Multi-aspect Ontology for Interoperability in
Human-machine Collective Intelligence Systems for Decision Support
Alexander Smirnov
a
, Tatiana Levashova
b
, Nikolay Shilov
c
and Andrew Ponomarev
d
St. Petersburg Institute for Informatics and Automation of the Russian Academy of Sciences,
39, 14
th
line, St. Petersburg, 199178, Russia
Keywords: Human-machine Collective Intelligence, Semantic Interoperability, Multi-aspect Ontology, Decision
Support.
Abstract: A collective intelligence system could significantly help to improve decision making. Its advantage is that
often collective decisions can be more efficient than individual ones. The paper considers the human-machine
collective intelligence as shared intelligence, which is a product of the collaboration between humans and
software services, their joint efforts and conformed decisions. Usually, multiple collaborators do not share a
common view on the domain or problem they are working on. The paper assumes usage of multi-aspect
ontologies to overcome the problem of different views thus enabling humans and intelligent software services
to self-organize into a collaborative community for decision support. A methodology for development of the
above multi-aspect ontologies is proposed. The major ideas behind the approach are demonstrated by an
example from the smart city domain.
1 INTRODUCTION
Collective intelligence is an emergent property from
the synergies among data-information-knowledge,
software-hardware, and humans with insight that
continually learns from feedback to produce just-in-
time knowledge for better decisions than any of these
elements acting alone. A collective intelligence
system could help organize all these elements to
improve decision making (Glenn 2013). The
Decision 2.0 framework shifting to collective
decisions in the era of Web 2.0, postulates three
general types of approach to accomplish the decision
making objectives. They are outreach, additive
aggregation, and self-organization. The former two
types suppose involvement of various sources
providing ideas and information. The latter type, self-
organization, is mechanisms that enable interactions
among community members, which can result in the
whole being more than the sum of its parts (Bonabeau
2009). That is, self-organization is the mechanism
that can help to achieve the main goal of collective
a
https://orcid.org/0000-0001-8364-073X
b
https://orcid.org/0000-0002-9380-5064
c
https://orcid.org/0000-0002-9264-9127
d
https://orcid.org/0000-0002-1962-7044
intelligence, that is to provide more knowledge than
any individual element provides.
In truly intelligent decision making systems the
elements above are interoperable only with a shared
understanding of the task, the context, and each
other’s perspectives and capabilities (van den Bosch
and Bronkhorst 2018). There are four levels of
interoperability (European Commission 2017):
technical, semantic, organizational and legislative.
Semantic interoperability is understood as shared
semantic interpretation of knowledge presented using
meta-models. The problem of shared knowledge
faces many obstacles in human-machine
environments. Namely, different meanings for terms
(Gruber 2008), diverse data formats, diverse
ontologies reflecting different contexts and area of
practice, diverse classification systems, diverse
folksonomies emerging from social tagging in
various social media (Halpin, Robu and Shepherd
2007), and multiple natural languages (Lévy 2010).
All these obstacles exist when heterogeneous teams
are aiming at providing collective intelligence.
458
Smirnov, A., Levashova, T., Shilov, N. and Ponomarev, A.
Multi-aspect Ontology for Interoperability in Human-machine Collective Intelligence Systems for Decision Support.
DOI: 10.5220/0008356304580465
In Proceedings of the 11th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2019), pages 458-465
ISBN: 978-989-758-382-7
Copyright
c
2019 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
In 2008, T. Gruber addressed the issue of
collective intelligence in the Web, where humans and
machines contribute actively to the resulting
intelligence, each doing what they do best (Gruber
2008). The Semantic Web was supposed the
technology enabling to provide interoperability
between humans and machines by utilizing
ontologies.
Most of the research on the human-machines
activities use multiple ontologies as a mechanism
enabling interoperability. Each ontology is a domain
representation reflecting specifics of a particular
problem this ontology was built for. Difficulties lie in
the necessity to operate not only with different
terminologies but also with different formalisms used
to describe different views. The terminologies and
formalisms, in turn, depend on the tools used for
efficient solving domains’ tasks. The problem of
heterogeneity of these ontologies can be addressed
through having multiple aspects within a common
multi-aspect ontology. The multi-aspect ontology is
defined as an ontology that specifies different
interrelated aspects (facets, constituents,
perspectives) of a complex problem domain. The
multi-aspect ontology provides for the common
vocabulary enabling the interoperability between
different decision-making processes and ontologies
supporting these, and it makes it possible to preserve
internal notations and formalisms suitable for
efficient support of these processes.
