Lila Rao
Mathematics and Computer Science Department, The University of the West Indies, Mona, Kingston 7, Jamaica
Han Reichgelt
School of Computing and Software Engineering, Southern Polytechnic State University, Marietta, GA, U.S.A.
Kweku-Muata Osei-Bryson
Department of Information Systems & The Information Systems Research Institute
Virginia Commonwealth University Richmond, VA, U.S.A.
Keywords: Competency Questions, Knowledge Elicitation Techniques, Ontologies.
Abstract: This research explores the applicability of existing knowledge elicitation techniques for the development of
competency questions for ontologies. This is an important area of research as competency questions are
used to evaluate an ontology. The use of appropriate knowledge elicitation techniques increases the
likelihood that these competency questions will be reflective of what is needed of the ontology. It thus
helps ensure the quality of the ontology (i.e. the competency questions will adequately reflect the end users
Knowledge elicitation (KE) involves the gathering
of knowledge from experts (Shadbolt and Burton,
1989). There is a number of existing knowledge
elicitation techniques (e.g. 20 questions, card sort,
repertory grid and laddering) (Hickey and Davis,
2003). These techniques have been used extensively
for various types of applications (e.g. expert
systems) (Nakhimovsky et al., 2006, Reichgelt and
Shadbolt, 1992, Bryrd et al., 1992).
Ontologies have been identified as important
components of a number of information systems
(Guarino, 1998, Pinto and Martins, 2004) such as
knowledge management systems (Sicilia et al., 2005,
Rao and Osei-Bryson, 2007), e-business applications
(Lee et al., 2006, Fensel et al., 2001, Papazoglou,
2001) and data warehouses (Critchlow et al., 1998,
Shah et al., 2005). Therefore, the quality of the
overall system is likely to be highly dependent on
the quality of the ontology.
There are many different definitions of the term
“ontology” and different proposals for what should
be represented in the ontology. However, most agree
that it is some formal description of a domain, which
can be shared among different applications and
expressed in a language that can be used for
reasoning (Noy, 2004).
As ontologies grow in size and complexity
because of the increasing number of demands are
being placed on them, ensuring their quality is an
important consideration in the development of these
systems. Quality is a multi-dimensional concept
(Wang et al., 1995, Wand and Wang, 1996), and, in
order to assess the quality of the ontology a set of
dimensions should be defined. These dimensions
can be used to derive metrics that can be used not
only to assess the quality of the ontology but also to
determine whether proposed quality improvement
techniques are actually effective. One of the
proposed quality dimensions is coverage/
completeness (Jarke et al., 1999) which has been
defined as the extent to which the ontology covers
the domain of interest (Rao and Osei-Bryson, 2007).
This can be measured as the difference between
what is required of the ontology and what is
available in the ontology.
Rao L., Reichgelt H. and Osei-Bryson K. (2008).
In Proceedings of the Tenth International Conference on Enterprise Information Systems - ISAS, pages 105-110
DOI: 10.5220/0001678701050110
One of the most commonly used techniques to
evaluate ontologies is competency questions (Staab
et al., 2001, Sure et al., 2002). Competency
questions define the ontology’s requirements in the
form of questions that the ontology must be able to
answer (Gruninger and Fox, 1994, Gangemi, 2005).
These competency questions are actually providing
an approach for measuring the coverage of the
ontology as the percentage of the total set of
competency questions posed that can be answered
by the ontology is indicative of coverage. However,
in order for the measure to be accurate we must
ensure that the set of competency questions are
complete. It would be misleading to measure the
coverage of the ontology using this set of
competency questions if it is not likely that this set
of questions is complete. Thus, appropriate
techniques are needed for identifying competency
questions. If we ensure that the techniques used are
likely to lead to a complete set of competency
questions then the coverage measure will be more
dependable, which will help ensure the quality of the
ontology and hence the quality of the overall system.
