4D-SETL

A Semantic Data Integration Framework

Sergio de Cesare

, George Foy

and Mark Lycett

College of Business, Arts and Social Science, Brunel University London, Uxbridge, U.K.

College of Engineering, Design and Physical Sciences, Brunel University London, Uxbridge, U.K.

Keywords: Foundational Ontology, Perdurantist, 4D, Semantic Data Integration, Modelling, Graph Databases,

Integration Frameworks.

Abstract: Although successfully employed as the foundation for a number of large-scale government and energy

industry projects, foundational ontologies have not been widely adopted within mainstream Enterprise

Systems (ES) data integration practice. However, as the closed-worlds of ES are opened to Internet scale

data sources, there is an emerging need to better understand the semantics of such data and how they can be

integrated. Foundational ontologies can help establish this understanding and therefore, there is a need to

investigate how such ontologies can be applied to underpin practical ES integration solutions. This paper

describes research undertaken to assess the effectiveness of such an approach through the development and

application of the 4D-Semantic Extract Transform Load (4D-SETL) framework. 4D-SETL was employed to

integrate a number of large scale datasets and to persist the resultant ontology within a prototype warehouse

based on a graph database. The advantages of the approach included the ability to combine foundational,

domain and instance level ontological objects within a single coherent system. Furthermore, the approach

provided a clear means of establishing and maintaining the identity of domain objects as their constituent

spatiotemporal parts unfolded over time, enabling process and static data to be combined within a single

model.

1 INTRODUCTION

An enterprise may acquire data from many sources

in many different forms (Zikopoulos and Eaton,

2011). Key considerations in integrating such data

include dealing with the diversity of representation

and the interpretation of the inherent explicit and

implicit semantics. The latter of these considerations

is particularly important in the context of ES

integration as, if left unrecognised, it can lead to the

things of importance (e.g., domain objects and their

relationships), their nuances and the state of affairs

they represent being misinterpreted (Lycett, 2013).

These considerations are well recognised within

database integration projects (Arsanjani, 2002;

Campbell and Shapiro, 1995; Sheth and Larson,

1990).

Ontology has emerged as a promising way of

dealing with such diversity, however many popular

domain ontologies have no grounding in a consistent

foundational view of reality (Cregan, 2007) and

therefore can add further diversity. A foundational

ontology can be employed to provide this ‘grounded’

view of reality and thus provide an explicit theory and

a common reference through which to interpret,

model and thus integrate data. Foundational ontology

“defines a range of top-level domain-independent

ontological categories, which form a general

foundation for more elaborated domain-specific

ontologies” (Guizzardi et al. 2008). From a

philosophical perspective, foundational ontologies

provide the criteria for ontological commitments –

statements on the things believed to exist within the

context of a particular theory (Bricker, 2014). Several

foundational ontologies currently exist (Gangemi et

al., 2002; Grenon and Smith 2004; Partridge 2005;

Guizzardi, et al., 2008; Herre 2010) which differ in

the ontological commitments they make but,

importantly, there is little existing work that examines

their suitability as an ultimate ‘mediating layer’

within a practical data integration context.

Here, we employ a 4D foundational ontology as

a means of dealing with the diversity of

representation and semantics within acquired data.

We do this in the context of a semantic Extract-

Transform-Load framework (called 4D-SETL from

Cesare, S., Foy, G. and Lycett, M.

4D-SETL - A Semantic Data Integration Framework.

In Proceedings of the 18th International Conference on Enterprise Information Systems (ICEIS 2016) - Volume 1, pages 127-134

ISBN: 978-989-758-187-8

127

this point) that uses a 4D foundational ontology to

harmonise data, then generates a graph database that

accords with the semantic commitments made by the

ontology. We examine the effectiveness of the

framework by applying it to semantically interpret

and integrate a number of large-scale datasets and to

instantiate a data warehouse based on a graph

database to persist the resultant ontology. In doing

this, the paper is structured as follows. Section 1

outlines the problem of variety in terms of the

semantic heterogeneity that exists within systems

modelling and foundational ontologies and also

identifies a number of the weaknesses in current

integration approaches. Section 2 describes the core

categories and foundational patterns of the BORO

foundational ontology. Section 3 introduces the 4D-

SETL framework. Section 4 provides details of the

experimental dataset integrated. Section 5 details the

outcomes and limitations.

