Grounding Conceptual Modelling Templates on Existing Ontologies

A Delicate Balance

Chiara Di Francescomarino, Chiara Ghidini and Muhammad Tahir Khan

FBK-IRST, Via Sommarive 18, 38050 Trento, Italy

Keywords:

Templates, Conceptual Modelling, Ontology.

Abstract:

Using templates for conceptual modelling is becoming a fashionable way to guide and facilitate Domain

Experts in providing rich and good quality knowledge. A possibility to build templates without starting from

scratch is grounding them on existing foundational and core ontologies. In this paper we investigate how

these ontologies can be effectively used for the construction of templates able to guarantee a balance between

usability and rigorousness. We report ﬁndings and lesson learned from a survey carried out for the evaluation

of templates built from existing foundational and core ontologies in the enterprise domain.

1 INTRODUCTION

Effectively involving domain experts in the process

of authoring OWL ontologies, making them able not

only to provide domain knowledge to knowledge en-

gineers but also to directly write ontologies on their

own or together with knowledge engineers, is in-

creasingly recognised in a number of works (see e.g.,

(Dimitrova et al., 2008; Tudorache et al., 2010)) as

a crucial and challenging step to make the construc-

tion of ontologies more agile and apt to the needs of

organisations and business enterprises.

A popular way of involving domain experts in au-

thoring conceptual models is to provide them with

graphical languages (diagrammatic representations).

Nevertheless, as argued in (Kop and Mayr, 2011) and

brieﬂy summarised in Section 2, diagrammatic repre-

sentations are often adequate to represent only part of

the knowledge that needs to be contained in a concep-

tual model, but are not used to render more complex

knowledge such as axioms and properties of relations.

Templates, usually constituted by tables, forms, con-

trolled language sentence patterns or any other kind

of structured textual information, have been proposed

in literature as an effective way of “encapsulating

knowledge and modelling recurrent sets of axioms”

in ontology engineering (Parreiras et al., 2010) and

have been used in existing tools such as MoKi (Di

Francescomarino et al., 2012), COE (Hayes et al.,

2005) and Populous (Jupp et al., 2012) to comple-

ment graphical languages and guide domain experts

to describe ontology elements in a structured textual

fashion. Templates, however, are difﬁcult to build.

They have to balance the needs of both domain ex-

perts and knowledge engineers, i.e., being usable but

also rigorous enough. Moreover, methodologies for

the deﬁnition of usable and rigorous templates need

to be put in place, in order to facilitate their adoption

and the customisation of tools that make use of them

to speciﬁc scenarios.

In this paper we propose our approach along with

a concrete method to build templates based on a

mixture of foundational, mid-level and domain spe-

ciﬁc ontologies, and we evaluate it through a survey

conducted with domain experts and knowledge en-

gineers. More in detail we investigate the ease of

understanding, usefulness, completeness and correct-

ness of templates for modelling key entities of an en-

terprise ontology based on 23 foundational, mid-level

and domain-speciﬁc ontologies. The main ﬁndings

of our analysis are the following: (i) grounding tem-

plates on existing ontologies allows the construction

of templates usable and rigorous enough; (ii) ground-

ing templates on foundational or mid-level ontologies

and characterising them with specialised domain on-

tologies allows us to reach the right level of gran-

ularity of templates, thus making them more usable

and precisely characterised; (iii) domain experts per-

ceive template usability more positively than knowl-

edge engineers, thus validating the key role of tem-

plates in supporting domain experts.

To the best of our knowledge, the evaluation per-

formed in this paper provides a ﬁrst empirically rig-

orous evaluation of the support provided by a wide

199

Di Francescomarino C., Ghidini C. and Khan M..

Grounding Conceptual Modelling Templates on Existing Ontologies - A Delicate Balance.

DOI: 10.5220/0004542501990206

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2013), pages 199-206

ISBN: 978-989-8565-81-5

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

range of existing ontologies to the deﬁnition of struc-

tured forms that can guide domain experts in ontol-

ogy modelling and therefore highlight, in a compre-

hensive manner, the potential and criticality of using

existing ontologies as a base for template-based ap-

proaches.