We will demonstrate the applicability of these
elicitation techniques by building an ontology and a
set of corresponding competency questions for a
university’s information technology (IT)
infrastructure domain. Knowledge about this domain
is routinely used to solve a number of different
problems, ranging from troubleshooting to network
redesign and decisions about software to acquire to
server administration. Any ontology for this domain
can therefore be shared by a range of different users
solving of different problems.
The rest of the proposal is organised as follows.
Section 2 provides a review of the literature that is
relevant to this research, including the ontology
literature (Section 2.1) and various knowledge
elicitation techniques (Section 2.2). Section 3
describes the applicability of various knowledge
elicitation techniques to the development of the
competency questions and provides an illustrative
example using a specific domain. Finally, section 4
provides some concluding remarks and some
directions for future research.
2.1 Ontology and Competency
An ontology has been defined as “a formal
description of entities and their properties,
relationships, constraints, behaviors” (Gruninger and
Fox, 1995). A number of approaches have been
proposed for developing ontologies (Gruninger and
Fox, 1995, Staab et al., 2001). Gruninger and Fox
(1995) propose an approach to engineering
ontologies that consists of three steps:
1) Defining an ontology’s requirements in the
form of questions that an ontology must be able
to answer (i.e. competency questions). This is
known as the competency of the ontology (Fox
et al., 1998).
2) Defining the terminology of the ontology - its
objects, attributes and behaviours. In this way
the ontology provides the language that will be
used to express the definitions in the
terminology and the constraints required by the
3) Specifying the definitions and constraints on
the terminology.
Staab et al. (2001) describe an ontology
development process consisting of 5 phases (i.e. the
feasibility study, the kickoff phase for ontology
development, refinement, evaluation and
Competency questions thus provide an important
tool to validate an ontology as they can be used to
evaluate the ontological commitments that have
been made, and are indeed generally accepted as a
verification technique for ontologies (Kim et al.,
2007). Staab et al. (2001) recommend using these
competency questions for the evaluation phase of
their proposed ontology development process. Thus
the evaluation process is highly dependent on the
competency questions that are formulated, and it is
therefore imperative that the process of deriving the
competency questions is thorough, and it is therefore
crucial that one identify a set of techniques for
reliably eliciting all the competency questions.
2.2 Knowledge Elicitation Techniques
There is a number of existing knowledge elicitation
techniques such as interviews (e.g. structured,
unstructured and semi-structured), case studies,
prototyping, sorting (e.g. card sorting), triad
analysis, 20 questions, laddering and document
analysis (Shadbolt and Burton, 1989, Nakhimovsky
et al., 2006).
Laddering is used to construct a graphical
representation of the concepts and relations in a
domain. The elicitor makes use of prompts to
explore the expert’s understanding of the domain. A
graph, consisting of a number of nodes and labelled
arcs, is constructed in the presence of the expert.
ICEIS 2008 - International Conference on Enterprise Information Systems
This technique involves three main steps. The first
step involves asking the expert to identify a starting
point (seed item) (i.e. a concept that is important in
the domain). The next step involves moving around
the domain using various prompts (i.e. asking
questions to move down, across and up the expert’s
domain knowledge). The final step involves the
elicitation of attributes for the various concepts
(Reichgelt and Shadbolt, 1992).
The card sort, triad analysis and twenty questions
techniques assume that the knowledge engineer has
some prior knowledge of the domain under
consideration. This initial knowledge can be
obtained through available documentation as well as
by conducting unstructured interviews. The
available documentation can be used to get a sense
of the domain under consideration (i.e. some of the
basic concepts and relationships within the domain).
Once the knowledge engineer has some
understanding of the domain, unstructured
interviews can then be used for providing high level
knowledge of the domain. Unstructured interviews
suit the early stages of elicitation when the
knowledge engineer is trying to learn about the
domain but does not know enough to set up indirect
or highly structured tasks (Cooke, 1999).
Card sort entails the use of a given set of cards
with the names of relevant domain elements or
problems written on them. Experts are used to sort
the cards into several piles according to whatever
criteria they choose. This process is repeated until
the expert the expert has exhausted the ways to
partition the elements (Shadbolt and Burton, 1989).