1.1 Semantic Data Integration

Data integration is problematic on several counts.

Firstly, people perceive and conceptualise reality in

different ways. Even when a set of models is

developed by the same individual, they can make

different (and sometimes arbitrary) choices about the

same reality at different times and in different

contexts (Kent, 1978). Secondly, in the course of

modelling reality, a designer may confuse what is

being represented with the representation itself

(Partridge et al., 2013). Thirdly, different structures

and restrictions are introduced by heterogeneous

modelling methods and languages (e.g., Entity-

Relationship, OWL etc.). Fourthly, it is common

practice to develop a number of models in systems

development – conceptual, logical and physical data

models for example (Codd, 1970). This layering can

have an adverse effect as the original semantic

structures may be distorted or lost completely as the

emphasis of the modelling activity moves from

representing the real world to representing data

structures. Consequently, when integrating data that

originates from different sources, the problem of

semantic heterogeneity arises – resolution is

required regarding differences in meaning,

interpretation or the intended use of related data

which forms a barrier to coherent semantic data

integration (Doan, Noy and Halevy, 2004).

1.2 Heterogeneous Foundational

Ontologies

Ontology provides a way of dealing with semantic

data integration. From a computational standpoint,

an ontology is generally taken as a ‘specification of

a conceptualization’ (Gruber, 1995) – that is, a

description of the concepts and relationships that are

considered legitimate within a particular system of

thought. In terms of the concrete implementation of

software systems, foundational ontologies can be

used to establish the fundamental ‘meta’ objects and

relations used to construct more specific domain

ontologies. If a common foundational theory is

extended and specialised to model a number of

domain ontologies, then objects common to each of

these domains will have the necessary (common)

grounding to enable semantic integration.

Consequently, foundational ontologies are important

as they provide a standpoint that underpins all the

domain models to be integrated – providing a

semantic grounding.

It is the case, however, that several such

standpoints (related to foundational ontologies)

exist, Each provides a criterion for the ontological

commitments made (implicitly or explicitly), which

are principally the things believed to exist within the

context of a particular theory such as four-

dimensionalism (Quine, 1952; Sider, 2003). An

understanding of ontological commitment, however,

means that the computational view of ontology

needs to defer to a philosophical one, which is more

specifically concerned with the nature of being

(metaphysics). As metaphysical theories differ on a

number of dimensions (realism versus idealism,

endurantism versus perdurantism to name but two)

differences thus appear in foundational ontologies.

Furthermore, and perhaps more importantly, the

degree to which foundational ontologies are actually

grounded in metaphysics varies. Clearly, a lack of

consensus at the metaphysical level introduces

obstacles to semantic integration (Campbell and

Shapiro, 1995) that result in weaknesses in

computational applications:

a) Lack of Grounding. Many current models

employed within information systems have no

form of grounding in a more fundamental theory

(Cregan, 2007). Thus the ontological

commitments underlying the model are

unknown. On examination of many Linked Open

Data ontologies, they are often ungrounded.

b) Integrating Elements from Models which are

Founded on Different Theories. There are

many automatic translation techniques for

translating RDBMS schema and data to an

OWL ‘ontology’. However, there is a lack of

recognition that the expressivity of Description

Logics (that underlie OWL) and RDBMS are

ICEIS 2016 - 18th International Conference on Enterprise Information Systems

128

different as are the unique naming and the

open/closed world-view assumptions.

c) Model Strata and Translations. As noted earlier,

the requirement to translate the high-level models

of reality created at the initial design to structures

that are focused on the execution environment can

result in semantic distortion. There is also the

problem of translating between run-time

representations; the often cited OO-RDBMS

impedance mismatch (Ireland et al., 2009).

d) Over Simplification to Fit a Model of Reality

to a Tractable First Order Logic (FOL)

Theory. The simplification of the abstraction of

reality to fit neatly into a FOL theory, thus

ignoring the fact that reality is not so simple and

higher order objects exist (Bailey, 2011).

e) Dividing Models into Static and Dynamic

Types. The separation of static and dynamic

aspects of reality into different structural and

process models leads to the development of

incompatible abstractions together with ‘exotic’

relations that are employed to bridge these static

and dynamic worlds.

f) Naming and Meaning Confusion. There is

often confusion between an entity’s naming and

meaning (Bailey and Partridge, 2009). An

object’s place in reality (and within ontology)

should define its meaning.

g) Establishing Identify. Many modelling and

information systems use ephemeral means of

establishing an entity’s identity which do not

function well over time.

h) Employing Techniques that do Not Scale.