2 TEMPLATE-BASED

CONCEPTUAL MODELLING

As extensively described in (Kop and Mayr, 2011),

template-based approaches have been used as an ef-

fective method to elicit knowledge in the area of soft-

ware engineering and database in the last 40 years.

By templates, we refer here to any kind of struc-

tured textual information (such as tables, forms, con-

trolled language sentence patterns) were the meaning

of the textual information contained in the template

can be determined according to its position within the

structure. The ﬁrst examples were tabular representa-

tions for the description of software functions (Janicki

et al., 1997), and glossaries for a structured descrip-

tion of terms relevant for a domain used in database

design methodologies such as DATAID (Albano et al.,

1985) and more recently in the KCPM (Klagenfurt

Conceptual Predesign Model) template-based mod-

elling language (V

ohringer and Mayr, 2006). With

the advent of object-oriented analysis and design, and

of popular graphical modelling languages for the rep-

resentation of use cases such as UML (Uniﬁed mod-

elling language), templates remain an effective way

of complementing the graphical use case description

with detailed information (Cockburn, 2000), which

usually constitutes the biggest portion of knowledge

of a conceptual model

In the last few years several works have started to

propose templates also to support Ontology Engineer-

ing activities in order to encapsulate complexity and

to better involve domain experts in the activity of on-

tology authoring (Parreiras et al., 2010; Di Francesco-

marino et al., 2012; Hayes et al., 2005; Jupp et al.,

2012). Among the main advantages of introducing

templates in Ontology Engineering listed in literature

are: (i) a decrease in complexity, as templates hide

the complexity of the axioms and expose only the re-

quired parameters to be ﬁlled in; (ii) an increase in

efﬁciency of modelling, as templates can be reused

in the modelling of several ontologies; and (iii) an

In (Kop and Mayr, 2011), Kop and Mayr portray the

information contained in a conceptual model as an iceberg,

whose visible part is constituted by the graphical represen-

tation, and whose hidden (biggest) part is contained in ad-

ditional textual descriptions such as templates.

Data Properties

Full Name

Location

Start Date

End Date

Event type

Event description

Event

Every Event has partecipant

Every Event is hosted by

Every Event is organized by

Every Event is associated to

Object Properties

Organization/People

Process

Is a

Every Event is a

Figure 1: A template for “Event”.

increase in reliability, as templates comprise set of

axioms developed by domain experts who, therefore

place domain knowledge at the heart of the modelling

process.

Templates in MoKi. MoKi

is a collaborative Me-

diaWiki-based

tool where teams of knowledge engi-

neers and domain experts, with different knowledge

engineering skills, can actively collaborate to author

ontologies. To foster collaboration MoKi provides

two different modes to access the formal content of

an ontology: a fully-structured access mode, where

knowledge engineers can edit the ontology content by

means of OWL axioms, and a lightly-structured ac-

cess mode, where domain experts can edit the ontol-

ogy content by means of a simpliﬁed, template-based,

view on the (same) formal description. As argued in

(Di Francescomarino et al., 2012), the ability of eas-

ily customising interfaces to create ad hoc templates

such as the one in Figure 1 is one of the advantages of

MoKi, and of MediaWiki-based tools in general.

However, deﬁning ad hoc templates such as the

one drafted in Figure 1 that can be used by domain

experts to: (i) reﬁne / adapt the characterisation of

the concept to the speciﬁc domain; (ii) model more

speciﬁc concepts (e.g., a concept “Product launch

event”); and (iii) populate the ontology with speciﬁc

instances of the concept (e.g., the instance “Fiat 500

UK launch event”) is a difﬁcult and time consuming

task, as templates need to be produced by balancing

usability and rigorousness aspects.

To address this problem we have investigated an

A comprehensive description of MoKi is omitted for

lack of space. The interested reader can refer to (Di

Francescomarino et al., 2012) and http://moki.fbk.eu

http://www.mediawiki.org

KEOD2013-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

200

approach which builds templates which we present

and evaluate in Sections 3 and 4. Note that while

the work presented here is motivated by the work on

MoKi, the approach and the evaluation do not depend

upon MoKi. Therefore they can provide an insight on

the potential and problems of using existing ontolo-

gies to structure templates for ontology engineering.