Card sort is useful when the aim is uncover the
different ways that an expert sees the relationships
between a set of concepts (Reichgelt and Shadbolt,
Triad analysis requires that the expert is given or
asked to generate a set of important elements. The
interviewer randomly selects three of these examples
and asks the expert to distinguish between them such
that two of the examples in the triad have a common
property not possessed by the third (Ryan and
Bernard, 2000). This distinguishing property is
known as the construct. This process continues with
different triads of elements until no further
discriminating constructs can be identified by the
expert (Reichgelt and Shadbolt, 1992).
20 questions require that the knowledge engineer
chooses an element from the domain or a problem.
The domain expert is then required to determine
what the element or problem is but is only allowed
to ask questions that the knowledge engineer can
answer either yes or no (Kemp, 1996). This allows
the knowledge engineer to determine the heuristics
that an expert uses in his or her problem solving
Although competency questions are seen as a viable
way to evaluate an ontology (Gruninger and Fox,
1994), (Staab et al., 2001), (Sure et al., 2002),
(Gangemi, 2005), (Kim et al., 2007) there is limited
work describing appropriate techniques for
developing them. Gruninger and Fox (1995) state
that motivating scenarios should be used for
generating informal competency questions (see
Figure 1). However they do not elaborate on how
these motivating scenarios will be identified.
Figure 1: Procedure for Ontology Design and Evaluation
(Gruninger and Fox, 1995).
Sure et al. (2002) stress the importance of the
domain expert as a valuable source of knowledge for
structuring the domain. Personal interviews are a
commonly used method for knowledge acquisition
from domain experts, thus, they propose that the
competency questions should be derived from
interviews with the domain expert. However, they
do not elaborate on how to conduct these interviews.
They use these competency questions to create the
initial version of the semi-formal description of an
ontology as well as for the evaluation of the
ontology. Noy and Hafner (2007) also point to the
need for interaction between the knowledge engineer
and domain expert for the development of
competency questions but do not mention any
techniques that can be used for facilitating this
It seems fair to say that the applicability of
existing knowledge elicitation techniques to the
development of competency questions has not been
fully explored. However, given the fact that
researchers have reported great success with the use
of more structured knowledge elicitation techniques,
such as laddering, card sort etc, in knowledge
elicitation for expert systems (Shadbolt and Burton,
1989, Wang et al., 2006), it seems reasonable to
expect that such knowledge elicitation techniques
will also prove useful in the elicitation of
competency questions. We will therefore explore the
applicability of three knowledge elicitation
techniques that seem applicable to the development
of competency questions, namely card sort, triad
analysis and 20 questions. In particular, we will
explore the use of laddering for the elicitation of an
initial ontology, and 20-questions, triad analysis and
card sort for the development of the competency
Given the nature of the knowledge obtained
through laddering, it will be clear why we use
laddering to elicit the initial ontology. The questions
that the domain expert generates in the twenty
questions technique are questions that he/she
considers important in that domain and therefore
should be able to be answered by the ontology.
Using card sort, the ontology engineer will be able
to determine criteria that are important to the domain
expert for grouping similar cases, and thus form the
basis of the competency questions as the expert will
expect that the ontology can answer queries about
these concepts. Similar considerations apply for triad
It is likely the case that a combination of the
existing techniques (Shadbolt and Burton, 1989,
Harper et al., 2003) may actually be most effective
for eliciting competency questions. As mentioned
previously, each of the three techniques requires
some knowledge of the domain which can be
captured by reviewing available documentation and
using unstructured interviews. To get a more
detailed description of the domain the use of card
sort, triad analysis and 20-questions will be
A number of domain experts will be used in this
exercise. Multiple experts will help to ensure that as
many competency questions as possible can be
identified. Various groups of experts are likely to be
concerned with specific tasks within the domain and
therefore the knowledge elicited will be specific to
those tasks. Multiple experts will provide a
consensus of the important concepts and
relationships within the domain.