Software tools such as OWL tableau calculus-

based reasoners are constrained by memory and

cannot be easily scaled to inference over

ontologies containing large instance populations

(Bock et al., 2008). The alternative is to use

simplified semantics and rule based reasoning -

that could in many cases employ standard

RDBMS techniques.

i) Semantic Integration Mismatches. For a more

extensive discussion on the types of semantic

integration mismatches see Visser et al., (1997)

who provide an extensive list of e semantic

mismatches that can occur when integrating

disparate datasets.

2 BORO FOUNDATIONAL

ONTOLOGY

Having examined several foundational ontologies

from a philosophical perspective, the research

described here adopts the Business Object Reference

Ontology (BORO) (Partridge, 2005) to semantically

interpret the original datasets and models. We adopt

BORO on the grounds that the ontology can

overcome the dichotomy that exists between

dynamic and static modelling paradigms and its

metaphysical thoroughness. Hence, the same model

can represent processes and things that are not

traditionally considered as processes (e.g., people,

products, machines, etc.). BORO represents all

individual elements (e.g. the activity, the person

assuming a role and the resource consumed) in

exactly the same way (i.e. as spatiotemporal

extents). BORO is based on a philosophical (rather

than computational) definition of ontology because

it requires more clarity on “the set of things whose

existence is acknowledged by a particular theory or

system of thought” (Lowe, 1998, p.634.). Key to

overcoming the dichotomy noted is the fact that

BORO is perdurantist (and thus extensionalist) in its

nature. In perdurantism (or 4D) an individual object

is never wholly present at one point is time, but only

partly present (a temporal part). For example, John

is not fully present in any given phase of his life

(e.g., childhood), he is fully present from his birth to

his death only – therefore, John’s childhood is a

temporal part of John. Identity is thus defined by an

individual object’s spatiotemporal extension (or

extent). Figure 1 represents the key part of the

foundational ontology relevant for the purposes of

this paper.

Figure 1: BORO Foundational Ontology (top level).

More in-depth discussions are provided in

Partridge, 2002; 2005; Bailey and Partridge 2009;

Bailey, 2011; Partridge et al., 2013). At the top level

4D-SETL - A Semantic Data Integration Framework

129

the BORO foundational ontology represents:

• Elements, which are individual objects or objects

with a spatiotemporal extent. For example, the

person John.

• Types, which are sets or objects that can have

instances. The identity of a type is also extensional

but, in this case, it is defined as the set of its

instances (i.e. members). For example, the extension

of the type Persons is the set of all people.

• Tuples, which are relationships between objects.

The identity of a tuple is defined by the places in the

tuple. An example is (Persons, John) in which the

type Persons and the element John occupy places 1

and 2 in the tuple respectively. This specific tuple is

an instance of the tuple type typeInstances in BORO.

In turn, Elements is subtyped by:

• Events: An event is an element that does not

persist through time (i.e. an event has zero

‘thickness’ along the time dimension). Events

represent temporal boundaries that either create

(CreationEvents) or dissolve (DissolutionEvents)

elements (e.g., a person or a person’s childhood

state).

• States: A state is an element that persists through

time. States (and elements in general) are bounded

by events. A state (like all elements) can have

further temporal parts (i.e. states and events).

Specific TupleTypes (or types whose instance are

tuples) relevant here are:

• temporalPartOf: This tuple type relates an

individual with its temporal parts (states and/or

events).

• happensTo: This tuple type relates an event with

one or more elements affected by the event.

happensTo has two subtypes:

- creates: Relates a creation event with the

element(s) whose creation is triggered by the event.

- dissolves: Relates a dissolution event with the

element(s) whose dissolution is triggered by the

event.

• happensIn: This tuple type relates an event with a

time instant or interval (TimeInstantsOrIntervals)

and it indicates the time in which an event takes

place.