3 BUILDING TEMPLATES FOR

ENTERPRISE ONTOLOGIES

In this section we provide an overview of the ap-

proach we followed to construct templates for the do-

main of Enterprise Ontologies. We selected this do-

main as this provides an important use case where do-

main experts are usually involved, but the suggested

method can provide suitable guidelines for the def-

inition of templates to support the task of ontology

engineering in a wide range of speciﬁc domains.

The Approach. The approach relies on building the

templates based on a model extracted from a range

of ontologies, which include foundational, mid-level

and domain-speciﬁc ontologies. The reason to ground

templates on existing ontologies, instead of building

them from scratch, is that of exploiting the ontologi-

cal analysis performed and modelling patterns already

encapsulated in well crafted ontologies, and of mak-

ing them available to domain experts to reuse and

adapt. The reason to exploit a number of existing on-

tologies, ranging from foundational to mid-level, to

speciﬁc ones is to evaluate the appropriateness of the

different resources to the speciﬁc task.

The Realisation. The method we have followed is

composed of four macro steps.

As a ﬁrst step, we focus on selecting entities rel-

evant to our problem domain (together with their re-

lated axioms) from foundational ontologies. The rea-

son to start from foundational ontologies is that they

contain a well founded characterisation of the founda-

tional ontological categories, and therefore provide a

good a starting point for building good quality ontolo-

gies. In our study we decided to use the DOLCE on-

tology (Masolo et al., 2003), which, we believe, pro-

vides a good reference ontology on a wide range of

domains which encompass the one investigated in our

experiment.

The second step concerns the complementation /

speciﬁcation of the entities selected from the DOLCE

ontology with entities selected from core (mid-level)

ontologies. In fact, while grounding the templates

on DOLCE helps us to start from a rigorous char-

acterisation of the entities they have to describe, we

need to produce templates for entities that are usu-

ally more speciﬁc than the categories contained in a

foundational ontology. Mid-level ontologies provide

the best candidate for this task as they usually de-

scribe speciﬁc (yet general) concepts for a particular

domain. Since we started from DOLCE we decided

to ﬁrst review mid-level ontologies that are produced

either as extensions of DOLCE (see e.g., the Ontol-

ogy of Social roles presented in (Masolo et al., 2004))

or grounded on DOLCE (see e.g., (Boella and van der

Torre, 2006)) and to select entities relevant to our do-

main together with their characterisation.

The third step concerns thebuilding of taxonomy

by using the is-A axioms (among entities) selected in

the ﬁrst two steps. In our case this resulted in a taxon-

omy of 20 entities, whose relevant part is illustrated

in Figure 2. The leaves of this taxonomy constitute

the entities to be characterised in the templates (7 in

our case). Note that while reviewing the mid-level

ontologies we found out that some of the entities we

were interested in are contained both in foundational

and mid-level ontologies. In our experiment that was

the case for “event” and “process”. In Figure 2 we

have decided to depict the entities in the most abstract

type of ontology in which they were found. Thus

“event” and “process” are listed among the entities

coming from foundational ontologies. We also found

out that several entities are contained in more than

one mid-level ontology. In these cases we decided

to merge the characterisation of the entities contained

in the different ontologies to produce a more compre-

hensive one, and possibly to solve some conﬂicting

issues, which were nevertheless extremely limited in

our experience.

The ﬁnal step concerns the complementation of

the characterisation of the different entities with

knowledge coming from domain-speciﬁc ontologies.

In our scenario we looked at domain-speciﬁc ontolo-

gies such as The Enterprise Ontology (Uschold et al.,

1998) and Schema.org

By applying the method above we built an on-

tology of 20 concepts, 86 relations and 159 axioms

which has provided the basis for the construction of

the templates for Organization, Actor, Role, Process,

Event, Task, and Artefact. In detail, the ontology De-

scription Logic axioms have been used for the auto-

matic construction of the templates. A description of

the ontology and of the resulting templates is omitted

for lack of space and can be found in (Ghidini et al.,

2012).

http://schema.org

GroundingConceptualModellingTemplatesonExistingOntologies-ADelicateBalance

201

Figure 2: A case for Enterprise Model.