This paper addresses the problem of semantic
interoperability support in human-machine collective
intelligence systems through application of multi-
aspect ontologies. The main research contribution is
a methodology for the above ontology development.
The paper is structured as follows. Section 2 presents
a review of related research. The developed
methodology is described in Section 3. It is followed
by the example of the ontology developed for the
smart city domain. Main results are summarized in
the Conclusion.
2 STATE-OF-THE-ART
The Section outlines various approaches to develop
ontologies representing different knowledge
perspectives. The most suitable ones are considered
in detail and analysed.
Ontologies support the formalization of rational
and intuitive decision behaviour in the Pi-Mind
technology (Terziyan, Gryshko and Golovianko
2018). This technology offers a compromise between
human-driven decision-making and machine-driven
decision-making with regard to Industry 4.0. Pi-Mind
captures the best humans’ decision models and
embeds them into decision making processes of
machines. As a result, the machines become able to
make decisions without any human accompany in
situations that are similar to the situations for which
humans’ decision models have been captured. The
technology relies upon three kinds of ontologies:
upper ontology providing basic means for describing
decisions and decision-making; Pi-Mind specific
ontology, describing a value based model of decision-
making; and domain ontologies describing the
structure of decision scenarios for specific domains.
In the automating design domain where
intelligent humanmachine interaction is supposed,
different approaches aiming at modelling the
automatic design knowledge represent different
aspects of design in their ontologies. Examples of
such aspects are process, function, physical product
and issue (Ahmed, Kim and Wallace 2005);
requirement ontology, product finish ontology and
machine motion ontology (Darlington and Culley
2008). The most recent approach (Yin et al. 2015)
distinguishes two aspects: the design ontology to
describe the product and the design process, and the
resource ontology to provide an integrated
representation of human and computer knowledge for
automating design.
The authors of a model-driven interoperability
framework for technical support of co-evolution
strategy of products and manufacturing systems
(Lafleur et al. 2016) address the interoperability
problem by connecting ontologies through
establishing “connector framework” matching these.
This framework connects ontology subclasses
representing product modules, manufacturing
alternatives, and operations. Interoperability between
the product life management tool and the production
capability tools is supported by the ontologies, that
are queried for assessment of the plant capabilities.
Ontology matching (Smirnov and Shilov 2013)
seems to be one of the solutions to the interoperability
problem. But in reality, automatic ontology matching
is still not reliable enough while manual ontology
matching takes too much efforts and time.
Two main and most promising approaches can be
distinguished among the studies on multiple domain
representations using ontologies. They are
multilingual ontology (Espinoza, Montiel-Ponsoda
and Gómez-Pérez 2009) and granular ontologies
(Calegari and Ciucci 2010).
The goal of multilingual ontologies is to resolve
terminological issues that arise due to usage of
different natural languages. Such ontologies are built
Multi-aspect Ontology for Interoperability in Human-machine Collective Intelligence Systems for Decision Support
459
as an ontology comprising language-specific
fragments with relationships between terms.
However, a multilingual ontology is formulated in a
single formalism and collecting together, for
example, knowledge about motivation strategies and
about structure of the problem under consideration
would not be possible without losing certain
semantics.
Granular ontologies rely on the integration of
ontology-based knowledge representation with the
concept of granular computing. Granular computing
is around the notion of granule that links together
similar regarding to a chosen criteria objects or
entities. However, different decision support
processes often overlap in terms of used information
and knowledge. This means that there exist multiple
processes that assume collaboration and usage of the
same information and knowledge. Granular
ontologies cannot solve the problem of terms having
different meaning in different processes.
3 MULTI-ASPECT ONTOLOGY
BUILDING
An analysis (Fernández-López and Gómez-Pérez
2002) of various ontology development
methodologies allows ones to distinguish 5 general
steps in this process: 1) identification of the purpose
and scope of the ontology; 2) identification of
concepts and relationships, and terms to name these
concepts and relationships; 3) ontology engineering;
4) ontology verification; 5) ontology validation.
These steps serve as the guide to develop the multi-
aspect ontology for interoperability support in
human-machine collective intelligence systems.