As a note of importance for the example shown

later, names are types in BORO. The instances of the

name of an individual (e.g. John Smith’s Name) are

all utterances (written, spoken, etc.) that name that

individual (e.g., John Smith). Therefore while a

name, is a type its instances are spatiotemporal

extents. To provide clarity within the ontology,

‘names’ as much elements of the ontology as the

things they name. A name object will belong to a

Name Space which holds all names related to a

particular naming authority or domain. As the

ontology adopts a theory of utterances – each

utterance of a name is an individual element and so

has an extent (Strawson, 1964). Therefore, a name is

a Type that has as instances all utterances of the

same name individuals.

3 4D SEMANTIC EXTRACT

TRANSFORM AND LOAD

FRAMEWORK (4D-SETL)

Given an outline understanding of the foundational

ontology, we now describe a Semantic Extract-

Transform-Load framework. Given a variety of data

input, 4D-SETL is designed to output a graph

database in accordance with the BORO foundational

ontology. The framework was designed around a

number of industry standard tools and technologies

(e.g., a UML design tool and a Graph Database),

supplemented where necessary with custom software

implemented in Java. The key technology choices

made for the initial implementation were threefold.

First spreadsheets were employed to document each

dataset. Second, a UML design tool (Enterprise

Architect) was selected as the graphical design tool

for the ontological models and a BORO custom

UML profile created: The advantage is that BORO

UML enables easy manipulation and design of each

of the required domain ontologies. Last, the Neo4J

Graph database was chosen for persistence, for

several reasons: (a) Primarily due to its flexibility in

enabling BORO to be used as the foundational

‘schema’ (both can be seen as graphs); (b)

scalability in order to handle model and instance

data volume appropriately; (c) Neo4J's web-based

interface also provides access to the graph database

for development testing; and (d) Neo4J Cypher

provides an appropriate means of querying and

updating the graph database resident data.

The Semantic Extract Transform Load (ETL)

process is shown in Figure 2, the key stages of the

process are as follows:

a) Semantic Extraction and Transformation.

The input data to a semantic integration process

may be structured in many forms –e.g., as fixed

record or delimited tabular files, RDF, RDFS,

OWL etc. – and may consist of both model

(schema) level and/or instance level data. Thus

the first stage in the semantic integration

ICEIS 2016 - 18th International Conference on Enterprise Information Systems

130

process begins with documenting the dataset

which can be considered a semantic extraction

and transformation process. The BORO

foundation provides a view of reality and the

patterns that can be employed perform this

interpretation and transformation. The

foundational ontology provides the equivalent

of a canonical data model (Saltor et al., 1991)

that can be employed to develop domain models

providing the semantics that are common to all

datasets that will be integrated. Thus the

translation process results in a new schema

(domain ontology) that extends the ontic

categories and patterns of the foundation.

Through this process, schemas are developed to

represent the entities and relationships that are

represented by the data. Finally this schema is

documented using a profile of UML that

conforms to BORO semantics.

b) Ontology Model ETL. Once a domain model

has been created in an ontologically consistent

form the semantic load and integration process

can be undertaken. Firstly the domain

ontological model, which includes such patterns

as type and classification taxonomies, is

translated from the BORO UML model and

loaded to the graph database. The 4D-SETL

framework provides a Java application to

translate the BORO UML and load it to the

graph database.

c) Ontology Data ETL. Next, the instance level

dataset is loaded and integrated. It is through

this process that the integration of individual

elements takes place. Integration can be

considered to take place within vertical and

horizontal planes. Initially the ‘vertical’

relationships between an individual element and

the domain ontology (and hence the foundation

ontology) is asserted, which consists of

establishing the individual relationships (such as

type instance, etc.). Then the ‘horizontal’

relationships that are deemed to hold between

individual domain level objects are established

(such as a company being located at a particular

geographic location). Foundational ontological

patterns can then be applied to simplify this

process. This can be a complex process that

requires both one-to-many and many-to-one

transformations. The 4D-SETL framework

provides a Java application to perform this

process.

Figure 2: 4D-SETL Framework.

4 EXPERIMENTAL DATA

As the foundational ontology is an integral part of

the framework, prior to processing any of the

domain ontological elements (model and datasets)

the foundational ontology is transformed to graph

format and loaded to the database. This is achieved

via the 4D-SETL framework which extracts the

BORO UML as XML (XMI), then transforms it to a

set of nodes and edges that are loaded to the

database. The graph database ontology also includes

the UML model identifiers as indexed node and

edge parameter key-values, these are employed to

enable the reproduction of the design time UML

models within the graph database runtime

environment and to establish the relationships

between the foundation and other subsequent

domain model elements that are loaded. The 4D-

SETL framework was applied to Extract, Transform

and Load (ETL) five datasets of varying scale and

complexity related to corporate data:

• Calendar: temporal locations (1856 to present).