4 EVALUATION

We are interested in investigating whether templates

built starting from existing ontologies can be used to

support domain experts in ontology modelling while

preserving the rigorousness and the accuracy charac-

terising the good principles of knowledge engineer-

ing. These templates should indeed be able to ﬁll the

gap between Domain Experts and Knowledge Engi-

neers while balancing usability and rigorousness. In

order to evaluate whether they can suit the different

requirements they are supposed to satisfy, we built

(Section 3) a set of templates grounded on different

existing ontologies ranging from foundational to spe-

ciﬁc domain ontologies and we asked experts to eval-

uate them.

4.1 Description

The goal of the evaluation is investigating (i) the us-

ability (in terms of ease of understanding and use-

fulness) of templates built starting from existing on-

tologies and (ii) their capability to completely and

precisely capture the modelling needs of the speciﬁc

domain, as perceived by both Domain Experts and

Knowledge Engineers. The research questions we are

interested to evaluate are:

RQ1. Are templates built from existing ontologies

perceived as easy to understand and useful by Do-

main Experts and Knowledge Engineers?

RQ2. Are templates built from existing ontologies

able to completely and precisely satisfy the mod-

elling needs of a speciﬁc domain according to Do-

main Experts and Knowledge Engineers?

In order to answer the above research questions,

we provided the seven templates described in Sec-

tion 3 and we asked a group of Domain Experts and

Knowledge Engineers to answer a questionnaire in-

vestigating four main factors: (i) ease of understand-

ing; (ii) usefulness for ontology modelling tasks; (iii)

characterisation completeness, i.e., how good is the

proposed characterisation to include relevant and nec-

essary properties and (iv) characterisation precision,

i.e., how good is the proposed characterisation to

omit redundant and useless properties for each tem-

plate. We used the subjective perception expressed

by subjects about the ﬁrst two factors for investigat-

ing whether templates built using well known existing

ontologies are actually easy to understand and use-

ful (RQ1) and the last two factors for a subjective

evaluation on the feasibility of the usage of ontologi-

cal resources for addressing domain modelling issues

(RQ2).

The subjects involved in the evaluation were 43:

19 Domain Experts (mainly enterprise workers) and

24 Knowledge Engineers (both modelling experts and

ontologists). We asked each of them to evaluate a set

of three (Role, Artefact and Actor) or four (Organiza-

tion, Task, Event and Process) templates.

Each template has been evaluated by means of

a set of questions belonging to two categories: (i)

closed-ended questions, where answers were selected

from a 5 point Likert-scale, (ii) open-ended questions,

where participants were allowed to provide sugges-

tions for improving the templates. In detail, we used

closed-ended questions and 5 point scales for ask-

ing experts their subjective perception about template

ease of understanding (0=immediate, ..., 2=reason-

able, ..., 4=complex), usefulness (0=absolutely use-

ful, ..., 2=neither useful nor useless, ..., 4=absolutely

useless) and completeness (0=totally complete, ...,

2=neither complete nor incomplete, ..., 4=totally in-

complete). For the precision, instead, subjects were

asked to evaluate, for each template characterisation,

whether they would keep or discard it (yes/no an-

swer). For each template, we used the collected an-

swers to compute the subjective precision of each ex-

pert. According to the classic information retrieval

precision metrics, we computed the subjective pre-

cision as the ratio between the number of character-

isation properties provided and evaluated as correct

by the expert and the total number of characterisa-

tion properties provided. We then classiﬁed the val-

ues of each subjective precision in three categories:

high (100% precision, i.e., the subject agreed on all

the provided template characterisations), reasonable

(the precision is between 66% and 100%, i.e., more

than 2/3 of the characterizations have been judged

correct), and low (less than 66% of precision).

KEOD2013-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

202

4.2 Results

Here we report the results of our observations and the

statistical analyses we applied to strengthen these re-

sults. In detail, to assess the positivity of the experts’

evaluations, since values are derived from an ordi-

nal scale, we test medians using a one-tailed Mann-

Whitney test (Wohlin et al., 2000), verifying the hy-

pothesis that

F ≤ 2, where F can be each of the four

factors under investigation,

F is its median and 2 the

intermediate value of the scale. All the analyses are

realized with a level of conﬁdence of 95% (p-value

< 0.05), i.e., there is only a 5% of probability that the

results are obtained by chance.