There have been a number of problems with
using these existing elicitation techniques. These
include, for example, the experts being adverse to
some of these techniques, the techniques being time
consuming and costly, combining the knowledge of
multiple experts, choosing the appropriate technique
(Cooke, 1999). However, tools and techniques have
been and are being developed to help address these
problems (Hickey and Davis, 2003, Harper et al.,
2003, Major and Reichgelt, 1990, Nakhimovsky et
al., 2006).
We will explore the usefulness of different
knowledge elicitation techniques to the development
and evaluation of ontologies by applying them to a
particular domain, namely the IT infrastructure
domain at a university campus in Jamaica.
Knowledge about the university’s IT infrastructure
can be used to solve various types of problems (e.g.
disaster recovery planning/business continuity
planning, security and risk management, training
and network design). Having a formal description of
the entities, relationships, constraints and behaviours
(Gruninger and Fox, 1995) in the domain ensures
that all the decisions are being made with the same
information. Additionally, various entities within the
university may need to communicate in order to
solve particular problems related to the IT
infrastructure. Having an ontology as a reference
will facilitate this communication as one of the main
purposes of an ontology is to formally describe the
domain of discourse so as to provide a common
language for all entities to communicate, thus
reducing the potential for ambiguity. Once
developed this ontology could then be used by other
universities that require the same types of problem
One of the problems requiring access to
information about the IT infrastructure domain is
disaster recovery planning (DRP). The aim of DRP
is to ensure that entities (i.e. the university) function
effectively during and following a disaster (Bryson
et al., 2002). A well-organized disaster recovery
plan will directly affect the recovery capabilities of
an entity. The contents of the plan should follow a
logical sequence and be written in a standard and
understandable format (Wold, 2002). For example,
in the case of the university campus in Jamaica there
is an annual threat of hurricanes. Therefore, the
disaster recovery plan should include procedures
that need to be followed in the event that a hurricane
becomes a threat to Jamaica. These procedures
would include, for example, the systems that would
need to be shut down, where they are physically
located, who is responsible for them being shut
down, who uses them so will be affected by their
shut down. Having this information readily available
in the ontology will make it possible to establish the
plan more effectively.
ICEIS 2008 - International Conference on Enterprise Information Systems
Additionally, the IT infrastructure domain
knowledge can be used to develop security plans for
the university’s systems. Both the information stored
in the system and the system as a whole needs to be
secured. In order to establish these security plans
decision makers would need to know, for example,
what information is stored in each system, the tasks
that the information is being used for, decisions that
these tasks are being used to make, the threat of
possible risks to the various systems. Again, if this
information is stored in an ontology then it will be
readily available to decisions makers, in a consistent,
standardised format.
The ontology for the IT infrastructure domain
will be developed using an extended form of
laddering as the main knowledge elicitation
technique. A number of the employees of the
university, playing various roles, will be used in the
knowledge elicitation process to develop this
domain ontology.
In order to evaluate the quality of the ontology
(i.e. the completeness/coverage of the ontology) a
set of competency questions for the IT infrastructure
domain will be developed. The ontology will be
considered to be of a high quality if it is able to
answer the competency questions. It is therefore
crucial that good methodologies for creating these
competency questions are found. Forced answer
techniques, such as twenty questions and card sort
may be good ways of doing this and their use will be
explored. The employees of the university will be
used as the domain experts for this process.
The evaluation of an ontology relies heavily on the
competency questions formulated, and the issue of
using appropriate knowledge elicitation techniques
to competency questions is therefore of central
importance in ontology development. This research
addresses this issue by exploring the applicability of
three specific knowledge elicitation techniques (i.e.
20 questions, triad analysis and card sort) to the
development of competency questions for an
If these techniques prove useful then this work
will help improve the quality of ontologies and in so
doing improve the quality of the systems (e.g.
knowledge management systems, e-commerce
systems and data warehouses) that they are a part of.
They will help derive a measure for the coverage of
the ontology which can help assess its quality.
In the future we will explore how the techniques
proposed in this paper can be used to develop an
approach for the development, representation and
evaluation of high quality ontologies.