The multi-aspect ontology is proposed to
comprise three levels: local, aspect, and global. The
local level represents concepts and relationships
observed only from one view. Each aspect can be
represented by specific formalism. The aspect level
represents concepts and relationships from local level
that are shared by two or more aspects. This level
provides a uniform shared ontology representation.
The global level is the common part of the multi-
aspect ontology represented using the representation
provided by the aspect level. The concepts
represented at this level are associated with those in
the aspects.
Development of the multi-aspect ontology
follows the proposed here methodology. At first, the
purpose and scope of the ontology are identified.
Then, the aspects of the ontology are defined based
on the information acquired at the first step and its
logical continuation. Next, ontologies for each of the
aspects are developed. These aspects are integrated
and “global level” is formed out of the concepts that
are considered to be common for the most of aspects.
The steps of verification and validation finalize the
ontology development.
3.1 Identification of the Purpose and
Scope
The purpose of the ontology is determined by the
research problem, i.e., support of interoperability in
human-machine collective intelligence systems
intended for decision support.
The ontology scope is identified based on the
information requirements specified with regard to the
ontology purpose. They include requirements
common for both humans and machines and
requirements having special importance for humans.
Common requirements for interoperability:
Motivation to participate in decision support.
Motivation is a precondition of success of the
collaboration. Moreover, the motivation
influences decision-making process.
Clarity of the problem. The decision support
problem must be clearly represented. The
representation must give to the community
members clear understanding of what they are
expected to do in the current situation (to provide
information, to choose an alternative, to perform
some computations, to do some activities, etc.)
As well, the information based on that decisions
are made must be understandable for the
members. That is, data, alternatives, constraints,
preferences, etc. must be explicitly represented.
Competences accounting. The competences of
the community members must be taken into
account to ensure appropriate decisions.
Negotiation patterns. In complex systems with
heterogeneous members, negotiation patterns
facilitate information/knowledge exchange and
especially useful to organize such exchange
between humans and machines.
Requirements important for humans:
Representations for the problem and associated
information must be human-readable.
Machines are expected to provide support for
complex (e.g., computational) tasks. They are
supposed to self-organize for human support.
KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development
460
3.2 Aspects Definition
The ontology scope is the source for aspects
definition. The set of questions has been formulated
to distinguish these aspects:
Which subproblems of the self-organization of a
community providing collective intelligence are
to be solved with the help of the ontology being
developed?
Which of the subproblems can be solved
separately, and which are inseparable?
Which formalisms are usually used for solving
identified subproblems?
As a result, identified subproblems form aspects
of the multi-aspect ontology, with inseparable ones
being integrated in one aspect, and others (especially
those, that use different knowledge models) into
separate ones.
The present research distinguishes two types of
aspects in the multi-aspect ontology supporting
interoperability in human-machine collective
intelligence systems intended for decision support.
They are basic and specific. The basic aspects are
usually task-independent. They represent concepts
and relationships needed to organize a community
supporting decisions in any domain. The specific
aspects are always task-dependent and make the
community task-oriented.
The set of basic aspects comprises Motivation,
Problem, Competence, and Negotiation protocol
(Figure 1).
Motivation is the reason for participation in the
decision support activity.
Competence is a quality made up of skill and
knowledge needed to successfully complete a task.
Negotiation protocol is a set of rules for
communication of negotiating parties towards
achievement of a desired final outcome.
Figure 1: Ontology aspects.
Problem is the decision support problem to be
solved in the current context. The corresponding
concept is used to represent conventional decision
support problems (situation awareness, problem
identification, development of alternatives, choice of
a preferred alternative, and decision implementation)
and the problem of community self-organization.
Besides, this concept include domain-specific tasks,
i.e. the user tasks for which the community provides
support.
The category of specific aspects is represented by
two concepts Input/Output and Task. The concept
Task represents the user task and the tasks related to
it. For instance, this concept is used for representation
of subtasks when the user task is decomposed.
Input/Output is intended to represent data and
information used at different stages of a decision
support process (context, alternatives, criteria,
preferences, constraints, etc.).
3.3 Development of Aspect Ontologies
At this step ontologies for each of the aspect are
developed. This can be done based on any ontology
development methodology since the aspects are
generally independent (i.e., they can be implemented
using different formalisms and representation
languages). Obviously, the ontology reuse is
beneficial for more or less typical subproblems that
have already been paid significant attention form the
research community (e.g., negotiation protocol
ontology); however, development of ontologies from
scratch is also possible if no appropriate existing
ontologies are found. Aspect ontologies are proposed
to be reused and further developed. Although, here,
the issue of development of these ontologies is not
considered, some thoughts which ontologies can be
reused to form the aspects described in the previous
Section are provided for.