Table 1 reports the percentage distribution related

to the answers provided by subjects for RQ1 and

RQ2. For the sake of readability, in reporting the

percentage distribution, we group together the two

positive and the two negative values of the three 5

point scales, thus reporting the percentages on 3 point

scales, the same used for the subjective precision

(high, reasonable and low). For each of the factors,

the median value of the provided evaluations is 1.

Considering the above results, it comes out that

subjects perceived the templates as overall easy to un-

derstand (the median value is 1, i.e., easy on the 5

point scale). Such a overall positive evaluation is also

conﬁrmed at statistical level (p-value < 0.05). In-

deed, by looking at the distribution of the provided

answers, more than half of them state the easy under-

standing of templates, while in only 11% of the cases

the templates have been evaluated difﬁcult to compre-

hend. Nevertheless, the number of answers that are

not clearly positive or negative is not trivial (about

33%). Considering the distribution of these “answers

in the middle” per template, we found that the high

number of reasonable answers, can be partially con-

nected to the Role template. Indeed by looking at Fig-

ure 3 (left), showing the subjective evaluation related

to ease of understanding for each template, the crit-

icality of the Role template is quite evident: 62.5%

of the subjects expressed a borderline answer for this

template.

Subjects provided an overall positive evaluation

also with respect to template usefulness. The median

value of the evaluation on the 5 point scale is indeed

1, i.e., useful. The result is also statistically conﬁrmed

with a level of conﬁdence of 95% (p-value < 0.05) by

the Mann-Whitney test. Templates have been judged

overall useful in almost 65% of the cases versus the

15% of useless evaluations. In this case, the border-

line answers were about 20% and, one of the reasons

for these uncertain answers, could be the lack of us-

age of the templates in a real setting. Indeed, most

Table 1: Statistical measure of the survey results

RQ Factor Rate Percentage

RQ1

Ease of Understanding

easy 55.3%

reasonable 33.3%

difﬁcult 11.4%

Usefulness

useful 64.9%

neither useful nor

useless

20.2%

useless 14.9%

RQ2

Completeness

complete 56.1%

neither complete nor

incomplete

25.4%

incomplete 18.4%

Precision

high 82.5%

reasonable 14.9%

low 2.6%

of the uncertain (neither useful nor useless) answers

(96% of the uncertain answers) has been provided by

Knowledge Engineers who have a less deep knowl-

edge of the speciﬁc domain, i.e., the enterprise one.

We can hence conclude that overall easy to under-

stand and useful templates can be built starting from

existing ontologies and hence positively answer RQ1.

The third row of Table 1, reports the distribution

of the survey answers related to the characterisation

completeness. Experts evaluated the characterisations

as overall complete: the median value is 1, i.e., com-

plete in the 5 point scale, with

Compl < 2 (p-value

< 0.05). In this case, although more than half of the

evaluations were positive, the percentage of the neg-

ative ones is slightly more than for the other factors

(18%). Similarly to the usefulness evaluation, most of

the 25% of uncertain answers (79.3%) was provided

by Knowledge Engineers, who are overall more rigor-

ous about what a complete characterisation is.

Finally, the fourth row of Table 1 reports the val-

ues related to the precision of the template character-

isation. Overall the precision is high: the median is 1,

i.e., the highest value and

P ≤ 2 (with p-value ≤ 0.05).

Indeed, most of the evaluations (82%) are high, while

only a small percentage (2.6%) is negative.

Considering the results related to characterisation

completeness and precision we can hence positively

answer RQ2: ontology-based template characterisa-

tions are perceived as overall complete and precise.

4.3 Findings and Discussion

The analyses of the data carried out for answering the

above research questions suggest that differences ex-

ist in the perception of Domain Experts and Knowl-

edge Engineers, as well as in their evaluation of the

different templates.