• Location: spatial locations (~2.5M locations).

• Standard Industrial Classification (taxonomy).

• UK Companies (~3.5M)

• UK corporate officers (~12M)

4D-SETL - A Semantic Data Integration Framework

131

5 OUTCOMES AND

LIMITATIONS

Having applied the framework, our experience is

that BORO provides a coherent lens through which

to view and model the world together with the

foundational ontological elements and patterns

through which the domain ontologies can be

developed to represent the datasets to be

semantically integrated (Partridge, 2002). In terms

of domain ontology development, this work concurs

with the work of Keet (2011), who stated that

employing foundational ontologies provides

advantages in terms of the quality and

interoperability of domain ontologies. Developing

such domain ontologies provided the means of

semantically integrating data conforming to different

models and theories – a necessary evil in dealing

with variety in big data.

Employing a graph database provided the means

of importing and restructuring data in a manner that

directly reflects the ontological model patterns

without the normal translation to tabular RDBMS or

Object Oriented form and not introducing the

‘impedance mismatch’ problem (Ireland et al.,

2009). Dispensing with RDBMS storage in favour of

a property graph data model removed the

partitioning of the storage structures between data

and schema and allows both ‘schema’ ontological

model objects and instance level objects to be

updated at run time. This supports the work related

to graph databases by Webber (2012). Related to this

finding, it was also demonstrated in this study that

patterns could be established within the warehouse

that directly reflected the physical or socially

constructed patterns of reality such as taxonomies

and taxonomic ranks, the latter of which employed

the powertype pattern (equivalent to the set theoretic

powerset) to more accurately reflect the nature of

such classification systems. These aspects of 4D

ontologies (along with others) provide a greater level

of flexibility and reusability when evolving the

warehouse system and therefore concur and take

forward the initial findings of Partridge (2002).

In practical terms, we propose that the data

structures resulting from the 4D-SETL process are

more suitable for discovering relationships within

data rather than for example processing aggregate

data (Vicknair et al., 2010). It is relatively easy, for

example, to discover all relationships that exist

between two elements using a standard algorithm

from the Neo4J library (designed to find all

available paths or the shortest path between two

nodes). Further, the Cypher graph database query

facilities provide the means of discovering more

complex patterns of relationships between the

people, company officers, company activities, events

and physical location. Finally, it was found through

the evaluation and empirical experiment on the

prototype warehouse (graph database) that data load

and information retrieval response times that the

prototype could be developed into a practical

information system. This was confirmed by

performing test data query (graph traversal)

experiments that for example, performed graph

traversals to retrieve all companies within a postcode

location (61 milliseconds) and all officers for a

specific business organisation (37 milliseconds) thus

the prototype produced indicative response times

within bounds that would support interactive

applications (Bhatti et al., 2000). Testing also

confirmed the graph database performance

evaluation undertaken by Vicknair et al. (2010).

Thus using a graph database and the parameter

graph model to store the ontology, alongside query

information via graph traversal, circumvents the

issues that limit the ability of systems built using

triple stores and tableau calculus-based reasoner

technology to deal with ontologies that are both

expressive and have with very large instance level

elements (arguably exactly what one would want

from big data). Neo4J is highly scalable and

provides capacities for Nodes/Edges of ~34 billion

and properties at least ∼ 68 billion respectively.

With the issue of disparate data sources in mind,

the work here has: (a) Examined the potential

contribution of foundational ontology; and (b)

described an implementation of a Semantic Extract-

Transform-Load framework (4D-SETL) based on

BORO, a 4D foundational ontology. Foundational

ontologies provide a ‘grounding’ for our view of

reality and thus provide a common reference through

which to model and integrate heterogeneous data.

The 4D-SETL framework uses the BORO

foundational ontology to harmonise data and then

generates a graph database that accords with the

semantic commitments made by that ontology. The

effectiveness of the framework was examined

applying it to large-scale open datasets related to

company information to semantically interpret and

integrate the datasets and to instantiate a prototype

graph database warehouse to persist the resultant

ontology. Our implementation is a prototype at this

stage and the use of foundational ontologies is not

without challenge (e.g., automation in the context of

real-time data streams). Accepting such limitations,

however, the potential utility of the 4D-SETL

framework can be seen in its ability to model and

ICEIS 2016 - 18th International Conference on Enterprise Information Systems

132

instantiate a number of complex ontological

structures, such as higher order taxonomic ranks.