We will also explore the additional benefits that
the knowledge elicitation process may provide. For
example, the process may help to identify the
various user groups within the domain. Identifying
these groups and their needs will help identify the
various user groups of the ontology. Those users that
formulate similar competency questions can be
classified as belonging to a particular group of users.
Based on these groups of users the ontology can then
be designed in a way (e.g. using subontologies) that
can maximise the efficiency of access to the
ontology. Further, as the system is used metadata
will be generated that will reflect the usage of the
system. This can be analysed to track the usage (i.e.
types of queries on the ontology) to determine if
they ontology needs restructuring (e.g. adding an
additional subontology for a frequently requested
type of query that was not identified in the initial
design). Therefore, the ontology will be maintained
as it is used. Thus, when a query is being processed
by the system, depending on the type of the query,
the appropriate subontology will be identified and
used for processing. Thus, the entire ontology will
not have to be searched. This can have significant
benefits as ontologies are becoming larger and more
Bryrd, T., Cossick, K. & Zmud, R. (1992) A Synthesis of
Research on Requirements Analysis and Knowledge
Aquisition Technqiues. MIS Quarterly.
Bryson, K.-M., Millar, H., Joseph, A. & Mobolurin, A.
(2002) Using Formal MS/OR Modeling to Support
Disaster Recovery Planning. European Journal of
Operational Research, 141, 679-688.
Cooke, N. J. (1999) Knowledge elicitation. IN DURSO, F.
T. (Ed.) Handbook of Applied Cognition. UK: Wiley.
Critchlow, T., Ganesh, M. & Musick, R. (1998) Automatic
Generation of Warehouse Mediators Using an
Ontology Engine. 5th International Conference on
Knowledge Representation Meets Databases Seattle,
Fensel, D., Van Harmelen, F., Universiteit, V., Horrocks,
I., Mcguinness, D. & Patel-Schneider, P. (2001) OIL:
An Ontology Infrastructure for the Semantic Web.
IEEE Intelligent Systems, 16, 38-45.
Fox, M. S., Barbuceanu, M., Gruninger, M. & Lin, J.
(1998) An Organization Ontology for Enterprise
Modeling. IN PRIETULA, M., CARLEY, K. &
GASSER, L. (Eds.) Simulating Organizations:
Computational Models of Institutions and Groups.
Menlo Park, CA., AAAI/MIT Press.
Gangemi, A. (2005) Ontology Design Patterns for
Semantic Web Content. 4th International Semantic
Web Conference (ISWC 2005). Galway, Ireland.
Gruninger, M. & Fox, M. S. (1994) The Design and
Evaluation of Ontologies for Enterprise Engineering.
Workshop on Implemented Ontologies, European
Conference on Artificial Intelligence (ECAI).
Amsterdam, NL.
Gruninger, M. & Fox, M. S. (1995) Methodology for the
Design and Evaluation of Ontologies. IJCAI'95,
Workshop on Basic Ontological Issues in Knowledge
Sharing. Montreal.
Guarino, N. (1998) Formal Ontology and Information
Systems. First International Conference on Formal
Ontologies in Information Systems. Trento, Italy, IOS
Harper, M. E., Jentsch, F. G., Berry, D., Lau, H. C.,
Bowers, C. & Salas, E. (2003) TPL–KATS-card sort:
A tool for assessing structural knowledge. Behavior
Research Methods, Instruments, & Computers, 35,
Hickey, A. M. & Davis, A. M. (2003) Elicitation
Technique Selection: How Do Experts Do It?
Requirements Engineering Conference. California.
Jarke, M., Jeusfeld, M., Quix, C. & Vassiliadis, P. (1999)
Architecture and Quality in Data Warehouses: An
Extended Repository Approach. Information Systems,
24, 229-253.
Kemp, E. A. (1996) The Role of the Individual Project in
Teaching Knowledge Acquisition. International
Conference on Software Engineering: Education and
Practice (SE:EP '96). Dunedin, New Zealand.