Surprisingly, it seems that, on average, Domain

Experts judged template ease of understanding, use-

GroundingConceptualModellingTemplatesonExistingOntologies-ADelicateBalance

203

Figure 3: Template usability subjective evaluation.

fulness and characterisation precision slightly more

positively than Knowledge Engineers. Indeed, the av-

erage evaluations of Domain Experts for the ease of

understanding, usefulness and characterisation preci-

sion (1.67, 1.61 and 1.24, respectively) are slightly

higher than those provided by Knowledge Engineers

(1.64, 1.58 and 1.17). On the contrary, the complete-

ness of the template characterisation has been evalu-

ated slightly more positively by Knowledge Engineers

(1.68 for Domain Experts versus 1.73 for Knowledge

Engineers). This result, that conﬁrms the validity of

ontology-based templates in supporting Domain Ex-

perts, can be partially explained by considering the

different competencies and objectives guiding the two

roles. Domain Experts, indeed, are interested in the

practical issues related to the modelling of the do-

main. They want templates that are useful to describe

the entities of the domain without caring whether

terms used for/in the template could also convey other

meanings in other contexts. On the contrary, Knowl-

edge Engineers are more focused on the formal part:

from their perspective, a term is easy to understand

if it is enough detailed and formal so that its seman-

tics cannot be confused; similar criteria also apply for

usability and precision of template characterisation.

Differences among templates, instead, clearly

come out by looking at the distribution of the users

evaluations about the four investigated factors across

the templates (Figure 3 and Figure 4). Figure 3 shows

that the evaluations related to ease of understanding

and usefulness of the Role template are the worst

ones. On the contrary, the Event and the Task tem-

plates seem the easiest to understand and the Event,

Task, Process and Actor templates the most useful.

Slightly different is the completeness case in which

(Figure 4, left) the most problematic template is the

Organization one, that is the template characterised

by the highest number of properties. Finally, looking

at the evaluations for the characterisation precision

(Figure 4, right), we found again that the Event and

the Process templates outperform the others, while

the Role and Actor characterisations are less precise.

With the aim of investigating the reasons behind

the criticality of some of the templates (e.g., the Role

template) and the positive perception of others (e.g.,

the Event template), we looked more in detail at how

they have been built and at the ontologies they have

been grounded on. Overall, two (including the out-

performing Event template) out of the seven templates

are deﬁned in foundational ontologies (other than

in mid-level and domain-speciﬁc ontologies). Table

2 reports for each template, the most abstract cate-

gory of ontology in which the entity is deﬁned, as

well as the percentage of characterisation attributes

and properties belonging to foundational, mid-level

or domain-speciﬁc ontologies, respectively. Finally,

it also reports the percentage of attributes and prop-

erties used to complete the characterisation, as well

as the percentage of properties common to more than

two ontologies.

The table reveals that differences exist in the tem-

plate composition (in terms of ontology categories

used for their construction). Crossing these data with

the users’ evaluations we found that the largest part

of the characterisation of the templates that has been

perceived easier to understand and more precisely de-

ﬁned than the others (i.e., the Event and the Task tem-

plates), comes from domain-speciﬁc ontologies. This

result is also conﬁrmed at statistical level. Indeed, we

found a strong positive correlation

between the per-

centage of characterisations deriving from domain-

speciﬁc ontologies and the average evaluation of both

ease of understanding and characterisation precision.

In other terms, the percentage of speciﬁc properties

is directly proportional to the template ease of under-

standing and precision. On the contrary, a negative

correlation exists between the percentage of attributes

and properties from mid-level ontologies and the av-

erage perception of ease of understanding, usefulness

and characterisation precision. A second factor that

seems inﬂuencing the perception of both ease of un-

derstanding and usefulness is the number of differ-

ent source ontologies sharing the same property or

attribute. Surprisingly, characterisations shared by

more than two ontologies seem to negatively affect

We performed a correlation analysis applying the Pear-

son’s coefﬁcient at 95% conﬁdence level.

KEOD2013-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

204

Figure 4: Subjective evaluation of the template characterisation completeness and precision.

Table 2: Source ontologies used for building templates.