The patterns specialised from the core foundational

BORO ontology patterns offer a high degree of

flexibility and reusability when evolving the graph-

based warehouse system. We have thus

demonstrated how a 4D (perdurantist) foundational

ontology can be employed to semantically interpret

and structure data, showing that a single coherent

ontology can be developed and loaded to a graph

database without the problems associated with

current approaches – e.g., model distortion, over

simplification or scalability problems.

Understandably, the work here is not without its

limitations, which may be summarised as follows.

First, and at the outset, the interpretation process is

manual. BORO encourages the development of

patterns (for ontological reuse), which allow for

partial automisation but skills in ontological

modelling are necessary throughout. In the context

of dealing with variety in big data automatic

translation of data is of particular importance. As a

consequence, pattern development and the extraction

of the rules associated with that are also of

importance for ongoing research. Second, as

previously noted, BORO is one of several

foundational ontologies and further work is required

to understand their relative comparative advantages

and disadvantages.

The work here was supported by funding from

the Engineering and Physical Sciences Research

Council (Project EP/L021250/1). The experimental

research data and metadata (Ontology) for this

project was sourced from the following

organisations: Companies House (2016), Company

Information; UK Office of National Statistics (2016)

Geographic Location (ONSPD Product); UK Office

of National Statistics (2016), Standard Industrial

Classification; Company Officers: (A commercial

credit reference agency); BORO Engineering

Limited (2016), Foundational Ontology.

The Companies House and ONS Datasets are

UK Open Government Data and can be freely

downloaded. The Company Officers and BORO

Ontology are commercial in-confidence.

REFERENCES

Arsanjani, A. (2002) 'Developing and Integrating

enterprise Components and Services', Communications

of the ACM, 45(10), pp. 30-34.

Bailey, I. (2011) 'Enterprise Ontologies–Better Models of

Business', in Intelligence-based systems engineering.

Springer, pp. 327-342.

Bailey, I. and Partridge, C. (2009) 'Working with

extensional ontology for defence applications',

Ontology in Intelligence Conference.

Bhatti, N., Bouch, A. and Kuchinsky, A. (2000)

'Integrating user-perceived quality into web server

design', Computer Networks, 33(1), pp. 1-16.

Bock, J., Haase, P., Ji, Q. and Volz, R. (2008)

'Benchmarking OWL reasoners', Proc. of the

ARea2008 Workshop, Tenerife, Spain (June 2008).

BORO Engineering Limited (2016) ‘BORO Ontology’.

Available from: < http://www.borosolutions.co.uk/so

lutions/resources/boro-presentations-and papers >. [16

February 2016].

Bricker, P. (2014) 'Ontological Commitment', in Edward

N. Zalta (ed.) The Stanford Encyclopedia of

Philosophy. Winter 2014 edn.

Campbell, A. and Shapiro, S. (1995) 'Ontological

Mediation: An Overview', IJCAI Workshop on Basic

Ontological Issues in Knowledge Sharing. 1995.

AAAI Press, Menlo Park, CA,.

Codd, E. (1970) 'A relational model of data for large

shared data banks', Communications of the ACM,

13(6), pp. 377-387.

Companies House (2016), Free Company Data Product.

Available from: < http://download.companieshouse.g

ov.uk/en_output.html >. [16 February 2016].

Cregan, A. (2007) 'Symbol grounding for the semantic

web', in The Semantic Web: Research and

Applications. Springer, pp. 429-442.

Doan, A., Noy, N.F. and Halevy, A.Y. (2004)

'Introduction to the special issue on semantic

integration', ACM Sigmod Record, 33(4), pp. 11-13.

Gangemi, A., Guarino, N., Masolo, C., Oltramari, A. and

Schneider, L. (2002) 'Sweetening ontologies with

DOLCE', in Knowledge engineering and knowledge

management: Ontologies and the semantic Web.

Springer, pp. 166-181.

