d4.4. ontology model for the episecc use case 4_deliverable_report.pdfepisecc_wp4_d4...
TRANSCRIPT
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |1
www.episecc.eu
D4.4. – Ontology model for the
EPISECC use case
Grant agreement number: 607078 Date of deliverable: 2016-04-30
Date of project start: 2014-06-01 Date of submission: 2016-06-15
Duration of project: 36 months Deliverable approved by: AIT
TCCA
Lead Beneficiary: UNIST
Contributing Beneficiaries: AIT, HITEC, DLR, HWC, FRQ, TCCA, KULeuven
Establish Pan-European Information Space to Enhance seCurity of Citizens
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |2
www.episecc.eu
Executive Summary
The deliverable provides detailed description of the Ontology model for the EPISECC use case
including methodology, model design and validation.
The methodology explains three levels of ontologies and their usage. Upper ontologies describe
common concepts across a wide range of human domains and domain ontologies narrow the scope
to the particular domain such as medicine or computer science. The EPISECC Ontology model is an
application ontology and it serves particular task or a software application. The method used for the
ontology construction can be summarized in two steps. The first step is the modelling of the
application knowledge, when experts define basic concepts and relations among them, and axioms
together with rules for data interpretation and reasoning. During the second step the application
concepts are linked with the concepts in both reference upper and domain ontologies.
Following this approach, the ontology scope is defined firstly. The EPISECC use case narrows down
the scope of the ontology to the scenario designed for the Proof of Concept. Deliverable 6.1 “Proof of
Concept design” elaborates the scenario and this deliverable summarises the exchanged and
dynamic data from the scenario’s episodes and use cases. Thus, the scope of the EPISECC Ontology
model encompasses the situational and operational data exchanged among the organisations in the
EPISECC use case. The EPISECC Ontology models the common operational picture with dynamic
spatial information or spatio-temporal/4D data, showing the situation in affected area and locations
of resources.
Knowing the scope of the ontology, a list of key concepts is defined. For the EPISECC Ontology model,
a subset of concepts of the EPISECC Taxonomy (D4.2) is selected according to the scope of the
EPISECC use case. Besides the concepts and their hierarchy, the relationships or properties between
classes are derived. Nine directed graphs are developed and presented in this deliverable. Two
domain ontologies, A Geographic Query Language for RDF Data (GeoSPARQL) and World Wide Web
Consortium (W3C) Time, describing spatial and temporal concepts are selected and referenced. As
the reference upper level ontology, the Descriptive Ontology for Linguistic and Cognitive Engineering
(DOLCE) Lite ontology is selected and referenced too.
This deliverable describes the first implementation of the EPISECC Ontology model. Once the
conceptual model is developed and upper and domain ontologies referenced, the ontology schema
as the formalisation of the conceptual model is created. Software-interpretable constructs are
defined by formal languages Resource Description Framework Schema (RDFS) and Ontology Web
Language (OWL). The EPISECC Ontology schema is created with Protégé Desktop and the logical
consistency is checked by applying reasoning algorithms.
The second implementation and validation will be done through the Proof of concept (Tasks 6.2 and
6.3) and the final Ontology model for the EPISECC use case will be elaborated in the Task 4.5
“Lessons learned from proof of concept and validation” and delivered in the Deliverable 4.5.
An excerpt of the EPISECC Ontology model in OWL file is given in the Annex.
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |3
www.episecc.eu
Table of Content
Table of Content ...................................................................................................................................... 3
List of Tables ............................................................................................................................................ 4
List of Figures ........................................................................................................................................... 5
List of Acronyms ...................................................................................................................................... 6
1. Introduction ..................................................................................................................................... 9
2. Methodology ................................................................................................................................. 11
2.1. The EPISECC Ontology model – an application ontology ....................................................... 11
2.2. Ontology development methodology .................................................................................... 12
2.3. Formal languages and tools ................................................................................................... 15
3. Ontology model ............................................................................................................................. 17
3.1. Ontology scope – the EPISECC use case................................................................................. 17
3.2. Ontology conceptual model ................................................................................................... 21
3.2.1. An overview of the existing ontologies ......................................................................... 22
3.2.2. Ontology model’s elements .......................................................................................... 23
3.2.3. Key concepts and their relations ................................................................................... 25
3.2.4. References to domain and upper ontologies ................................................................ 31
3.2.4.1. References to domain ontologies ................................................................................. 31
3.2.4.2. References to upper level ontology .............................................................................. 35
4. Ontology schema ........................................................................................................................... 38
4.1. Implementation in Protégé Desktop ...................................................................................... 38
4.2. Validation ............................................................................................................................... 40
5. Conclusions .................................................................................................................................... 41
Bibliography ........................................................................................................................................... 42
Annex ..................................................................................................................................................... 45
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |4
www.episecc.eu
List of Tables
Table 1: The EPISECC use case data ...................................................................................................... 20
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |5
www.episecc.eu
List of Figures
Figure 1: The ontology dimensions [4] .................................................................................................. 10
Figure 2: The ontology levels (adopted from [6]) ................................................................................. 11
Figure 3: Two-steps approach for building ontology ............................................................................ 13
Figure 4: The common operational picture data classes (adopted from [24]) ..................................... 19
Figure 5: The EPISECC Ontology model’s scope .................................................................................... 21
Figure 6: Ontology model’s main elements and their relations (adopted from [27])........................... 23
Figure 7: Example of the directed graph: generic (a) and specific one (b) ........................................... 24
Figure 8: Directed graph of the main classes and properties ............................................................... 25
Figure 9: Directed graph of the class Disaster ....................................................................................... 26
Figure 10: Directed graph of the class Process ..................................................................................... 27
Figure 11: Directed graph of the class Resource ................................................................................... 28
Figure 12: Directed graph of the class Organisation ............................................................................. 29
Figure 13: Directed graph of the class Common operational picture ................................................... 30
Figure 14: Directed graph of common operational picture containing 4D data [34] ........................... 32
Figure 15: GeoSPARQL directed graph [15]........................................................................................... 33
Figure 16: Subclasses of the class ‘Spatiotemporal part’ (adopted from [34]) ..................................... 34
Figure 17: W3C Time directed graph [16] ............................................................................................. 35
Figure 18: DOLCE basic categories of particulars [14] ........................................................................... 36
Figure 19: Directed graph of main EPISECC classes and with references to DOLCE-Lite, GeoSPARQL
and W3C Time ontology classes (adopted from [34]) ........................................................................... 37
Figure 20: Protégé Desktop interface showing imported ontologies ................................................... 38
Figure 21: Protégé Desktop interface showing the class hierarchies ................................................... 39
Figure 22 Protégé Desktop interface showing the object and data properties .................................... 40
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |6
www.episecc.eu
List of Acronyms
Abbreviation Description
BFO Basic Formal Ontology
CAP Common Alerting Protocol
CIS Common Information System
COP Common Operational Picture
DOLCE Descriptive Ontology for Linguistic and Cognitive Engineering
EMSI Emergency Management Shared Information
GeoSPARQL A Geographic Query Language for RDF Data
GML Geography Markup Language
ISO International Organisation for Standardization
LEMA Local Emergency Management Authority
OGC Open Geospatial Consortium
OWL Ontology Web Language
PPDR Public Protection and Disaster Relief
QUDT Quantities, Units, Dimensions and Types
RDF Resource Description Framework
RDFS Resource Description Framework Schema
RIF Rule Interchange Format
SKOS Simple Knowledge Organisation System
SKOS-thes SKOS extension for thesauri
SPARQL SPARQL Protocol and RDF Query Language
SQL Structured Query Language
SSN Semantic Sensor Network
SUMO Suggested Upper Merged Ontology
SWEET Semantic Web for Earth and Environmental Terminology
TSO Tactical Situation Object
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |7
www.episecc.eu
URI Uniform Resource Identifier
W3C World Wide Web Consortium
WGS 84 World Geodetic System 84
WKT Well-known text
XML Extensible Markup Language
XSD XML Schema Definition
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |8
www.episecc.eu
List of Terms
Term Definition
Concept Concepts are the units of thought or ideas, meanings, or (categories of) objects and events.
