that was thought to be present in all flammable materials
and was necessary for burning to occur. Becher’s theory of phlogiston was dominant until
Lavoisier proved that combustion requires oxygen.
• Statistical regularities that characterize the domain;
• Inconclusive, ambiguous, incomplete, unreliable, and dissonant knowledge
related to entities of the domain; and
• Uncertainty about all the above forms of knowledge;
where the term entity refers to any concept (real or fictitious, concrete or abstract)
that can be described and reasoned about within the domain of application.
Probabilistic ontologies are used for the purpose of comprehensively describing
knowledge about a domain and the uncertainty associated with that knowledge in a
principled, structured and sharable way, ideally in a format that can be read and proc-
essed by a computer. They also expand the possibilities of standard ontologies by
providing a means of representing the statistical regularities and the uncertain evi-
dence about entities in a domain of application.
It is important to emphasize that a probabilistic ontology is not a probabilistic
model (e.g. a model built using applications such as Netica, Hugin, or
Quiddity*Suite), in the same way that an ontology is not a database application.
Ontologies and database schemas are sometimes confused because they are expressed
using similar formalisms. The real differentiation between the two resides in their
respective intended purposes. Ontologies represent domains in a way that should
facilitate interoperability with other representations of that domain (i.e. other ontolo-
gies built by different people with different views and interests) or of domains that are
not directly related but share some concepts. When a database solution for a given
domain is conceived, its primary focus is not in representing all concepts of a domain
in a way that makes it interoperable with current or future views of that domain, but in
defining the concepts of that domain which would enable storage and retrieval of the
information the database stakeholders (and their customers) want to store and retrieve,
in a way that best fits their requirements.
In a similar vein, when a probabilistic model is built to solve (say) a radar data fu-
sion problem, the main interest driving its creators is not in making sure that their
definitions about radar domain concepts are interoperable with other definitions that
might exist for those same concepts. In contrast, interoperability would definitely be a
primary focus when building a probabilistic ontology for the domain of radar data
fusion. Ontology engineers would attempt to express one view of that domain in a
way that others (with possibly different views) may use/understand and thus build
applications (databases, decision systems, etc) that are compatible with anything built
under that view.
Furthermore, it is not necessary for an ontology to be a running database, yet a da-
tabase application can be built on top of an ontology. Likewise, a probabilistic ontol-
ogy does not necessarily need to be a running probabilistic model, yet a running prob-
abilistic model (i.e. an executable application built using a probabilistic package) can
be built on top of a probabilistic ontology if that fits the objectives of the application
at hand. A subtle difference here is that anything built on top of a traditional ontology
can be built on top of a probabilistic ontology, but the converse is not always true,
since the latter is an extension of the former that adds the above mentioned features of
a probabilistic framework.
As a means to develop probabilistic ontologies that might be used in a framework
for semantic matching of Web Services, we are using and developing PR-OWL [16,
17], an extension that enables OWL ontologies to represent Bayesian probabilistic
models in a way that is flexible enough to be used by diverse Bayesian probabilistic
tools based on different probabilistic technologies. That level of flexibility can only
be achieved using the underlying semantics of first-order Bayesian logic [14], which
is not a part of the standard OWL semantics and abstract syntax. Therefore, it seems
clear that PR-OWL can only be realized via extending the semantics and abstract
syntax of OWL. However, in order to make use of those extensions, it is necessary to
develop new tools supporting the extended syntax and implied semantics of each
extension. Such an effort would require commitment from diverse developers and
workgroups, which falls outside our present scope.
The major advantages of using PR-OWL are its flexibility and representational
power, both inherited from the fact that the language is based on MEBN, a full inte-
gration of first-order logic and probability that merges the expressiveness of the
former with the inferential power of the latter. MEBN provides: (1) a means of ex-
pressing a globally consistent joint distribution over models of any consistent, finitely
axiomatizable FOL theory; (2) a proof theory capable of identifying inconsistent
theories in finitely many steps and converging to correct responses to probabilistic
queries; and (3) a built in mechanism for adding sequences of new axioms and refin-
ing theories in the light of observations. Thus, any knowledge can be represented in
MEBN, provided it can be represented in FOL. Furthermore, because MEBN is a first
order Bayesian logic, using it as the underlying semantics of PR-OWL not only guar-
antees a formal mathematical foundation for a probabilistic extension to the OWL
language (PR-OWL), but also ensures that the advantages of Bayesian Inference (e.g.
natural “Occam’s Razor”, support for learning from data, etc.) will accrue to PR-
OWL probabilistic ontologies. A comprehensive explanation of MEBN logic is out-
side the scope of this work, but the interested reader is directed to [14], [18].