Grenon, P. and Smith, B. (2004) 'SNAP and SPAN:

Towards dynamic spatial ontology', Spatial cognition

and computation, 4(1), pp. 69-104.

Gruber, T.R. (1995) 'Toward principles for the design of

ontologies used for knowledge sharing?', International

journal of human-computer studies, 43(5), pp. 907-

928.

Guizzardi, G., de Almeida Falbo, R. and Guizzardi, R.

(2008) 'Grounding Software Domain Ontologies in the

Unified Foundational Ontology (UFO): The case of

the ODE Software Process Ontology.', CIbSE. , 127-

140.

Herre, H. (2010) 'General Formal Ontology (GFO): A

foundational ontology for conceptual modelling', in

Theory and applications of ontology: computer

applications. Springer, pp. 297-345.

Ireland, C., Bowers, D., Newton, M. and Waugh, K.

(2009) 'A classification of object-relational impedance

mismatch', Advances in Databases, Knowledge, and

4D-SETL - A Semantic Data Integration Framework

133

Data Applications, 2009. DBKDA'09. First

International Conference on. IEEE, 36-43.

Keet, M. (2011) 'The use of foundational ontologies in

ontology development: an empirical assessment', in

The Semantic Web: Research and Applications.

Springer, pp. 321-335.

Kent, W. (1978) Data and reality : basic assumptions in

data processing reconsidered. Amsterdam ; Oxford:

North-Holland Publishing Co.

Lowe, E.J. (1998) 'Ontology.', in Hondreich, T. (ed.) The

Oxford Companion to Philosophy. New York: Oxford

University Press, pp. 634.

Lycett, M. (2013) ''Datafication': Making sense of (big)

data in a complex world', .

Partridge, C. (2002) 'The role of ontology in integrating

semantically heterogeneous databases', Rapport

technique, 5(02).

Partridge, C., Mitchell, A. and de Cesare, S. (2013)

'Guidelines for developing Ontological Architecures in

Modelling and Simulation', in Tolk, A. (ed.) Ontology,

Epistemology, and Teleology for Modeling and

Simulation. Berlin Heidelberg: Springer, pp. 27-57.

Office of National Statistics (2016), Postcode Data

Product. Available from: < http://www.ons.gov.uk/ons

/guide-method/geography/products/postcode-directorie

s/-nspp-/index.html>. [16 February 2016].

Office of National Statistics (2016), Standard Industrial

Classification System 2007. Available from: <

http://www.ons.gov.uk/ons/guide-method/classificatio

ns/current-standard-classifications/standard-industrial-

classification/index.html>. [16 February 2016].

Partridge, C. (2005) Business objects. 2nd edn. Oxford:

Butterworth Heinemann.

Quine, W.V. (1952) Methods of logic. Routledge and

Kegan Paul.

Saltor, F., Castellanos, M. and Garcia-Solaco, M. (1991)

'Suitability of Data models As Canonical Models for

Federated Databases', SIGMOD Rec., 20(4), pp. 44-48.

Sheth, A.P. and Larson, J.A. (1990) 'Federated Database

Systems for Managing Distributed, Heterogeneous,

and Autonomous Databases', ACM Comput.Surv.,

22(3), pp. 183-236.

Sider, T. (2003) Four-dimensionalism: An ontology of

persistence and time. Oxford.

Strawson, P. F. "Identifying reference and truthvalues",

Theoria, 30(2), 1964 pp. 96-118.

Vicknair, C., Macias, M., Zhao, Z., Nan, X., Chen, Y. and

Wilkins, D. (2010) 'A comparison of a graph database

and a relational database: a data provenance

perspective', Proceedings of the 48th annual Southeast

regional conference. ACM, 42.

Visser, P.R., Jones, D.M., Bench-Capon, T. and Shave, M.

(1997) 'An analysis of ontology mismatches;

heterogeneity versus interoperability', AAAI 1997

Spring Symposium on Ontological Engineering,

Stanford CA., USA. , 164-172.

Webber, J. (2012) 'A programmatic introduction to Neo4j',

Proceedings of the 3rd annual conference on Systems,

programming, and applications: software for

humanity.

ACM, 217-218.

Zikopoulos, P. and Eaton, C. (2011) Understanding big

data: Analytics for enterprise class hadoop and

streaming data. McGraw-Hill Osborne Media.

ICEIS 2016 - 18th International Conference on Enterprise Information Systems

134