Data type properties Data type properties relate individuals to data values.
Facet A particular aspect or feature of something (Oxford dictionary).
Object properties Object properties relate individuals to individuals.
Ontology A set of concepts and categories in a subject area or domain that shows their properties and the relations between them (Oxford dictionary).
Ontology model Ontology model is a conceptual model of a domain consisting of classes, properties and individuals.
Ontology schema The description of the ontology in a formal language is named ‘ontology schema’.
Resource Description Framework (RDF) data model
RDF is a standard model for data interchange on the web (W3C).
Resource Description Framework Schema (RDFS)
RDF's vocabulary description language. RDF Schema defines classes and properties that may be used to describe classes, properties and other resources (W3C).
Reasoning algorithms A set of algorithms that use rules of logic to infer new statements from available knowledge.
Simple Knowledge Organisation System (SKOS)
A representation of knowledge organisation systems within the framework of the Semantic web.
SPARQL query language A query language for RDF.
Taxonomy A scheme of classification (Oxford dictionary).
Term Term is a label to a concept (D4.2).
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |9
www.episecc.eu
1. Introduction
A response to a critical event, as the EPISECC project’s universe of discourse, is a complex human
domain described as “a complex dynamic system composed of actions taken in a certain spatial,
technical, organisational, and legal environment during a disaster, including one or more situations
which straightforwardly lead to a disaster, as well as handing over to a recovery phase” (D4.2).
Access to information and communication with other first responders are key factors for effective
emergency operations and the current status of sharing information between the teams is not
satisfactory. The FP7 projects ACRIMAS [1] and CRISYS [2] recognised that as a major gap. The
interoperability gap is further described in the EPISECC project Deliverable 2.1 “PPDR Information
Space - Status quo of commercial, research and governmental projects and applications”. Results
revealed that more than 70% of the investigated tools reach physical and syntactical interoperability,
but only 30% (12 of 41 investigated tools) are covering the semantic interoperability. Therefore, the
EPISECC project elaborates three semantic structures facilitating semantic interoperability:
taxonomy, semantic repository and ontology model.
The backbone semantic structure is the EPISECC Taxonomy describing concepts used in “a response
to a critical event” and organizing them in hierarchies and facets (D4.2). The EPISECC Data model
provides database structure called Semantic Repository for the storage of the semantic descriptions
in the Common Information Space (CIS). The Semantic Repository stores the EPISECC Taxonomy, the
emergency organisations concepts and the semantic mappings of organisations concepts to the
EPISECC Taxonomy. The third semantic structure is the Ontology model for the EPISECC use case
(hereinafter also referred to as the EPISECC Ontology model). The EPISECC Ontology model describes
the EPISECC use case as a model of domain knowledge. The main goal is to enrich semantic
descriptions of the EPISECC Taxonomy by describing an existing situation that is highly challenging in
terms of number of agencies involved, cross-border situation and interoperability needs.
Deliverable 6.1 “Proof of Concept design” elaborates the EPISECC use case that defines a scope of the
EPISECC Ontology model. The focus of the EPISECC use case is situational awareness and definition of
the most urgent resources requested by first responders. Deliverable 6.1 defines situational and
operational data as the most important data to be exchanged and both are dynamic data including
geolocations. Therefore, the EPISECC Ontology model has main goal to model spatio-temporal data
or 4D data.
What is ontology, ontology model and ontology schema?
Ontology is the branch of metaphysics dealing with the nature of being. In the domain of computer
sciences ontologies are used for knowledge representations. Ontology is a set of concepts and
categories in a subject area or domain that shows their properties and the relations between them
(Oxford dictionary). Tom Gruber defines ontology as a formal specification of a shared
conceptualisation [3]. Today, the term ontology covers various things, from formal upper-level
ontologies expressed in first order logic such as DOLCE-Lite to the simple list of keywords used to
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |10
www.episecc.eu
annotate resources on the Web. In between the two are vocabularies and taxonomies. Ontologies
represent subsumption or ‘is-a’ relationships among entities but also other kinds of relationships
such as functional or physical. The Ontology Summit 2007 [4] has proposed so called ‘Framework’
which defines a limited number of dimensions along which ontologies can be characterized. The
Framework is comprised of semantic and pragmatic dimensions as shown on Figure 1. Semantic
dimensions include expressiveness, structure, and representational granularity. Pragmatic
dimensions include intended use, use of automated reasoning, and prescriptive versus descriptive.
Figure 1: The ontology dimensions [4]
In the context of information systems development such as CIS, the common goal of developing
ontologies is sharing common understanding of information among people or software agents [5].
Ontology model is a conceptual model of a domain consisting of classes, properties and individuals.
Classes describe concepts in the domain, properties describe relations among classes and individuals
are realisations of classes as data items. A class can have subclasses and a property can have
subproperties, thus constituting hierarchies of classes and properties.
Description of the ontology in a formal language is named ‘ontology schema’. Ontology schemas can
be compared to physical schemas or table definitions in relational database. Formal languages such
as Resource Description Framework Schema (RDFS) and Ontology Web Language (OWL) are used to
describe ontology models with software-interpretable constructs and thus to create an ontology
schema.
This report delivers the methodological approach for the development of the EPISECC Ontology
model, its content and relations to upper ontology models. The last chapter presents the EPISECC
ontology schema: a description of the EPISECC Ontology model by OWL constructs applicable to be
used by software application.
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |11
www.episecc.eu
2. Methodology
The development of the ontology model follows the methodology described in this chapter. The
overall approach includes two steps: modelling application knowledge and linking the model with
reference upper and domain ontologies. Reuse of existing ontologies should be considered as well.
The chapter starts with a brief explanation of the ontology levels, continues with the ontology
development methodology and ends with the formal languages and tools used for the ontology
development.
2.1. The EPISECC Ontology model – an application ontology
The EPISECC Ontology model is an application ontology what means that it serves a particular
software application. Figure 2 shows three levels of ontologies [6]:
upper ontologies,
domain ontologies, and
application ontologies.
An upper ontology (or foundation or top-level ontology) describes common concepts across a wide
range of human domains. There are several upper ontologies developed so far such as Basic Formal
Ontology (BFO), Suggested Upper Merged Ontology (SUMO) and Descriptive Ontology for Linguistic
and Cognitive Engineering (DOLCE). Upper ontologies are developed by philosophers and cognitive
engineers and they provide concepts and relations to domain ontologies for their merging.
Figure 2: The ontology levels (adopted from [6])
A domain ontology describes concepts of a particular human domain such as medicine, computer
science etc., and it is developed by domain experts. Domain concepts could be referenced to the
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |12
www.episecc.eu
concepts of upper ontologies and thus, the domain ontologies that use the same upper ontology can
be merged automatically. The domain of emergency management does not have a specifically
intended ontology model. Liu et al. in [7] present a comprehensive review of the ontologies for crisis
management. The review identifies a total of 11 subject areas represented in contemporary crisis
information systems including critical infrastructures, resource management, decision support,
situation awareness, response coordination, etc. but the authors state the following: “Moreover, no
formal ontologies are specifically intended for representing the emergency and disaster domain. The
existing ontologies that represent the disaster subject area are mainly in the form of database
schemas, which are not publicly available.” [7]. The World Wide Web Consortium (W3C) published
the report “Emergency Information Interoperability Frameworks” in 2009 [8]. This document
describes candidate components of the emergency management ontology based on common use
cases.
Application ontology describes concepts used by a particular software application e.g. data and
software application concepts used by a web service. Application concepts could be referenced to
the concepts of domain ontologies and the application ontologies that use the same domain
ontology can be merged automatically. Application ontologies are developed by software developers.
The EPISECC Ontology model is an application ontology serving emergency response tasks for the
EPISECC use case. Sharing situational and operational data among the first responders is considered
as the critical one for the success of rescue operations. Therefore, the focus of the EPISECC Ontology
is modelling of spatio-temporal or 4D data showing locations of the most urgent resources requested
by the first responders.
2.2. Ontology development methodology
Ontology could be built either from scratch, by reusing existing ontologies, or by modifying existing
ontologies to the application domain. There are four ontology development methodologies recently
developed: Methontology [9], On-To-Knowledge [10], Diligient [11] and NeOn Methodology [12].