3 The Role of Probabilistic Ontologies in SOA
In order to envision the applicability of POs in SOAs, it is necessary to first under-
stand what kind of uncertainties might be present in a service-oriented environment.
SOA, as defined in its reference model [19], is a paradigm for bringing together needs
and capabilities to address those needs. It requires establishing an execution context
(EC), which is an alignment of all technical and policy-related aspects, including
vocabularies, protocols, licensing, quality of service (QoS), etc. Much of this specific
information is contained in or linked to the service description and/or the description
of underlying capabilities. Considering the complexity involved, many forms of
uncertainty can be present within a given execution context. For example, uncertainty
may arise in the description content (e.g. information annotated with its source but
there is no way to verify whether the identity of the source is correct), in the way
information is captured as part of a description (e.g. information annotated with its
respective source but with no indication of whether it is raw or processed data), or in
the applicability of information to current need (e.g., information on recording
equipment that does not indicate whether the recorded data fall within a reasonable
range for the recording conditions). An ontology that represents statistical information
can enable a reasoner to draw inferences about the missing information. For example,
consider a report that a given device has recorded an ambient temperature of 28 de-
grees Celsius was reported at Washington Dulles Airport (IAD) on 23 January. This is
a highly unlikely, but not impossible, temperature reading for January in the DC Met-
ropolitan area. Statistical information about climate, sensor reliability, and data re-
cording error rates, if represented in the relevant domain ontologies, could be used to
draw inferences about the about the likely temperature at IAD on 23 January that
appropriately account for the possibility of various kinds of error. This example illus-
trates the need for a principled representation of uncertainty in service descriptions, a
feature not found in current specifications.
A typical Web Services scenario can be seen as publish-find-bind triangle, in
which a service provider publishes a service description, a consumer searches a serv-
ice registry for a service satisfying his criteria, analyzes the included information (or
link to information) on the message structure to be exchanged and the address to
exchange it, and interacts with the service to retrieve the resources needed. In this
triangle, there are implicit, unspoken challenges for which a principled representation
of uncertainty is needed. For example:
• The publisher has to choose vocabulary with which to describe the service (or
some other resource related to the service). The vocabulary sets the properties
for that class of item. Service ontology developers attempt to define the “right”
set and structure of properties for the anticipated users. The consumer must know
and understand the semantics of the chosen property vocabulary because these
are the properties used to describe the class and its instances, and the consumer
must understand and use the same vocabulary or there must be a known and ac-
cessible mapping between the properties used for description and those used as
search categories. There are many opportunities for uncertainty about intended
meanings.
• The publisher uses the chosen property vocabulary as the basis to describe and
register instances of that class. This means that the publisher associates values
with the properties and registers the instance. But what is the vocabulary for the
values? All parties may agree that something has the property color and on the
meaning of that property, but if the publisher uses only primary colors and the
subscriber’s search criterion asks for the color pink, the latter will never find a
match for items the first had catalogued. How does a client’s requested value re-
late to a provider’s published values? Do they agree on the vocabulary? Do they
agree on the mechanism to mediate vocabulary mismatches?
• The publisher chooses a property vocabulary and creates instance descriptions by
associating values. One can infer what properties the publisher considers impor-
tant by which properties s/he chooses to populate, assuming values are not neces-
sarily assigned for all possible properties. But what of the consumer’s priorities
when assigning search criteria? If the consumer assigns relative importance, how
does the search engine trade off among different combinations of matches across
the consumer’s search criteria, and how are missing attribute values handled?
Beyond publish-find-bind for a single service, the vision is to provide services at
the appropriate granularity, combining atomic services into more complex tasks. For
example, suppose a supplier needs to find the dimensions and weight limits for cargo
containers for future shipments of items it produces. In today’s integration paradigm,
the supplier would need to query specific shipping agents directly, and might need to
develop special-purpose software interfaces to support interactions with individual
shipping agents. In the envisioned architecture, the supplier would invoke a service
that (i) searches a UDDI registry for shipping agents; (ii) queries each for its respec-
tive restrictions; (iii) compares with the supplier’s requirements; and (iv) selects a
shipper that meets the requirements.