Template

Most abstract

Foundational Mid-level

Domain-

Other

Shared by > 2

ontology category speciﬁc ontologies

Role mid-level 0 93.8% 0 6.3% 37.5%

Artefact mid-level 0 75% 8.3 % 16.7% 14.3%

Actor mid-level 23.5% 58.8% 5.9% 11.8% 0

Organization mid-level 10 % 40% 30% 20% 0

Task mid-level 0 40% 40% 20% 0

Event foundational 7.7% 15.4% 46.2% 30.8% 0

Process foundational 8.3% 50% 33% 8.3% 0

the overall perception of the ease of understanding

and usefulness of the template. This result can also

be connected to the negative correlation between too

generic characterisations and their usability.

Summarizing, hence, (i) we found that ground-

ing templates on foundational and core ontologies, al-

lows the construction of templates usable (6 out of 7

templates were perceived as easy to understand and

useful by more than 50% of the subjects) and rigor-

ous (the characterisation of 6/7 of the 7 templates has

been judged complete/precise by more than 50% of

the evaluators). Moreover, (ii) we observed that to

improve the (perceived) usability of the templates, the

right level of granularity has to be devised by oppor-

tunely grounding templates on foundational or mid-

level ontologies and characterising them also with

specialized domain ontologies. Finally (iii) we val-

idated the usability of ontology-based templates by

Domain Experts.

Threats to Validity. Despite the rigorousness and

the care used for conducting the survey and for ana-

lyzing the results some threats can hinder the result

validity. A ﬁrst threat is related to the lack of usage of

the templates in real-life tasks. Nevertheless, the tem-

plates were accompanied with a description of a pos-

sible scenario, thus helping the subjects to ﬁgure out

a possible practical use case for the template usage.

A second threat is the subjectivity that has been used

for building the templates. The impact of this threat,

however, is partially limited by the deﬁnition and the

use of method for the construction of the templates.

Finally, the speciﬁcity of the domain considered is a

further threat to the validity of the study. On the other

hand, the enterprise domain is a quite general domain

in which Domain Experts are often involved.

5 RELATED WORK

Works dealing with templates based approaches in the

ontology engineering ﬁeld can be mainly classiﬁed in

two categories: those focused on the ontology struc-

ture and those oriented to both structural and domain

knowledge.

In (Parreiras et al., 2010), authors presents a ap-

proach to support ontology engineers in modelling of

ontology templates. They have demonstrated their ap-

proach with templates for ontology design patterns.

This work is different from our approach as it is more

focused toward knowledge engineers and it needs to

be extended, the usability aspect have not been stud-

ied and it has not been used for top level ontologies.

Also in MoKi (Di Francescomarino et al., 2012) tem-

plates are mainly used for providing structural sup-

port to the ontology authoring. The lightly-structured

view, indeed, provide a semi-formal rendering of the

OWL content in a predeﬁned template to support do-

main experts access to the content of the ontology.

The work of this paper will extend a tool like MoKi

with a set of templates grounded on existing founda-

tional ontology and core ontologies instead of starting

from scratch.

A slightly different approach is the one of Popu-

GroundingConceptualModellingTemplatesonExistingOntologies-ADelicateBalance

205

lous (Jupp et al., 2012), a template based approach

designed for domain experts to gather knowledge that

can be used to build ontologies. In Populous the au-

thors advocate the use of spreadsheets as templates to

gather and organise information about concepts and

their relationships as it provides a simple and intu-

itive form ﬁll-in style of user interface. However, the

main purpose of this approach is knowledge gather-

ing, the population of the templates at the point of

data entry, and not ontology building. In this paper,

instead, we focus on the complexity of building us-

able and rigorous templates. Another approach that

uses templates to capture experts knowledge is pre-

sented in (Hayes et al., 2005). In order to acquire

knowledge from experts this approach allows users to

structure their knowledge by using concepts maps or

the formalised knowledge structures deﬁned as tem-

plates. The templates in this approach corresponds

to commonly used owl structures (e.g., subclass, in-

stance, owl restrictions), they do not help very much

in the overall structure of the ontology.

6 CONCLUSIONS

The paper shows, through a rigorous evaluation, that

templates based on existing foundational and core on-

tologies are able to reconcile usability and formality

dimensions, ensuring ease of understanding and use-

fulness while guaranteeing completeness and preci-

sion. As a side effect of our evaluation we provide

some insights about the relation between the differ-

ent types of ontologies used for building templates

and their resulting perception by users. Future work

includes the investigation of the impact of different

categories of ontologies on template characteristics,

thus (i) reﬁning the proposed approach for the con-

struction of ontology-based templates; and (ii) devis-

ing measures and criteria of their appropriateness to

deﬁne templates that meet certain requirements.