Methontology is a comprehensive methodology used to build ontologies from scratch and at the
conceptual level. On-To-Knowledge methodology includes a tool environment for semantic
information processing of large numbers of heterogeneous documents and it speeds up knowledge
management in large companies. Diligent methodology is focussing on the evolution of ontologies
instead of the initial design. NeOn Methodology is a scenario-based methodology that supports the
collaborative aspects of ontology development and reuse, as well as the dynamic evolution of
ontology networks in distributed environments. Based on the above mentioned methodologies and
taking into account the specifics of spatial data, the authors Kun et al. in [13] have proposed the
method of ontology construction for spatial data combining bottom-to-top and top-to-bottom
approach in two steps as follows.
The first step (or bottom-to-top) models the application knowledge. Experts define basic
concepts and relations among them, and define axioms and rules for data interpretation and
reasoning.
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |13
www.episecc.eu
The second step (or top-to-bottom) links the application concepts with concepts in the
reference upper and domain ontologies. One should consider reuse of the already developed
ontological resources.
The above two-steps approach is selected for the development of the EPISECC Ontology model
because the approach suits the ontology scope and application level. Figure 3 shows the sources of
application knowledge for the EPISECC Ontology model:
the EPISECC Taxonomy (D4.2),
the EPISECC use case (D6.1).
The upper and domain ontologies used are the following:
Descriptive Ontology for Linguistic and Cognitive Engineering (DOLCE)-Lite (upper ontology)
[14],
A Geographic Query Language for RDF data (GeoSPARQL) (domain ontology for spatial data)
[15], and
W3C Time (domain ontology for temporal data) [16].
Figure 3: Two-steps approach for building ontology
The selected methodology for the EPISECC Ontology model development consists of the following
seven tasks.
1. Define the scope of ontology
2. List the key concepts in the selected domain/application
3. Establish the conceptual model for ontology
4. Reuse the existing ontologies and/or make references to domain and upper ontologies
5. Formalize ontology and populate it with individuals
6. Validate ontology
7. Revise and refine ontology
A brief description of the above development tasks follow.
Ontology development starts by defining its domain and scope. Answering on several basic questions
can help limiting the scope of ontology. Examples of such questions are given in [17]:
What is the domain that the ontology will cover?
For what we are going to use the ontology?
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |14
www.episecc.eu
For what types of questions the information in the ontology should provide answers?
Who will use and maintain the ontology?
The EPISECC Ontology model scope is defined by the EPISECC use case (D6.1) and described in
Chapter 3.
Knowing what is the scope of ontology, a list of key concepts and terms can be defined. Concepts are
selected if they are used in writing statements or explaining the ontology scope to a user. What are
the terms and their properties used in explanations? For the EPISECC Ontology model, a subset of
concepts and terms of the EPISECC Taxonomy (D4.2) is selected according to the scope of the
EPISECC use case.
To establish a conceptual model for ontology, classes and properties (also called relations) with their
hierarchies (or “is-a” relation) must be defined. Classes and class hierarchy are defined by the
EPISECC Taxonomy (D4.2), but properties, subproperties and property hierarchy must be derived.
Properties can be seen as four types as follows [17]:
intrinsic properties such as the capacity of a vehicle;
extrinsic properties such as a rescuer’s name;
parts, if the object is structured; these can be both physical and abstract parts (e.g., the
actions of a decontamination);
relationships between individual members of the class and other items (e.g., the owner of a
vehicle, representing a relationship between a vehicle and a organization).
Property can have other aspects such as cardinality, value type, domain, range etc. Besides classes
and properties, a conceptual model can define axioms and rules for data interpretation and
reasoning. Conceptual model for ontology is expressed as one or more directed graphs. Additional
descriptions such as domains and ranges are documented by a text coupled with the directed graphs.
Conceptual model of the EPISECC Ontology model is described in Chapter 3.
Many ontologies are already available and there are libraries of reusable ontologies (also called
design patterns) on the Web such as Ontology design pattern Web portal [18]. It is worth considering
existing work and checking if this can be refined and extended to suit the scope of new ontology. In
case that an application ontology needs to interact with other applications, their ontologies must be
considered and referenced. The common way to merge application level ontologies is by referencing
them to the same upper and domain level ontologies. The EPISECC Ontology model is an application
level ontology and there are no existing ontologies to be reused, but there are existing domain and
upper ontologies to which the EPISECC Ontology is referenced.
Once the conceptual model is developed and upper and domain ontologies selected, the ontology
schema can be created and populated with individuals. Ontology schema is a formalisation of the
conceptual model. Software-interpretable constructs are defined by formal languages RDFS [19] and
OWL [20], the W3C standards. The ontology schema of the EPISECC Ontology model is described in
Chapter 4.
The next step is to populate ontology schema with test data describing individuals, the instantiations
of classes. The EPISECC Ontology model will be populated with test data through the Proof of
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |15
www.episecc.eu
Concept (Tasks 6.2 and 6.3). The test data will reflect the EPISECC use case and will be collected
during the field exercise.
The final step is to validate ontology schema. Logical consistency of an ontology schema is checked
by applying reasoning algorithms. The schema is not consistent if reasoning algorithms infer that one
individual belongs to two disjoint classes. Ontology modelling is subjective work of an expert and
there is a need for an objective way to validate the model. The most common approach is to create
test use cases and a list of related competency questions or questions the model should answer
correctly [21]. The EPISECC Ontology model implementation and validation will be done along with
CIS functions design. The validation will include end users and project’s Advisory Board members as
planned within WP6 through the Proof of Concept (Tasks 6.2 and 6.3). The Proof of Concept scenario
describes data exchanged between actors and data needed by responders. Therefore, based on the
Proof of Concept scenario and use cases, a list of competency questions will be built. The ontology
model will be validated through its ability to answer the competency questions.
Ontology development is an iterative process. The first version of ontology is revised and filled with
more details according to validation results, thus the final EPISECC Ontology model will be elaborated
in Task 4.5 “Lessons learned from proof of concept and validation” and delivered in the Deliverable
4.5.
There are many modelling decisions to be made during the ontology development process. Whether
to model a specific distinction (such as human, financial and material resources) as a property value
or as a set of classes? Whether a particular concept is a class in an ontology or an individual instance
of a class? e.g. shell ’water tank truck’ be represented as a class or as an instance of the class
‘vehicle’? The decisions should be based on criteria which solution would work better for the given
tasks, be more extensible and maintainable. Ontology as a model of reality must contain concepts
which reflect this reality. The first version of developed ontology must be evaluated by using it in
applications and discussing it with users and experts. This iterative process will continue through the
entire lifecycle of the ontology.
2.3. Formal languages and tools
RDFS and OWL are W3C standard languages for formal description of RDF database and ontology
schema. The EPISECC Data model and the EPISECC Ontology model are formalised by RDFS and OWL.
Although already described in Deliverable 4.3 “Data model”, the main characteristics of RDFS and
OWL are given briefly as follows.
RDFS and OWL are based on RDF vocabulary that defines Subject, Predicate and Object, which
constitute Triples, the building blocks in RDF database [21].
The RDFS language has a small set of constructs and rules [19]. The constructs are the following:
definitions of classes and hierarchy between classes via sub-classes,
definitions of properties and hierarchy between properties via sub-properties, and
definitions of domain and range describing relationships between classes and properties.
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |16
www.episecc.eu
OWL extends RDFS with many modelling constructs such as: operations over sets (union,
intersection, complement), cardinality, characteristics of properties (inverse, symmetric, reflexive,
functional, transitive), restrictions, etc. [20]. The key construct of OWL is a restriction, which
describes classes by restricting the values allowed for certain properties. A combination of RDF, RDFS
and OWL constructs can be used to model complex relationships between classes, properties and
individuals.
The EPISECC project consortium decided to use open source tools for both data model and ontology
development and implementation. The Protégé Desktop [22] tool is selected for ontology schema
development because it supports RDFS and OWL languages and includes reasoning algorithms used
for ontology schema consistency checking. Protégé Desktop is described in detail in the deliverable
D4.2 “Taxonomy model”.