This simple scenario does not include other actions that must be included in such a
transaction. For example, security will be needed to authenticate the supplier to the
shipping agent and the shipping agent to the supplier. Other actions may be required
to establish that each party is authorized to engage in business with the other. The
interaction itself may require a guaranteed level of service that would fall into the
realm of reliable messaging to guarantee delivery. Additionally, the response from
the shipping agent could optionally include video showing details of container pack-
ing and handling, and these would not be appropriate to send if the supplier is using a
low bandwidth communications link.
Security, reliable messaging, and results dissemination are examples of general-
purpose services that could be combined with services for specific business functions,
thus alleviating the business service from the need to create and maintain all support-
ing services. All of these services will have associated service descriptions so that
someone composing a robust service combination can identify the appropriate serv-
ices and the process by which these will work together to provide the higher-level
functionality. That said, what are the uncertainties in identifying the correct services
and combining these to form a consistent package? Is uncertainty even a relevant
concept, or is it a black-and-white issue of whether the pieces fit or not? When trying
to decide among several services that appear to satisfy aspects of the same needed
function, does the ability to reason under uncertainty come into play in identifying the
component services to use and how to combine these?
The above questions do not have simple, universally valid answers. Nevertheless,
we expect that there will be problems for which deterministic implementations of
SOA elements will suffice to build viable solutions, but it is clear that there are issues
that cannot be satisfactorily solved without a principled representation of uncertainty.
Probabilistic ontology languages such as PR-OWL can fulfill this requirement.
Providing a detailed account of how to use PO languages to build standards for
SOA elements, or even examples of (say) service descriptions with probabilistic ele-
ments would require detailed explanation that goes beyond the limits of this paper.
Thus, as a means to explore another possible use of POs in a SOA environment, we
now present a possible framework using a federation of ontologies (common and
probabilistic) for tackling the problem of semantic mapping among concepts used in
Web Services (WS) descriptions within a WS repository.
Figure 1 shows a simplified scheme for SOA using probabilistic semantic map-
ping. As a means to illustrate this scheme, we will devise fictitious examples involv-
ing Web Service providers within the geospatial intelligence domain. In this scheme,
a service consumer or provider that conveys semantic information (ontology that it
abides to, metadata about its requests, parameters, etc.) is called a SOA node Level 1,
whereas a SOA node that has no semantic awareness is called a SOA node Level 0.
Figure 1 – Probabilistic Semantic Mapping for Web Services
In our first use case, S1 needs to plan a convoy route and requests a service for as-
sessing the possibility of finding guerrilla insurgents in a given region. Being a Level
1 client, S1 sends its request with embedded data about the ontology it references and
other semantic information regarding its request (e.g. coordinate system used, ex-
pected QoS, etc.). The WS repository, which itself uses an ontology, finds S4, another
Level 1 client using the same ontology as S1. This ontology is the PR-OWL ontology
“OntB”, which represents a probabilistic model of the geospatial domain and has the
ability to perform a probabilistic assessment of the requested information. In this case,
the request was probabilistic, but the uncertainty involved was related to the service
itself (a probabilistic query on a uncertainty-laden domain), and not to the service
exchanging process. In other words, the exchange was completed using the logical
reasoner alone, since there was a perfect matching in terms of ontologies (both S1 and
S4 abide to the same PR-OWL ontology) and the parameters of the requested service,
and thus no probabilistic mapping was necessary. (yet, note that S1’s query made use
of OntB’s ability to represent uncertainty about the geospatial intelligence domain.)
In a variation of the previous case, let’s suppose that no perfect match between the
request and the available providers is found. In this case, the probabilistic reasoner
accesses the WS repository to search for the most suitable service given the parame-
ters of S1’s request. During that process, it analyses the mapping ontologies related to
“OntB” (the ontology referenced by S1) and the domain ontologies related to the
services it deemed promising to fit S1’s request. In the end, an ordered list of possible
providers is built, and the best possible answers will be returned to S1. This simple
example shows that there might be many combinations of the use of logical and prob-
abilistic reasoners and ontologies to match the needs of a specific request.
4 Conclusion
Our main objective was to discuss the validity of probabilistic ontologies as a prin-
cipled representation of uncertainty in a given domain, and its uses in extending the
reach of Service Oriented Architecture. Although the concept of a semantic-enabled
SOA is in its infancy, we believe much can be achieved by employing both complete
and incomplete knowledge to optimize the way resources are exchanged.