Kim, H. M., Fox, M. S. & Sengupta, A. (2007) How to
Build Enterprise Data Models to Achieve Compliance
to Standards or Regulatory Requirements (and share
data). Journal of the Association for Information
Systems, 8, 105-128.
Lee, T., Lee, I., Lee, S., Lee, S., Kim, D., Chun, J., Lee, H.
& Shim, J. (2006) Building an Operational Product
Ontology System. Electronic Commerce Research and
Applications, 5, 16–28.
Major, N. & Reichgelt, H. (1990) ALTO: An Automated
Laddering Tool. Current Trends in Knowledge
Acquisition. O C S L Press
Nakhimovsky, Y., Schusteritsch, R. & Rodden, K. (2006)
Scaling the Card Sort Method to Over 500 Items:
Restructuring the Google AdWords Help Center.
Conference on Human Factors in Computing Systems
CHI '06 Montréal, Québec, Canada
Noy, N. (2004) Semantic integration: A survey of
ontology based approaches. SIGMOD Record, 33, 65-
Papazoglou, M. P. (2001) Agent-Oriented Technology in
Support of E-Business. Communications of the ACM,
44, 71-77.
Pinto, H. S. & Martins, J. P. (2004) Ontologies: How Can
They Be Built? Knowledge and Information Systems
6, 441-464.
Rao, L. & Osei-Bryson, K.-M. (2007) Towards Defining
Dimensions of Knowledge Systems Quality. Expert
Systems with Applications, 33, 368-378.
Reichgelt, H. & Shadbolt, N. (1992) ProtoKEW: A
Knowledge-Based System for Knowledge Acquisition.
IN SLEEMAN, D. & BERNSEN, O. (Eds.) Research
Advances in Cognitive Science volume 5: Artificial
Intelligence. Hillsdale, NJ., Lawrence Erlbaum.
Ryan, G. W. & Bernard, H. R. (2000) Data Management
and Analysis Methods. IN DENZIN, N. & LINCOLN,
Y. (Eds.) Handbook of Qualitative Research.
Thousand Oaks, CA, Sage Publications, Inc.
Shadbolt, N. & Burton, A. M. (1989) The Empirical Study
of Knowledge Elicitation Techniques. ACM SIGART
Bulletin, 15-18.
Shah, S., Huang, Y., Xu, T., Yuen, M., Ling, J. &
Ouellette, F. (2005) Atlas – A Data Warehouse for
Integrative Bioinformatics. BMC Bioinformatics 6, 34.
Sicilia, M.-A., Lytras, M., Rodriguez, E. & Garcia-
Barriocanal, E. (2005) Integrating descriptions of
knowledge management learning activities into large
ontological structures: A case study. Data &
Knowledge Engineering, article in press.
Staab, S., Schnurr, H.-P., Studer, R. & Sure, Y. (2001)
Knowledge Processes and Ontologies. IEEE
Intelligent Systems, 16, 26-34.
Sure, Y., Erdmann, M., Angele, J., Staab, S., Studer, R. &
Wenke, D. (2002) OntoEdit: Collaborative Ontology
Development for the Semantic Web. First
International Semantic Web Conference (ISWC 2002).
Sardinia, Italy.
Wand, Y. & Wang, R. Y. (1996) Anchoring Data Quality
Dimensions in Ontological Foundations.
Communications of the ACM, 39, 86-95.
Wang, R. Y., Storey, V. C. & Firth, C. P. (1995) A
Framework for Analysis of Data Quality Research.
IEEE Transactions on Knowledge and Data
Engineering, 7, 623-640.
Wang, Y., Sure, Y., Stevens, R. & Rector, A. (2006)
Knowledge Elicitation Plug-in for Prot´eg´e: Card
Sorting and Laddering. Asian Semantic Web
Conference (ASWC'06). Beijing, China.
Wold, G. H. (2002) Disaster Recover Planning Process
Disaster Recovery Journal 5, 29-34.
ICEIS 2008 - International Conference on Enterprise Information Systems