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |17
www.episecc.eu
3. Ontology model
This chapter explains the conceptual model of the EPISECC Ontology consisting of classes and
properties. Following the methodology described in Chapter 2, the scope of the ontology is defined,
the key concepts are selected, and the ontology model is established and explained. According to the
defined scope, the EPISECC Ontology models the common operational picture (COP) with dynamic
spatial information (or spatio-temporal/4D data) showing the situation in the affected area and the
locations of the resources.
The last section describes references to the DOLCE-Lite upper ontology and two domain ontologies:
GeoSPARQL for spatial concepts and W3C Time for temporal concepts. The EPISECC Ontology model
described in this chapter is the first version of the ontology model since it will be further validated in
WP6 and elaborated in the Deliverable D4.5.
3.1. Ontology scope – the EPISECC use case
The first step in ontology development is definition of its domain and scope.
The EPISECC project has a distinctive universe of discourse defined as ‘a response to a critical event’
which is analysed and described in Deliverable 4.2 “Taxonomy model”. Focus is on the disasters’
relief period, namely first 72 hours after the disaster’s sudden impact or early warning system alert,
and it concerns many concepts like processes, data, organizations, disasters, resources. The EPISECC
project keeps focus on the response to a disaster and that represents a domain of the EPISECC
Ontology model.
To define a scope of ontology, one should define its purpose and tasks the ontology should serve.
Questions the ontology should answer and the users should be defined also. The EPISECC Ontology
model is an application ontology serving emergency response tasks for the EPISECC use case. Thus,
the scope of the EPISECC Ontology model is defined by the following:
tasks defined by the CIS design, and
the EPISECC use case actors and information needed.
According to the EPISECC project proposal, the CIS should include appropriate semantic definitions
by taxonomies and/or ontologies. The proposal has recognised several challenges for informational
interoperability. The following three are the most important for the semantic structures in the CIS:
utilisation of existing and emerging standards,
utilisation of the EPISECC Taxonomy, and
dynamic information sharing.
The above stated challenge of dynamic information sharing is recognized as an important challenge
for informational interoperability as explained in the EPISECC project proposal: “Within this project,
the common information space is not regarded as something static, but more as a dynamic space,
where additional participants can be added spontaneously as required by the mission.”
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |18
www.episecc.eu
The EPISECC project elaborates three semantic structures facilitating semantic interoperability in CIS:
taxonomy, semantic repository and ontology model. The backbone semantic structure is the EPISECC
Taxonomy describing concepts used in ‘a response to a critical event’ and organizing them in
hierarchies and facets (D4.2). The EPISECC Data model provides database structure called Semantic
Repository for the storage of the semantic descriptions: the EPISECC Taxonomy, concepts used in
standards such as CAP and EMSI, and emergency organisations concepts. The EPISECC Semantic
Repository supports the semantic mapping services in CIS (D4.3). The third semantic structure of the
EPISECC project is the EPISECC Ontology model and its main objective is to enhance dynamic
information sharing. The EPISECC Ontology model describes the EPISECC use case as a model of
domain knowledge with references to the upper ontologies what enables dynamic merging of data.
Semantically corresponding concepts are identified and associated and the heterogeneous data can
be integrated. Thus, the task of the EPISECC Ontology model within the CIS design is to ease adding
of additional participants, dynamically and as required by the mission.
The EPISECC use case narrows down the domain of emergency response to a scenario designed for
the Proof of Concept. Deliverable 6.1 “Proof of Concept design” elaborates the scenario and a brief
description of the scenario follows.
The scenario describes the earthquake in Northern Italy followed by a dam break in Austria and
flooding and it covers the real field exercise which will be held in Friuli region in Italy in September
2016. The exercise will be monitored and analysed by the EPISECC consortium as described in D6.1.
The scenario and the field exercise involve the following:
cross-border situation between Italy, Slovenia, Austria and Croatia;
various organisations such as Local Emergency Management Authorities (LEMA), 112 centres,
national civil protection services, humanitarian organisations and first responders;
various tasks such as command and coordination on tactical level, assessment of the
situation, planning and conducting the response (search and rescue, medical service,
evacuation, protection and managing supply).
The scenario is divided in two episodes and use cases as follows.
• Episode 1: Main earthquake and aftershock causing partial collapse of roads and railways
Use-cases:
o Local Emergency Management Authority (LEMA) collects early warnings and
situation assessment info and compiles situation awareness
o Italian government requests help from Austria and Slovenia
o Italian government requests help from the Emergency Response Communications
Centre
• Episode 2: Dam break and flooding
Use-cases:
o Austria informs Slovenia and Italy
o Austria, Slovenia and Italy prepare for the dam break
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |19
www.episecc.eu
The episodes and use cases are described in D6.1 “Proof of Concept design” including descriptions of
the course of events, tools used, information needed and exchanged. The analysis of the use cases
shows that the information that needs to be exchanged compiles the common operational picture.
Thus, the interoperability needs will focus on exchanging data that composes the common
operational picture.
Common operational picture is defined in [23] as follows: “A common operating (operational) picture
is established and maintained by the gathering, collating, synthesizing, and disseminating of incident
information to all appropriate parties involved in an incident. Achieving a common operating picture
allows on−scene and off−scene personnel to have the same information about the incident, including
the availability and location of resources, personnel, and the status of requests for assistance.
Additionally, a common operating picture offers an overview of an incident thereby providing incident
information which enables the Incident Commander (IC), Unified Command (UC), and supporting
agencies and organizations to make effective, consistent, and timely decisions. In order to maintain
situational awareness, communications and incident information must be updated continually.”
Zlatanova et al. [24] model the data of the common operational picture as shown on Figure 4.
Figure 4: The common operational picture data classes (adopted from [24])
Data composing the common operation picture can be classified to static and dynamic data. Static
data exists prior the disaster occurred and it includes topographic maps, critical infrastructure data,
vulnerable objects data, population data, risk maps etc. Dynamic data consists of operational and
situational data and is collected during the emergency response. Operational data describes the
emergency response operations such as locations of rescue teams and other resources that can be
commanded. Situational data describes the situation in affected area such as disaster location,
effected people and properties, meteorological data etc.
The EPISECC use case focuses on the dynamic data and the static data is omitted from further
analysis. Table 1 summarises the exchanged and dynamic data from the scenario’s episodes and use
cases. Where applicable, data description uses the concepts and terms from the EPISECC Taxonomy.
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |20
www.episecc.eu
Table 1: The EPISECC use case data
No. Data class Data description
1 Operational data
Emergency response process (operation)
type of process (rescue, evacuation, etc.)
status of process (planned, in progress, etc.)
priority of process (high, medium, low)
capability of process (yes, no, partially)
responsible organization
2 Operational data
Resource needed/deployed
type of resource (human, material, machinery, vehicle, etc.)
capability of resource (yes, no, partially)
status of resource (available, needed, in use, etc.)
quantity of resource
3 Situational data
Disaster type of disaster (earthquake, flood, etc.)
progress of disaster (in progress, stopped, etc.)
cause of disaster (nature, humans, technology)
earthquake epicentre
earthquake magnitude
water reservoir water level
water reservoir risk level
4 Situational data
Affected area size of affected area
number of affected inhabitants
type of affected critical infrastructure
status of affected critical infrastructure
type of affected vulnerable objects
status of affected vulnerable objects
5 Situational data
Damages type of damaged objects
estimation of damages (number of objects, financial value, etc.)
6 Situational data
Casualties type of casualties
status of casualties
number of casualties
7 Situational data
Weather (actual and forecast)
temperature
humidity/rain
wind speed
wind direction
pressure
All the EPISECC use case data is spatio-temporal data or 4D data which means that includes
geolocation and reference to time. Geolocation could be represented with a point, line, polygon or a
collection of the points, lines and polygons. Different systems for the geolocating could be used such
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |21
www.episecc.eu
as the coordinates in one of the reference coordinate systems (e.g. WGS 84), place names,
addresses, the grid/sector codes defined by the emergency and risk plans, etc. Time reference could
include the time interval or the instant in time and be described by year, month, week, day, hour,
minute and second.