References
[1] Mitra, N. “SOAP Version 1.2 Part 0: Primer”, W3C Recommendation 24 June 2003, available at
http://www.w3.org/TR/soap12-part0/
[2] Chinnici, R.; Moreau, J.; Ryman, A.; and Weerawarana, S. “Web Services Description Language
(WSDL) Version 2.0 Part 1: Core Language”, W3C Candidate Recommendation 27 March 2006.
Available at http://www.w3.org/TR/wsdl20
[3] M. Paolucci, T. Kawamura, T. Payne, and K. Sycara. (2002) “Importing the Semantic Web in
UDDI”. In C. Bussler, R. Hull, S. McIlraith, M. Orlowska, B. Pernici, and J. Yang, editors, Proceed-
ings of the Workshop on Web Services, E-Business and Semantic Web, CAiSE. Toronto, Canada.
[4] Martin, D. (ed.) “OWL-S: Semantic Markup for Web Services”. Available at
http://www.daml.org/services/owl-s/1.1/overview/
[5] Domingue, J.; Lausen, H., and Fensel, d. (Chairs) ESSI Web Services Modeling Ontology Working
Group. Website, available at http://www.wsmo.org.
[6] Battle, S. et al. (2005) “Semantic Web Services Framework (SWSF) Overview”. Available at
http://www.daml.org/services/swsf/1.0/overview/
[7] W3C (2006) Semantic Annotations for WSDL Working Group. Website, available at
http://www.w3.org/2002/ws/sawsdl.
[8] McGuinness, D; and van Harmelen, F. (eds.) “OWL Web Ontology Language Overview”, W3C
recommendation, 10 February 2004. Available at http://www.w3.org/TR/owl-features/
[9] Minsky, M. L. (1975) Framework for Representing Knowledge. In The Psychology of Computer
Vision. P. H. Winston (Eds.), pages 211-277. New York, NY: McGraw-Hill.
[10] Brachman, R. J. (1977) What's in a Concept: Structural Foundations for Semantic Networks. Interna-
tional Journal of Man-Machine Studies, 9(2), 127-152.
[11] Carey, S.; Kleiner, M., Hieb, M. R.; and Brown R. (2001) Standardizing Battle Management Lan-
guage – A Vital Move Towards the Army Transformation, Paper 01F-SIW-067, in: Proceedings of
the IEEE 2001 Fall Simulation Interoperability Workshop.
[12] Field Manual No. FM 1-02 (FM 101-5-1) MCRP 5-12A, Operational Terms and Graphics, Head-
quarters, Department of the Army, Washington, DC, 21 September 2004.
[13] Bittner, T.; and Smith, B. (2001) Vagueness and Granular Partitions. In Proceedings of the Second
International Conference on Formal Ontology in Information Systems (FOIS 2001). Alpha, NJ:
Sheridan Printing Company, Inc.
[14] Laskey, Kathryn B. MEBN: A Logic for Open-World Probabilistic Reasoning. The Volnegau School
of Information Technology and Engineering. George Mason University, Fairfax, VA, USA. Avail-
able at http://ite.gmu.edu/∼klaskey/index.html.
[15] Heckerman, D., Meek, C., & Koller, D. (2004). Probabilistic models for relational data. Redmond,
WA: Microsoft Corporation.
[16] Costa, P. C. G (2005) Bayesian Semantics for the Semantic Web. Doctoral Thesis, School of Infor-
mation Technology and Engineering, George Mason University. Fairfax, VA, USA. Available at
http://hdl.handle.net/1920/455.
[17] Costa, P. C. G.; Laskey, K. B.; and Laskey, K. J. (2005) PR-OWL: A Bayesian Framework for the
Semantic Web. Proceedings of the first workshop on Uncertainty Reasoning for the Semantic Web
(URSW 2005), held at the Fourth International Semantic Web Conference (ISWC 2005). November,
6-10 2005, Galway, Ireland. Available at http://hdl.handle.net/1920/454.
[18] Costa, P. C. G.; Laskey, K. B.; and Laskey (2005) Multi-Entity Bayesian Networks without Multi-
Tears. Draft, Department of Systems Engineering and Operations Research, George Mason Univer-
sity: Fairfax, VA, USA, 2005. Available at http://hdl.handle.net/1920/456.
[19] Mackensie, C. M.; Laskey, K. J.; McCabe, F.; Brown, P. F.; and Metz, R. (2006) Reference Model
for Service Oriented Architecture 1.0. Committee Specification 1, 19 July 2006. Available at
http://www.oasis-open.org/committees/download.php/19361/soa-rm-cs.pdf