REFERENCES

Albano, A., Antonellis, V. D., and Leva, A. D., edi-

tors (1985). Computer-Aided Database Design: the

DATAID approach. North-Holland.

Boella, G. and van der Torre, L. (2006). A foundational

ontology of organizations and roles. In Proceedings

of DALT’06, volume 4327 of LNCS, pages 78–88.

Springer-Verlag.

Cockburn, A. (2000). Writing Effective Use Cases. Addison

Wesley Publ. Comp.

Di Francescomarino, C., Ghidini, C., and Rospocher, M.

(2012). Evaluating wiki-enhanced ontology author-

ing. In Proceedings of EKAW2012, Galway, Ireland,

volume 7603 of LNCS, pages 292–301. Springer.

Dimitrova, V., Denaux, R., Hart, G., Dolbear, C., Holt, I.,

and Cohn, A. G. (2008). Involving domain experts

in authoring owl ontologies. In Proc. of ISWC 2008,

number 5318 in LNCS, pages 1–16. Springer.

Ghidini, C., Khan, M. T., and Di Francescomarino,

C. (2012). Grounding conceptual modeling tem-

plates on existing ontologies: a delicate balance.

Technical report. https://dkm.fbk.eu/images/a/a4/

TechicalReport paper.pdf.

Hayes, P., Eskridge, T. C., Saavedra, R., Reichherzer, T.,

Mehrotra, M., and Bobrovnikoff, D. (2005). Collab-

orative knowledge capture in ontologies. In Proceed-

ings of the K-CAP ’05, pages 99–106. ACM.

Janicki, R., Parnas, D. L., and Zucker, J. (1997). Tabu-

lar representations in relational documents. Relational

methods in computer science, pages 184–196.

Jupp, S., Horridge, M., Iannone, L., Klein, J., Owen,

S., Schanstra, J., Wolstencroft, K., and Stevens, R.

(2012). Populous: a tool for building owl ontologies

from templates. BMC Bioinformatics, 13(Suppl 1):S5.

Kop, C. and Mayr, H. C. (2011). The evolution of concep-

tual modeling. volume 6520 of LNCS, chapter Tem-

plates in domain modeling - a survey, pages 21–41.

Springer-Verlag.

Masolo, C., Borgo, S., Gangemi, A., Guarino, N., and

Oltramari, A. (2003). WonderWeb deliverable D18

ontology library (ﬁnal). Technical report.

Masolo, C., Vieu, L., Bottazzi, E., Catenacci, C., Ferrario,

R., Gangemi, A., and Guarino, N. (2004). Social roles

and their descriptions. In Proceedings of KR2004,

pages 267–277.

Parreiras, F. S., Gr

oner, G., Walter, T., and Staab, S. (2010).

A model-driven approach for using templates in owl

ontologies. In Proceedings of EKAW’10, volume 6317

of LNCS, pages 350–359. Springer-Verlag.

Tudorache, T., Falconer, S. M., Noy, N. F., Nyulas, C.,

Ust

un, T. B., Storey, M.-A. D., and Musen, M. A.

(2010). Ontology development for the masses: Cre-

ating icd-11 in webprot

e. In EKAW 2010. Proceed-

ings, volume 6317 of LNCS, pages 74–89. Springer.

Uschold, M., King, M., House, R., Moralee, S., and Zor-

gios, Y. (1998). The enterprise ontology. The Knowl-

edge Engineering Review, 13:31–89.

ohringer, J. and Mayr, H. C. (2006). Integration of

schemas on the pre-design level using the kcpm-

approach. In Advances in Information Systems Bridg-

ing the Gap between Academia & Industry, pages

623–634. Springer, Heidelberg.

Wohlin, C., Runeson, P., H

ost, M., Ohlsson, M. C., Regnell,

B., and Wessl

en, A. (2000). Experimentation in soft-

ware engineering: an introduction. Kluwer Academic

Publishers.

KEOD2013-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment

206