Some examples of the spatio-temporal data in the common operational picture follows. The
emergency response process could be geolocated by the point or polygon showing the operation’s
scene. As the process evolve in time, its scene changes and the polygon representing operation’s
scene changes its geometry in time. The location of the rescue team could be represented by the
moving point, a flood by spreading polygon etc.
Figure 5 shows the EPISECC Ontology model as spatio-temporal ontology that models data for the
EPISECC use case. This model should answer on question: Who, where and when is conducting what
operations?
Figure 5: The EPISECC Ontology model’s scope
To conclude, the scope of the EPISECC Ontology model encompasses the situational and operational
data exchanged among the organisations in the EPISECC use case. The EPISECC Ontology models the
common operational picture with dynamic spatial information (or spatio-temporal/4D data) showing
the situation in affected area and the locations of the resources.
3.2. Ontology conceptual model
The ontology development starts from reusing already developed ontologies and a brief description
of existing and relevant ontologies is given in the next section. Forthcoming sections explain scope
and key concepts of the ontology, and the last section gives references to the upper and domain
ontologies.
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |22
www.episecc.eu
3.2.1. An overview of the existing ontologies
There are libraries of reusable ontologies (or design patterns) such as Ontology design pattern Web
portal [18], but there is still no reference ontology specifically intended for the emergency or disaster
management. One of the reasons could be that the domain is quite complex and it could be seen as a
composition of several human domains such as disasters, organisations, people, events, workflows,
resources, infrastructures, policy, etc.
The importance of having a common understanding in the emergency management has been
recognised and different vocabularies have been created, but further development towards
reference ontology model has just started. An overview of efforts is given in [25] and it shows that
several attempts are made but they are limited either to one disaster type, to one infrastructure
system, to the method of development etc. Another overview of existing ontologies for disaster
management is elaborated by the FP7 project DRIVER and is given in [7]. The authors identified a
total of 11 subject areas represented in contemporary crisis information systems including critical
infrastructures (20%), resource management (13%), decision support (13.3%), situation awareness
(13.3%), response coordination (13.3%), command and control (6.7%) and other types (26.7%) such
as humanitarian response and relief. But, there is still no one model attempting to cover the
emergency and disaster management.
The FP7 project DISASTER developed the EMERGEL ontology [26] based on the Simple Knowledge
Organisation System (SKOS) model. The EMERGEL interprets a disaster as a kind of event. Three
ontology’s parts are developed as follows:
core module, referenced to the classes from the upper-level ontology DOLCE+DnS Ultralite;
transversal modules, used for space-time representation; and
vertical modules, used for representation of standard classifications of map symbol sets,
chemical substances, transport codes, etc.
The World Wide Web Consortium (W3C) published the report “Emergency Information
Interoperability Frameworks” in 2009 [8]. The undertaken work included several approaches: from
building the conceptual mind map, through using Wordle for generating the most used terms on the
Web, to studying the existing standards, information systems and research projects such as Tactical
Situational Object (TSO), OCHA, Sahana and the FP6 project ORCHESTRA. Still, the report describes
only candidate components for the emergency management ontology based on the common use
cases.
For the EPISECC Ontology model development, the above mentioned work is analysed and none of
the proposed ontologies could be reused. There are reference ontologies to be used as domain and
upper-level ontologies. Looking at the emergency management sub-domains, two domain ontologies
are selected for the development of the EPISECC Ontology model:
GeoSPARQL as domain ontology for spatial data [15], and
W3C Time as domain ontology for temporal data [16].
As a reference upper level ontology, the DOLCE-Lite ontology is selected [14].
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |23
www.episecc.eu
3.2.2. Ontology model’s elements
Ontology model is a conceptual model of a domain and it could be decomposed in three main
elements as follows [27]:
classes,
properties, and
individuals.
Classes describe concepts in the domain and they classify individuals or other classes. Individuals are
instantiations of the classes and are represented by data items. Properties describe relations and are
expressed as verbs, e.g. the verb ‘uses’ describes relation between an organisation and a resource.
Properties could be established between the following:
classes, e.g. ‘Vehicle’ is subclass of ‘Resource’;
individuals and classes, e.g. ’Water tank id 345’ is of type ‘Vehicle’;
individuals, e.g. ‘Water tank id 345’ belongs to ‘Fire brigade of the City of Split’.
Further more, OWL language in ontology schemas use two types of properties:
object properties for relating individuals to individuals, and
data type properties for relating individuals to data values.
Class could have subclasses and property could have subproperties, thus constituting the hierarchies
of the classes and properties. Although both the object-oriented and the ontology models use
classes, properties and inheritance, there are significant differences in these two paradigms. Some of
the ontology model’s characteristics that differ from object-oriented are the following:
properties exist independently of classes and can form a hierarchy,
an individual can belong to more then one class,
classes can be defined by set-theoretic constructs, and
inheritance propagates only on class membership up to the class hierarchy tree.
Figure 6 shows three main elements of the ontology model and the relations between them.
Figure 6: Ontology model’s main elements and their relations (adopted from [27])
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |24
www.episecc.eu
Individual is an instance of a class. Class has a property. Individual is a subject and an object of a
property. Property can have a class for its domain and range e.g. the class ‘Vehicle’ can have the
property ‘belongs to’. The property ‘belongs to’ can have a range: the class ‘Organisation’; and a
domain : the class ‘Resources’. ’Water tank id 345’ is an instance of the class ‘Vehicle’ and is subject
of the property ‘belongs to’.
Properties could have attributes such as cardinality, value type, inverse, etc. Classes could be defined
by property restriction. Property restriction describes the class by restricting the values allowed for
certain properties or putting the constraints on individuals for being a member of the class e.g. class
‘Blocked emergency ingress’ are emergency routes with at least one blocked road along the route.
Besides classes and properties, a conceptual ontology model can define axioms and rules for data
interpretation and reasoning. Axioms describe additional rules which exist between concepts and
properties. Based on the ontology model elements and by use of inferencing, the reasoning
algorithms can discover new relationships which are not explicitly stated in the model, and thus
extend semantic description of the domain. Reasoning algorithms use rules of logic to infer new
statements from available knowledge stored in the ontology.
Ontology conceptual model is expressed as one or more directed graphs. Figure 7a shows the
elements of directed graph as follows:
subject representing class or individual,
predicate representing property, and
object representing class or individual.
Figure 7: Example of the directed graph: generic (a) and specific one (b)
Figure 7b shows an example of the directed graph representing that the individual ’Water tank id
345’ belongs to the individual ‘Fire brigade of the City of Split’. Additional descriptions such as
domains and ranges are documented by a text coupled with the directed graphs.
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |25
www.episecc.eu
3.2.3. Key concepts and their relations
Knowing what is the scope of ontology, a list of key concepts can be defined. For the EPISECC
Ontology model, a subset of concepts of the EPISECC Taxonomy (D4.2) is selected according to the
scope of the EPISECC use case. The ontology model does not completely follow the EPISECC
Taxonomy classification because the ontology model is not a classification system but it should
describe the reality by modelling classes and relationships between them. E.g. for the EPISECC Use
case, the concept ‘Process’ is not a subclass of the EPISECC Taxonomy concept ‘Resource’ but the
main class of the ontology model. Besides the concepts and their hierarchy, the relationships or
properties between classes are derived. Figure 8 shows the main concepts modelled as ontology
classes (shown as ovals) and properties (shown as links with arrows).
PROCESS
DISASTER
ORGANISATION
RESOURCE
COMMON OPERATIONAL
PICTURE
invokes
executed by
uses
uses
has available
Figure 8: Directed graph of the main classes and properties
Five main classes are as follows:
Disaster
Any situation which has or may have a severe impact on people, the environment, or
property, including cultural heritage (D4.2).
Process
Process is a set of actions aiming for a certain result, executed by an organisation during a
response to a critical event (D4.2).
Resource
Assets an organisation has available for the response to a critical event (D4.2).
Organisation
An organisation is a unit established to meet goals related to disaster management. It is
structured along its management, which defines the relationships between responsibilities,
tasks and its structure (D4.2).
Common operational picture
A single identical display of relevant information shared by more than one command [28].