Results obtained in the research on modelling the
motivation domain in Enterprise Architecture
(Azevedo et al. 2011) and on development of
ontologies to represent human emotional, cognitive,
and motivational processes (López-Gil, Gil and
García 2016) can give some ideas of what concepts
and relationships can be used to represent Motivation.
Sources for an ontology to model Competences
can be found in the competence management domain.
Examples of such sources are the ontology for skill
and competence management (Fazel-Zarandi and Fox
2012), the competence model (Miranda et al. 2017),
the competence ontology (Brandmeier et al. 2017),
etc.
Basic
Specific
Aspect
Motivation
Negotiation
protocol
Competence
Problem
Input/Output
Task
is-a
Multi-aspect Ontology for Interoperability in Human-machine Collective Intelligence Systems for Decision Support
461
There are several efforts on development of
ontologies supporting negotiations. An ontology for
automated negotiation in open environments (Tamma
et al. 2002) and its future application to e-commerce
(Tamma et al. 2005) provides different aspects of
negotiation protocol. The negotiation ontology
(Wang, Wong and Wang 2013) supports an ontology
based approach to organize the multi-agent assisted
supply chain negotiations. The mentioned efforts as
well as some others can be used to model the concept
of negotiation protocol.
The concepts of Task and Input/Output are
domain specific and out of the research scope.
3.4 Aspect Integration
At this step, the aspects are analysed with regard to
common concepts that need to be identified and often
taken to the common part of the multi-aspect
ontology. It is useful to write down a list of all such
concepts and then to form a “global level” out of
these. After that, these terms are associated with those
in the aspects. Besides, horizontal relationships
should also be defined at this step for concepts that
are common for two or more aspects, but which are
not high-level enough to be taken into the global
level. A common formalism to represent the common
concepts and the horizontal relationships is defined.
This step is partially described in Section 4.
3.5 Verification
The goal of this step is to ensure the internal
consistency of the developed global level as well as
internal consistencies of the separate aspects taking
into account their relations to other aspects. The step
of ontology verification involves special techniques
and is out of the paper scope.
3.6 Validation
Validation usually takes place during the usage of the
developed multi-aspect ontology in a real-life or
modeled environment. The accumulated issues are
collected, analyzed, and the corresponding
modifications are introduced into the ontology.
Currently, this step is going on and its results will be
available upon completion of this activity.
4 CASE STUDY
New information technologies enable various new
possibilities enhancing our lives. One of products of
this development is appearance of the notion of
“smart city” (Dustdar, Nastić and Šćekić 2017). There
is no common definition of this notion, however, its
common understanding is a coherent urban
development methodology heavily relying on
information and communication technologies to
gather necessary input and provide information for
decision making. Intelligent decision support
collecting information related to the current situation
analysis and assisting in solving various typical
problems becomes essential since otherwise, one can
sink in the ocean of the available today information
and problems to be solved (Anagnostopoulos et al.
2007; Gallacher et al. 2014). Thus, the ontology
purpose is defined as support of interoperability in
human-machine collective intelligence systems
intended for decision support in the smart city
domain.
The scope and aspects do not depend on a
particular domain and are as described in Section 3.
Several representation formalisms for multi-
aspect ontologies have been analysed. The most
progress in this direction is achieved by M. Hemam
who in co-authorship with Z. Boufaïda proposed in
2011 a language for description of multi-viewpoint
ontologies MVP-OWL (Hemam and Boufaïda
2011) extended in 2018 with probability support
(Hemam 2018). In accordance with this notation, the
OWL-DL language was extended in the following
way (only some of the extensions are listed here; for
the complete reference, please, see (Hemam and
Boufaïda 2011)). First, the viewpoints were
introduced (in the current research they correspond to
ontology aspects). Classes and properties were split
into global (observed from two or several viewpoints)
and local (observed only from one viewpoint).
Individuals could only be local. However, taking into
account the possibility of multi-instantiation, they
could be described in several viewpoints and at the
global level simultaneously. Also, four types of
bridge rules were introduced that enable links or
“communication channels” between viewpoints (only
the bidirectional inclusion bridge rule stating that two
concepts under different viewpoints are equal is used
in the example below, indicated with the symbol ̅
̿
).