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |26
www.episecc.eu
The main classes are related by the properties: ‘invokes’, ‘uses’, ‘executed by’ and ‘has available’.
Class ‘Disaster’ is further elaborated as shown on Figure 9. Subclasses of ‘Disaster’ follow the EPISECC
Taxonomy classification of the facet ‘Disaster type’ which has 18 subclasses on the first level of the
classification (e.g. ‘Avalanche’, ‘Earthquake’, etc.) and 26 subclasses on the second level of the
classification (e.g. ‘Flash flood’, ‘Urban flood’, etc.). Their definitions are given in the Annex of the
Deliverable 4.2 “Taxonomy model”.
AVALANCHE
DISASTER DISASTER CAUSE
FLASH FLOOD
EARTHQUAKE FLOOD TSUNAMI
URBAN FLOOD
... ...
...
DISASTER PROGRESS
EPICENTRE
MAGNITUDE
has e
picen
tre
has
mag
nitu
de
has cause
has pro
gress
WATER RESERVOIR RISK LEVEL
WATER LEVEL
has risk level
has wa
ter lev
el
DISASTER OBJECT has affected
object
PROPERTYPEOPLE ...
...
subclass of ... classes not shown due to the limited space
Figure 9: Directed graph of the class Disaster
Due to the limited paper space, only few ‘Disaster’ subclasses are shown on Figure 9. Class ‘Disaster
object’ follows the EPISECC Taxonomy classification while the classes ‘Disaster progress’ and ‘Disaster
cause’ are not further classified and they could contain individuals (or data items) such as
‘technological cause’ or ‘disaster in progress’, respectively. Additional classes are introduced to the
ontology model beside the ones from the EPISECC Taxonomy. ‘Water reservoir’ is introduced as a
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |27
www.episecc.eu
type of property affected by a disaster and thus it represents a subclass of the EPISECC Taxonomy
class ‘Property’. ‘Water level’, ‘Risk level’, ‘Magnitude’ and ‘Epicentre’ are classes containing
individuals (or data items) coming from early warning systems and Earth observations systems and
could be considered as subclasses of the EPISECC Taxonomy class ‘Data’.
Concepts describing observations and measurements are complex concepts and they could be
further elaborated by use of domain ontologies such as Semantic Web for Earth and Environmental
Terminology (SWEET) [29] or Semantic Sensor Network (SSN) [30].
Figure 9 shows all properties between the classes with the verb explaining their meanings (shown as
links with arrows). Link ending with arrow without fill colour represents ‘is-a’ or ‘subclass of’
property.
Class ‘Process’ is elaborated on the Figure 10. Subclasses of ‘Process’ follow the EPISECC Taxonomy
classification of the facet ‘Process type’ which has 5 subclasses on the first level of the classification
(e.g. ‘Command & control’, ‘Physical response’, etc.), 22 subclasses on the second level of the
classification (e.g. ‘Decontamination’, ‘Search for people’, etc.), and 2 subclasses on the third level of
the classification. Their definitions are given in the Annex of the Deliverable 4.2 “Taxonomy model”.
Due to the limited paper space, only few ‘Process’ subclasses are shown on the Figure 8. The classes
‘Process status’, ‘Priority’ and ‘Capability’ are not further classified and they could contain individuals
(or data items) such as ‘planned’ or ‘high priority’ or ‘partially capable’, respectively.
Figure 10. shows all properties between the classes with the verb explaining their meanings.
COMMAND & CONTROL
PROCESS CAPABILITY
DECONTA-MINATION
INTERACTION WITH PEOPLE
PHYSICAL RESPONSE
RESOURCES MANAGEMENT
SEARCH FOR PEOPLE
... ...
...
PRIORITY
has capability
has pri
ority
PROCESS STATUS
has s
tatu
s
ORGANISATION executed by
subclass of
... classes not shown due to the limited space
Figure 10: Directed graph of the class Process
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |28
www.episecc.eu
Class ‘Resource’ is elaborated on the Figure 11. Subclasses of ‘Resource’ follow the EPISECC
Taxonomy classification of the facet ‘Resource type’ which has 5 subclasses on the first level of the
classification (e.g. ‘Animal’, ‘Physical’, etc.), 11 subclasses on the second level of the classification
(e.g. ‘Decontamination’, ‘Search for people’, etc.), 31 subclasses on the third level and additional
classes on the fourth and fifth levels of the classification.
Subclass ‘Process’ is omitted from the classification of the ‘Resource’ class because the ‘Process’ is
modelled as the main class in the EPISECC Ontology model.
The classes ‘Resource status’ and ‘Capability’ are not further classified and they could contain
individuals (or data items) such as ‘in use’ or ‘partially capable’, respectively.
Class ‘Quantity’ is also not further classified and it could contain data items such as numbers
representing quantities. The concept of quantity is a complex concept including measuring units etc
and it could be further elaborated by use of domain ontologies such as Quantities, Units, Dimensions
and Data Types Ontologies (QUDT) [31].
Figure 11. shows all properties between the classes with the verb explaining their meanings.
ANIMAL
RESOURCE CAPABILITY
CIVIL PROTECTION MEMBER
POLICEMAN...
QUANTITY
has capability
has qu
antity
RESOURCE STATUS
has s
tatu
s
PROCESS uses
subclass of
... classes not shown due to the limited space
FINANCIAL HUMAN PHYSICALINSTITUTIONAL
Figure 11: Directed graph of the class Resource
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |29
www.episecc.eu
Class ‘Organisation’ is elaborated on the Figure 12. Subclasses of ‘Organisation’ follow two EPISECC
Taxonomy facets classifications. Facet ‘Organisation type’ has 3 subclasses: ‘Governmental’, ‘Non-
governmental’ and ‘Private’. Facet ‘Specialisation’ has 5 subclasses (e.g. ‘Civil protection’, ‘Emergency
medical assistance’, etc.). The classes ‘Operational scope’ and ‘Spatial scope’ are not further
classified and they could contain individuals (or data items) such as ‘tactical’ or ‘international’,
respectively.
Figure 12. shows all properties between the classes with the verb explaining their meanings.
CIVIL PROTECTION
ORGANISATION SPATIAL SCOPE
OPERATIONAL SCOPE
has spatial scope
has op
eratio
nal sco
pe
PROCESS executed by
EMERGENCY MEDICAL
ASSISTANCE
HUMANITARIAN ASSISTANCE
SEARCH AND RESCUE
OPERATIONS
RECOVERY OF CULTURAL HERITAGE
GOVERNMENTALNON-
GOVERNMENTALPRIVATE
subclass of ... classes not showndue to the limited space
Figure 12: Directed graph of the class Organisation
Class ‘Common operational picture’ is elaborated on the Figure 13. ‘Common operational picture’
consists of ‘Static data’ and ‘Dynamic data’ according to the data model explained in Section 3.1.
‘Static data’ exists prior the disaster occurred such as topographic maps, population data, evacuation
routes, etc. and it is omitted from further analysis.
‘Dynamic data’ has two subclasses: ‘Situational data’ and ‘Operational data’.
‘Situational data’ describes situation in the affected area including disaster location, affected people
and properties, meteorological data, etc. and is further classified according to the ‘Data set content’
facet of the EPISECC Taxonomy. ‘Data set content’ facet contains 17 subclasses on the first level of
the classification and 7 subclasses on the second and third level (e.g. ‘Casualties’, Weather forecast’
etc.).
‘Operational data’ describes the emergency response operations such as the locations of rescue
teams and other resources that can be commanded.
Figure 13 shows the main classes ‘Process’ and ‘Resource’ as subclasses of the ‘Operational data’
which means that data describing processes and resources are considered as operational data. The
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |30
www.episecc.eu
main class ‘Disaster’ is subclass of ‘Situational data’ what means that data describing a disaster is
considered as situational data.
Figure 13 shows all properties between the classes with the verb explaining their meanings.