The presented below ontology is based on
integration of several existing ontologies. Due to the
space restrictions, only three aspects are considered
to illustrate the developed multi-aspect ontology
(Figure 2): Competences, Negotiation Protocol, User
KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development
462
Figure 2: Multi-aspect ontology for three viewpoints.
Task. The three aspects are aimed at different tasks
and, as a result, they use different formalisms (below,
these are described with the most illustrative
concepts).
The task considered in the Negotiation Protocol
aspect is providing agents with ability to
communicate and reach the desired result. Inference
rules are defined on top of the negotiation ontology to
guide agents’ reasoning ability. The negotiation
protocol aspect makes agents’ negotiation behaviours
more adaptive to various negotiation environments
utilizing corresponding negotiation knowledge, that
does not need to be hard-coded in agents, but it is
represented by an ontology (Wang, Wong and Wang
2013), (Tamma et al. 2005). The formalism used in
this aspect is OWL, and the example classes are
Community Member (agent (representing software
service) or human participant of the community
providing for collective intelligence; subject of the
negotiation process), Human (subclass of Community
Member), Agent (subclass of Community Member),
Strategy (outlines the overall coordination method
governing the set of negotiation interactions), Utility
Function (specifies the method to evaluate a proposal
comprising multiple negotiation issues), Parameter
(various variables affecting the negotiation process)
and Role (played by the community members
involved in a negotiation process).
The User Task aspect (category Task in Section
3) is aimed at definition of the user tasks in the
considered domain (in the given case study the
domain is the smart city user information support),
their interdependencies and subtasks, as well as
functional dependencies between their parameters.
Multi-aspect Ontology for Interoperability in Human-machine Collective Intelligence Systems for Decision Support
463
The formalism of object-oriented constraint networks
makes it possible to define functional dependencies
(represented by constraints) between different
parameters of the smart city environment then process
these via a constraint solver when a particular situation
takes place. As a result, the internal representation
basically consists of entities, their parameters and
constraints defined between them. However, for the
interoperability reasons, the following connecting
classes are defined at the aspect level: Entity, Social
(subclass of Entity), Physical (subclass of Entity),
Cyber (subclass of Entity), Parameter, Domain,
subclasses of the Domain class (e.g., Healthcare,
Education, etc.), Rule.
The third example aspect is Competence where
competences of the members of the human-machine
community. The competences are organized into a
hierarchy for facilitating tasks of matching between
competences and tasks to be solved. The following
classes are considered in this aspect: Community
member, Competence, Domain, Competence Level,
Competence Statement (a more detailed description of
this ontology can be found in (Brandmeier et al. 2017)).
In this aspect, an OWL ontology is used.
In accordance with (Hemam and Boufaïda 2011)
the following ontology elements have been defined:
Aspects: Competence, Negotiation Protocol, User
Task.
Global classes: Thing, Parameter, Community
Member, Role, Domain.
Local Classes:
Negotiation Protocol: Human, Agent, Strategy, Utility
Function
User Task: Entity, Social, Physical, Cyber, Rule,
Healthcare, Education, etc.
Competences: Competence, Competence Level,
Competence Statement
Bridge Rules are presented in Figure 3.
When the aspects and bridge rules are defined, one
can use any required formalism inside each of the
aspects. Besides, the existing models can be integrated
into such a multi-aspect ontology without significant
modification.
5 CONCLUSIONS
The paper suggests a methodology for building multi-
aspect ontologies for interoperability support in a
collective intelligence community aimed for decision
support. The suggested methodology consists of six
steps: interoperability requirements definition, aspect
definition, development of aspect ontologies, aspect
integration, verification, and validation.
Parameter
̅
̿
Parameter
NegotiationProtocol
Parameter
̅
̿
Parameter
UserTask
Parameter
̅
̿
CompetenceLevel
Competences
CommunityMember
̅
̿
CommunityMember
NegotiationProtocol
CommunityMember
̅
̿
Entity
UserTask
CommunityMember
̅
̿
CommunityMember
Competences
Role
̅
̿
Role
NegotiationProtocol
Role
̅
̿
Role
UserTask
Domain
̅
̿
Domain
UserTask
Domain
̅
̿
Domain
Competences
i.e., the Roles from different aspects are the same
roles, and Entity from the User Task aspect is
Community Member from the Negotiation
Protocol aspect.