COMMON OPERATIONAL
PICTURE
STATIC DATA DYNAMIC DATA
uses
SITUATIONAL DATA
OPERATIONAL DATA
PROCESSRESOURCEAFFECTED
PEOPLE DATADAMAGE DATA
WEATHER FORECAST
...
containscont
ains
subclass of
... classes not showndue to the limited space
DISASTER
CASUALTIES HOMELESS MISSING TOTAL AFFECTED
Figure 13: Directed graph of the class Common operational picture
Additional to the classes and object properties shown on Figure 13, the following data type
properties of situational and operational data are used for detailed description of the situation:
‘has size’ (e.g. for ‘Affected area’),
‘has number’ (e.g. for ‘Casualties’), and
‘has value’ (e.g. for ‘Damages data’).
Weather is described by two classes: ‘Weather data’ and ‘Weather forecast’. As weather is a complex
concept, these two classes could be further elaborated by use of domain ontologies such as Semantic
Web for Earth and Environmental Terminology (SWEET) [29] or Semantic Sensor Network (SSN) [30].
The above sections use directed graphs to present the conceptual model of the EPISECC Ontology
model with its classes, subclasses and properties. The presented model completely covers the
EPISECC use case data listed in Table 1. The next section gives the EPISECC Ontology model
references to the selected domain and upper ontologies.
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |31
www.episecc.eu
3.2.4. References to domain and upper ontologies
For the EPISECC Ontology model development, two domain ontologies are selected and referenced:
GeoSPARQL as domain ontology for spatial data [15], and
W3C Time as domain ontology for temporal data [16].
As a reference upper level ontology, the DOLCE-Lite ontology [14] is selected and referenced.
The following sections explain the extension of the EPISECC Ontology model towards spatio-temporal
data and its references to the domain and upper level ontologies.
3.2.4.1. References to domain ontologies
The focus of the EPISECC Ontology is given to the modelling of spatio-temporal or 4D data showing
situational and operational data e.g. locations of the most urgent resources requested by the first
responders. Spatio-temporal data includes geolocation and reference to time. Geolocation could be
represented with different geometry entities such as points or surfaces, and different geolocating
systems such as WGS 84, the national coordinate system or addresses. Time reference could include
the time interval or the instant in time described by any time unit such as a day or a day plus hours
and minutes e.g. the emergency response process could be geolocated by the point or polygon
showing the operation’s scene. As the process evolve in time, the polygon representing the scene
changes its shape and position in time. Another example is the location of the rescue team which
could be represented by the moving point, a flood could be represented by spreading polygon etc.
Representation of 4D or dynamic features asks for mechanisms allowing representation of the
properties varying in space and time. Although the reference ontologies exist for the spatial and
temporal data, putting these two concepts together requires solution for a new concept: evolution of
concepts in time and space.
Several approaches for the 4D data ontology modelling have been reported, which a briefly explained
in [32]. Welty and Fikes in [33] have demonstrated that 4D Fluent model is the most appropriate one
for ontology modelling of 4D data using the OWL language. The 4D Fluent model maintains that all
entities are perdurants. The DOLCE ontology defines perdurants as follows [14]: “Perdurants (also
called occurrents) are entities that ‘happen in time’, they extend in time by accumulating different
‘temporal parts’, so that, at any time ‘t’ at which they exist, only their temporal parts at ‘t’ are
present.” Thus, perdurants have temporal parts and can be thought as ‘space-time worms’ whose
temporal parts are slices of the worm.
To represent entities varying also in space, Baučić in [34] has extended 4D Fluent model with spatio-
temporal parts of ‘space-time worms’. The spatio-temporal part represents 4D entity in the certain
moment in time with its location and geometry. E.g. flood can have its spatio-temporal part
consisting of polygon showing flooded area and time record stating when the flood had that extent,
as well as stating other flood characteristic valid for the same instant of time. Figure 14 shows the
model for common operational picture proposed in [34]. Ovals coloured in grey represent main
classes of the common operational picture and white ovals are classes of the reference domain
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |32
www.episecc.eu
ontologies. For the representation of dynamic or 4D data, the class ’Spatio-temporal part’ is
introduced and related with the property ‘is spatio-temporal part of’ to the class ‘Dynamic data’.
Figure 14 shows property ‘has spatiotemporal part’ relating ‘Dynamic data’ and ’Spatio-temporal
part’ and it is the inverse property of the ‘is spatio-temporal part of’.
COMMON OPERATIONAL
PICTURE
containscon
tain
s
DYNAMIC DATASTATIC DATASPATIOTEMPORAL
PART
has spatio-temporal part
has geometry
SPATIAL OBJECT
GEOMETRYFEATURE
Spatial relations(contains,intersects, within, touches, etc.)
Temporal relations(precedes, during,
started by, overlaps, etc.)
is inside
TEMPORAL ENTITY
INTERVALINSTANThas beginning
has end
SERIALIZATIONhas serialization
WKT GMLPOINT CURVE SURFACEGEOMTRY
COLLECTION
has serialization
XSD
DATA
TIME
has instant
Geospatial ontology(OGC GeoSPARQL, 2012)
Time ontology(W3C, 2006)
subclass of
Figure 14: Directed graph of common operational picture containing 4D data [34]
The selected domain ontologies further explain the spatial and temporal concepts. Thus, ’Spatio-
temporal part’ is referenced as a sublass of the class ‘Feature’ of the GeoSPARQL ontology and also
of the class ‘Temporal entity’ of the W3C Time ontology. Class ‘Static data’ is referenced as a sublass
only to the class ‘Feature’ because it does not contain temporal properties. A short description of
GeoSPARQL ontology and W3C Time ontology follows.
GeoSPARQL is Open Geospatial Consortium (OGC) standard [15]. This standard defines a core
RDF/OWL vocabulary for geographic information based on the General Feature Model [35], Simple
Features [36], Feature Geometry [37] and SQL for Multi Media [38]. Also, it includes a set of SPARQL
Protocol and RDF Query Language (SPARQL) extension functions and a set of Rule Interchange
Format (RIF) rules. GeoSPARQL defines a core set of classes, properties and datatypes that can be
used to construct query patterns. Figure 15 shows three main classes as follows:
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |33
www.episecc.eu
Spatial Object
This class represents everything that can have a spatial representation. It is superclass of
feature and geometry.
Feature
This class represents the top-level feature type and it is superclass of all feature types.
Geometry
The class represents the top-level geometry type and it is superclass of all geometry types.
has geometry
SPATIAL OBJECT
GEOMETRYFEATURE
Spatial relations(contains,intersects, within, touches, etc.)
Figure 15: GeoSPARQL directed graph [15]
The main property is ‘has geometry’ which links a feature with a geometry that represents its spatial
extent. GeoSPARQL does not restrict the cardinality of the ‘has geometry’ property. It is thus possible
for a feature to have many associated geometries. This standard includes properties describing
spatial relations of several models such as Simple Features Relation Family, Egenhofer Relation
Family and RCC8 Relation Family (e.g. ‘equal’, ‘disjoint’, ‘intersects’, ‘crosses’, etc.).
Figure 14 shows ‘has serialization’ property that connects a geometry object with its text-based
serialization. GeoSPARQL includes two serialization types which strongly affects the geometry
conceptualization. The serializations are as follows.
Well-known Text (WKT)
WKT aligns the geometry types with Simple Features standard [36].
Geography Markup Language (GML)
GML aligns the geometry types with geometry types of ISO 19107 Spatial Schema [37] and
uses Geography Markup Language Encoding Standard [39].
Well-known Text serialization includes definition of the coordinate reference system for the
geometry. It is expressed by URI and the OGC maintains a set of coordinate reference system URIs
under the http://www.opengis.net/def/crs/ namespace e.g. http://www.opengis.net/def/crs/OGC/
1.3/CRS84 denotes WGS 84 longitude-latitude. Additional to the coordinate reference systems, first
responders could use other systems for geolocations such as place names, addresses, the grid/sector
codes defined by the emergency and risk plans, etc. Therefore, the EPISECC Ontology model is
extended with the classes representing various types of the geolocation systems.
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |34
www.episecc.eu
Figure 16 shows three types of geolocation systems: ‘Address’, ‘Location ISO 125’ and ‘Emergency
sector’. Property ‘has location’ relates the ‘Spatiotemporal part’ with its geolocation system. Each
geolocation system is then related to the ‘Feature’ class of GeoSPARQL as a subclass.