Figure 3: Bridge Rules.
At the current stage of the research, the developed
methodology has proved its eligibility to building
multi-aspect ontologies supporting interoperability in
collective intelligence communities. However, the
“validation” step is currently going on and its results
will be available upon completion of this activity. After
that an analysis of the strong points and weaknesses of
the developed methodology and multi-aspect ontology
for interoperability support in a collective intelligence
community will be performed.
ACKNOWLEDGEMENTS
The research is funded by the Russian Science
Foundation (project # 19-11-00126).
REFERENCES
Ahmed, S., Kim, S. and Wallace, K.M., 2005. A
Methodology for Creating Ontologies for Engineering
Design. In: Volume 3: 25th Computers and Information
in Engineering Conference, Parts A and B. 2005 ASME.
pp. 739750. DOI: https://doi.org/10.1115/DETC2005-
84729.
Anagnostopoulos, A., Broder, A.Z., Gabrilovich, E.,
Josifovski, V. and Riedel, L., 2007. Just-in-time
contextual advertising. In: Proceedings of the sixteenth
ACM conference on Conference on information and
knowledge management - CIKM ’07. 2007 New York,
New York, USA: ACM Press. p. 331. DOI:
https://doi.org/10.1145/1321440.1321488.
Azevedo, C.L.B., Almeida, J.P.A., van Sinderen, M.,
Quartel, D. and Guizzardi, G., 2011. An Ontology-Based
Semantics for the Motivation Extension to ArchiMate.
In: 2011 IEEE 15th International Enterprise Distributed
Object Computing Conference. August 2011 IEEE. pp.
2534. DOI: https://doi.org/10.1109/EDOC.2011.29.
KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development
464
Bonabeau, E., 2009. Decisions 2.0: The power of collective
intelligence, MIT Sloan Management Review, 50(2): 44
53.
van den Bosch, K. and Bronkhorst, A., 2018. Human-AI
cooperation to benefit military decision making. In:
Proceedings of Specialist Meeting Big Data & Artificial
Intelligence for Military Decision Making. 2018 p. DOI:
https://doi.org/10.14339/STO-MP-IST-160.
Brandmeier, M., Neubert, C., Brossog, M. and Franke, J.,
2017. Development of an ontology-based competence
management system. In: 2017 IEEE 15th International
Conference on Industrial Informatics (INDIN). July 2017
IEEE. pp. 601608. DOI: https://doi.org/10.1109/
INDIN.2017.8104840.
Calegari, S. and Ciucci, D., 2010. Granular computing
applied to ontologies, International Journal of
Approximate Reasoning, 51(4): 391409. DOI:
https://doi.org/10.1016/j.ijar.2009.11.006.
Darlington, M.J. and Culley, S.J., 2008. Investigating
ontology development for engineering design support,
Advanced Engineering Informatics, 22(1): 112134.
DOI: https://doi.org/10.1016/j.aei.2007.04.001.
Dustdar, S., Nastić, S. and Šćekić, O., 2017. Smart Cities.
Cham: Springer International Publishing. DOI:
https://doi.org/10.1007/978-3-319-60030-7.
Espinoza, M., Montiel-Ponsoda, E. and Gómez-Pérez, A.,
2009. Ontology localization. In: Proceedings of the fifth
international conference on Knowledge capture - K-CAP
’09. 2009 New York, New York, USA: ACM Press. pp.
3340. DOI: https://doi.org/10.1145/1597735.1597742.
European Commission, 2017. New European
Interoperability Framework: Promoting seamless
services and data flows for European public
administrations. 2017.
Fazel-Zarandi, M. and Fox, M., 2012. An ontology for skill
and competency management. In: Formal Ontology in
Information Systems. Frontiers in Artificial Intelligence
and Applications. IOS Press. p.
Fernández-López, M. and Gómez-Pérez, A., 2002. Overview
and analysis of methodologies for building ontologies,
The Knowledge Engineering Review, 17(2): 129156.
DOI: https://doi.org/10.1017/S0269888902000462.
Gallacher, S., Papadopoulou, E., Abu-Shaaban, Y., Taylor,
N.K. and Williams, M.H., 2014. Dynamic context-aware
personalisation in a pervasive environment, Pervasive
and Mobile Computing, 10: 120137. DOI:
https://doi.org/10.1016/j.pmcj.2012.11.002.