DYNAMIC DATA
SPATIOTEMPORAL PART
LOCATION ISO 19125
ADDRESSEMERGENCY
SECTOR
has l
ocat
ion
has location
has lo
cation
... ...
Temporal relations(precedes, during,
started by, overlaps, etc.)
is inside
TEMPORAL ENTITY
INTERVALINSTANT
has beginning
has end
has instanthas geometry
SPATIAL OBJECT
GEOMTERYFEATURE
Spatial relations(contains,intersects, within, touches, etc.)
has spatio-temporal part
subclass of
... classes not showndue to the limited space
Figure 16: Subclasses of the class ‘Spatiotemporal part’ (adopted from [34])
W3C Time ontology is W3C standard [16]. This standard presents an ontology of temporal concepts
including temporal relations among instants and intervals, together with information about durations
and date-time information. Figure 17 shows three main classes as follows:
Temporal entity
This class represents everything that can have a temporal representation. It is superclass of
instant and interval.
Instant
This class represents a precise moment of time.
Interval
This class represents a period of time.
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |35
www.episecc.eu
The standard further explains that “Intervals are, intuitively, things with extent and instants are,
intuitively, point-like in that they have no interior points. It is generally safe to think of an instant as
an interval with zero length, where the beginning and end are the same.”
Temporal relations(precedes, during,
started by, overlaps, etc.)
is inside
TEMPORAL ENTITY
INTERVALINSTANT
has beginning
has end
has instant
Figure 17: W3C Time directed graph [16]
Four main properties are as follows:
‘is inside’ describes relation between an instant and an interval where instant is placed inside
interval but it is not intended to include beginnings and ends of intervals;
‘has instant’ describes relation between an instant and a temporal entity where instant
represents a unique moment of time;
‘has beginning’ describes relation between an instant and a temporal entity where instant
represents a unique and beginning point of the entity;
‘has end’ ‘describes relation between instant and temporal entity where instant represents a
unique and end point of the entity.
Figure 14 shows ‘has serialization’ property that connects a instant with its text-based serialization.
W3C Time defines XSD Date Time serialization that is XML representation of date and time.
Temporal relations are based on a calculus of binary relations on intervals (e.g., meets, overlaps) for
representing qualitative temporal information [40]. W3C Time ontology provides the interval
relations such as ‘equals’, ‘before’, ‘meets’, ‘overlaps’, ‘starts’, during’, etc.
3.2.4.2. References to upper level ontology
As a reference upper level ontology, the DOLCE-Lite ontology is selected [14]. DOLCE is an acronym
for Descriptive Ontology for Linguistic and Cognitive Engineering and it is clear that DOLCE ontology
aims at capturing the ontological categories underlying natural language and human common sense.
Suffix ‘Lite’ stands for encoding version with most of predicates formalised in DOLCE.
Figure 18 shows DOLCE basic categories of a particular. The distinction between endurants (also
called continuants) and perdurants (also called occurrents) is related to their behaviour in time.
Endurants are wholly present at any time. Perdurants extend in time by accumulating different
temporal parts, so that, at any time they are only partially present, in the sense that some of their
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |36
www.episecc.eu
proper temporal parts (e.g., their previous or future phases) may be not present [14]. Qualities are
the basic entities we can perceive or measure: shapes, colours, sizes, sounds, weights, lengths, etc.
Abstract entities do not have spatial nor temporal qualities, and they are not qualities themselves.
Figure 18: DOLCE basic categories of particulars [14]
Figure 19 shows the main classes of the EPISECC Ontology model extended with 4D data
representation and referenced to the DOLCE-Lite ontology. Grey ovals represent DOLCE-Lite classes
and white ovals represent the EPISECC Ontology classes. The model corresponds to the one
described in [34] and is accommodated to the main classes of the EPISECC Ontology model. Due to
the limited paper space only main classes are shown on Figure 19.
References to domain and upper level ontologies enable the EPISECC Ontology model to be merged
with other semantic structures using the same reference ontologies. The semantic is expressed
formally and can be used to negotiate meaning, either for enabling cooperation between multiple
artificial agents or between artificial agents and humans.
There are other domain ontologies which could be referenced such as Semantic Web for Earth and
Environmental Terminology (SWEET) [29], Semantic Sensor Network (SSN) [30] and Quantities, Units,
Dimensions and Data Types Ontologies (QUDT) [31]. These ontologies include large number of
concepts for Earth observations, measurements, units etc. Because the focus of the EPISECC
Ontology model was on 4D data, they are omitted from the further study.
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |37
www.episecc.eu
PROCESS
DISASTER
ORGANISATION
RESOURCE
COMMON OPERATIONAL
PICTURE
invo
kes
executedBy
uses
use
s
containscont
ains
DYNAMICDATA
STATICDATA
SPATIOTEMPORAL PART
is spatio-temporal part of
hasGeometry
SPATIALOBJECT
GEOMETRYFEATURE
Spatial relations
(contains,
intersects, within,
touches, etc.)
Geospatial ontology(OGC GeoSPARQL, 2012)
Temporal relations
(precedes,
during, started by,
overlaps, etc.)
Time ontology(W3C, 2006)
isInside
TEMPORALENTITY
INTERVALINSTANT
hasB
eginn
ing
hasE
nd
STATE
PROCESS
AGENTIVE PHYSICAL OBJECT
NON – AGENTIVE PHYSICAL OBJECT
INFORMATION OBJECT
SPATIAL – LOCATION_q
TEMPORAL – LOCATION_q
PERDURANT
ENDURANT
ENDURANT
HUMAN
PHYSICAL
NON – AGENTIVE SOCIAL OBJECT
QUALITY
subclassOf
has
av
aila
ble
Figure 19: Directed graph of main EPISECC classes and with references to DOLCE-Lite, GeoSPARQL and W3C Time
ontology classes (adopted from [34])
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |38
www.episecc.eu
4. Ontology schema
This chapter describes the first implementation and validation of the EPISECC Ontology model. Once
the conceptual model is developed and upper and domain ontologies referenced, the ontology
schema can be created. Ontology schema is a formalisation of the conceptual model. Software-
interpretable constructs are defined by formal languages RDFS [19] and OWL [20]. The EPISECC
Ontology schema is created with Protégé Desktop software already used for the EPISECC Taxonomy
and the EPISECC Semantic repository. Protégé Desktop description is given in the Deliverable 4.2.
Advanced functions and plug-ins are used for the input of EPISECC Ontology model classes and
properties, search and reasoning. The EPISECC Ontology schema is stored in an OWL file locally on
the computer in RDF/XML syntax and an excerpt is given in the Annex. The first validation includes
testing of logical consistency by applying reasoning algorithms.
The second implementation and validation will be done through the Proof of Concept and validation
and described in the Deliverable 6.3. The final ontology schema will be elaborated in Task 4.5.
“Lessons learned from proof of concept and validation” and delivered in the Deliverable 4.5.
4.1. Implementation in Protégé Desktop
The EPISECC Ontology model classes and properties are inserted to the OWL model. The following
figures illustrates the Protégé Desktop interfaces.
Figure 20 shows imported domain ontologies GeoSPARQL and W3C Time. Their RDF files are
downloaded from the web and imported into the OWL file of the EPISECC Ontology schema. These
ontologies already contain other ontologies and their prefixes so they are indirectly imported too (e.g
GML and Simple Feature ontology).
Figure 20: Protégé Desktop interface showing imported ontologies
http://www.episecc.eu/
-
EPISECC_WP4_D4 4_Deliverable_Report.docx 119 |39
www.episecc.eu
Figure 21 shows two class hierarchies. Left hierarchy shows the first level of classes of the EPISECC
Ontology schema. Classes written in bold letters are defined by the EPISECC Ontology conceptual
model and classes written in regular letters are defined by GeoSPARQL and W3C Time ontologies e.g.
class ‘Spatiotemporal part’ is the EPISECC Ontology conceptual model class, ‘Spatial object’ is the
GeoSPARQL class and ‘Temporal entity’ is the W3C Time class. Right side of the Figure 21 shows the
class ‘Common