Glenn, J.C., 2013. Collective Intelligence and an Application
by The Millennium Project, World Futures Review, 5(3):
235243. DOI: https://doi.org/10.1177/19467567
13497331.
Gruber, T., 2008. Collective knowledge systems: Where the
Social Web meets the Semantic Web, Journal of Web
Semantics, 6(1): 413. DOI: https://doi.org/10.1016/j.
websem.2007.11.011.
Halpin, H., Robu, V. and Shepherd, H., 2007. The complex
dynamics of collaborative tagging. In: Proceedings of the
16th International WWW Conference. 2007 p.
Hemam, M., 2018. An Extension of the Ontology Web
Language with Multi-Viewpoints and Probabilistic
Reasoning, International Journal of Advanced
Intelligence Paradigms, 10(1): 1. DOI: https://doi.org/
10.1504/IJAIP.2018.10003857.
Hemam, M. and Boufaïda, Z., 2011. MVP-OWL: a multi-
viewpoints ontology language for the Semantic Web,
International Journal of Reasoning-based Intelligent
Systems, 3(3/4): 147. DOI: https://doi.org/10.1504/
IJRIS.2011.043539.
Lafleur, M., Terkaj, W., Belkadi, F., Urgo, M., Bernard, A.
and Colledani, M., 2016. An Onto-Based Interoperability
Framework for the Connection of PLM and Production
Capability Tools. In: PLM 2016: Product Lifecycle
Management for Digital Transformation of Industries.
2016 Springer. pp. 134145. DOI: https://doi.org/
10.1007/978-3-319-54660-5_13.
vy, P., 2010. From social computing to reflexive collective
intelligence: The IEML research program, Information
Sciences, 180(1): 7194. DOI: https://doi.org/
10.1016/j.ins.2009.08.001.
pez-Gil, J.-M., Gil, R. and García, R., 2016. Web
Ontologies to Categorialy Structure Reality:
Representations of Human Emotional, Cognitive, and
Motivational Processes, Frontiers in Psychology, 7: .
DOI: https://doi.org/10.3389/fpsyg.2016.00551.
Miranda, S., Orciuoli, F., Loia, V. and Sampson, D., 2017.
An ontology-based model for competence management,
Data & Knowledge Engineering, 107: 5166. DOI:
https://doi.org/10.1016/j.datak.2016.12.001.
Smirnov, A. and Shilov, N., 2013. Ontology matching in
collaborative recommendation system for PLM,
International Journal of Product Lifecycle Management,
6(4): 322338. DOI: https://doi.org/10.1504/
IJPLM.2013.063210.
Tamma, V., Phelps, S., Dickinson, I. and Wooldridge, M.,
2005. Ontologies for supporting negotiation in e-
commerce, Engineering Applications of Artificial
Intelligence, 18(2): 223236. DOI: https://doi.org/
10.1016/j.engappai.2004.11.011.
Tamma, V., Wooldridge, M., Blacoe, I. and Dickinson, I.,
2002. An Ontology Based Approach to Automated
Negotiation. In: Agent-Mediated Electronic Commerce
IV. Designing Mechanisms and Systems: AAMAS 2002
Workshop on Agent-Mediated Electronic Commerce.
2002 pp. 219237. DOI: https://doi.org/10.1007/3-540-
36378-5_14.
Terziyan, V., Gryshko, S. and Golovianko, M., 2018.
Patented intelligence: Cloning human decision models
for Industry 4.0, Journal of Manufacturing Systems, 48:
204217. DOI: https://doi.org/10.1016/j.jmsy.
2018.04.019.
Wang, G., Wong, T.N. and Wang, X., 2013. An ontology
based approach to organize multi-agent assisted supply
chain negotiations, Computers & Industrial Engineering,
65(1): 215. DOI: https://doi.org/10.1016/j.cie.
2012.06.018.
Yin, Y.H., Nee, A.Y.C., Ong, S.K., Zhu, J.Y., Gu, P.H. and
Chen, L.J., 2015. Automating design with intelligent
humanmachine integration, CIRP Annals, 64(2): 655
677. DOI: https://doi.org/10.1016/j.cirp.2015.05.008.
Multi-aspect Ontology for Interoperability in Human-machine Collective Intelligence Systems for Decision Support
465