Task Knowledge Model for Triage Decision-Support

Shamimi A. Halim

, Muthukkaruppan Annamalai

, Rashidi Ahmad

and Mohd Sharifuddin Ahmad

Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA, Shah Alam, Malaysia

Trauma and Emergency Department, University Malaya Medical Centre, Kuala Lumpur, Malaysia

Centre for Agent Technology, Universiti Tenaga Nasional, Kajang, Malaysia

Keywords: Triage Decision-Support, Task Modelling, Commonkads.

Abstract: The paper discusses the construction of the task knowledge model to support the development of a triage

decision-support system. Knowledge rather than experience is predominant in the triage decision making.

Since the triage decision making knowledge is complex, we resort to knowledge modelling to help in the

systematisation of the knowledge. The paper concentrates on the modelling of the task knowledge model to

back triage decision-support. We adopted the CommonKADS methodology as a basis of modelling and

engineering the knowledge. The top-down modelling approach availed general task structures that could be

reused and adapted to engineer the triage decision-support task knowledge model. Consequently, the

resulting task model informs the engineering of the triage decision-support domain knowledge model.

1 INTRODUCTION

The triage decision making that takes place in

hospital Emergency Department (ED) involves

clinical judgments that need to be made quickly

under conditions of uncertainty. The decisions

however, have a major impact on the mortality and

mobility of the patients. A common characteristic in

drawing the triage decision depends on the

knowledge and experience of the triage officers

(Considine et al., 2007). Then again, studies have

shown that knowledge plays a more important role

compared to experience in determining the triage

decisions (LeVasseur et al., 2001).

In our previous work (Halim et al., 2011), we

have noted that the benefits of implementing a

uniform and more robust triage in Malaysian EDs

that can ensure consistent triage decisions, which is

possible with the aid of a Triage Decision-Support

System (TDSS). The TDSS can help to determine

the ‘right’ triage level of a patient by reasoning over

the represented triage decision-support knowledge.

However, the triage decision making knowledge is

rather complex. It consists of factual and procedural

knowledge gathered from decision rules and clinical

practices and guidelines. Therefore there is a need to

organise the supporting knowledge before we

develop the TDSS. For that reason, we resorted to

knowledge modelling to help in the systematisation

of the triage decision-support knowledge.

Two crucial models of knowledge in a

knowledge-based system are the task and the domain

knowledge models (Annamalai, 2006). These

models can help in the understanding of a

knowledge intensive process, and lead the way to

interact with them. The modelling of the domain

knowledge is directed by its purposive mechanism

(Annamalai and Sterling, 2003). In this regard, the

triage decision-support task knowledge model

informs the representation of the knowledge

resources to support the task. Cognitive tasks

involve inferences. Some knowledge modelling

methods additionally advocate the use of an

inference knowledge model to elaborate the

inference knowledge (Tarta, 2004).

The paper is organised as follows. Section 2

analyses the top-down knowledge engineering and

modelling methods. Section 3 describes the

modelling of the triage decision-support task.

Section 4 briefly discusses its validation. Finally,

section 5 concludes the paper and points to future

work.

261

A. Halim S., Annamalai M., Ahmad R. and Sharifuddin Ahmad M..

Task Knowledge Model for Triage Decision-Support.

DOI: 10.5220/0004546802610268

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2013), pages 261-268

ISBN: 978-989-8565-81-5

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

Table 1: Contemporary knowledge engineering methods.

Knowledge Engineering

Method

Strength

Our

requirement

Generic Task

Inspired by diagnostic and design tasks ×

Include fixed PSM strategy which specify the inference steps ×

Provides task specific vocabulary √

Role-limiting

Views PSM as fixed ×

Provides a set of predefined terminology √

MIKE

Proposes informal and semi-formal specification techniques to describe knowledge √

Uses executable KARL ×

Applies reversible process in system development ×

Protégé-II

Allows development of PSM independently from the knowledge base √

Decomposable of PSM into sub-method enable a configuration of generic problem-solver √

The domain layer is comprise in domain ontology √

Platform specific design and implementation ×

KADS

CommonKADS

Has a broad view of the process during early stage of requirement and analysis √

The Model of Expertise provides four layers of knowledge: domain, inference, task and

strategy layer

√

Provides general knowledge-intensive task templates √

Emphasize on the internal control √

Platform independent design and implementation √

2 KNOWLEDGE ENGINEERING

AND MODELLING METHODS

Knowledge engineering extracts the concepts and

relationships among them from the knowledge

sources and resources, and defines them in

knowledge models.

We propose to adopt a top-down knowledge

engineering approach to the modelling of the triage

decision-support knowledge. The top-down

modelling approach avail general task structures that

could be reused and adapted to engineer the

knowledge models (Kingston, 2007). Knowledge

acquisition is directed and focussed to knowledge

that is relevant to the problem in hand. As a result,

the time required for knowledge acquisition and

analysis will be reduced.

Contemporary top-down knowledge modelling

and engineering methods are Generic task

(Chandrasekaran, 1986), Role-limiting (Marcus,

1988), MIKE (Angele et al., 1998), Protégé-II

(Gennari et al., 2003), KADS (Schreiber et al., 1993)

and CommonKADS (Schreiber et al., 2000). Table

1 highlights the strengths of each method and

indicate the ones (with

√), which we think are useful

to support the development of the TDSS.

Inspired by diagnostic and design task, the

Generic Task try to solve different types problems

by creating a taxonomy or vocabulary which

appropriate for a particular domain knowledge.

However, the clarity of knowledge representation is

weak because the fact that the languages to

implement the expert system is not standardized

across the tasks. On the other hand, the Role-limiting

method separates the Problem Solving Method

(PSM) from the domain knowledge where the object

and their relation including the environment are

fixed building block. It also provides a predefined

terminology. The orientation of the terminology is a

problem-solving-method-specific, and not domain-

specific. This feature gives flexibility to knowledge

engineer to accommodate a particular domain

MIKE proposes the informal and semi-formal

specification techniques to describe the knowledge.

MIKE uses KARL (Knowledge Acquisition and

Representation Language) to describe the

functionality of the knowledge precisely. Since it is

an executable language, the specification will be

developed based on prototyping approach and the

functionality can be tested by a running prototype.

Another special feature of MIKE is the ability to

increment and reverse during system development

process. However, the ability to develop the system

is not required since we do not intend to develop the

TDSS in this manner to test its functionality.

Protégé-II provides several PSMs which were

developed separately from the knowledge base and

those PSMs can be used to work with different

knowledge bases and solve different real-world

problems. Ever though this feature is not significant

in the development of TDSS, the decomposable of

generic PSM featuring in Protégé-II will help a lot.

The domain layer in Model of Expertise in Protégé-

II was captured in domain ontology and the other

three layers were kept optional and can be used for

any appropriate PSM. This separation however

limits the system-level view of the process,

particularly in triage process. The implementation is

however within the Protégé knowledge acquisition

environment.

KADS on the other hand provides all layers of

knowledge in Model of Expertise. Therefore the

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

262

Figure 1: Triage Decision-support Task model.

domain, procedural, inferences needed and any

election of tasks can be viewed. The KADS and the

evolved CommonKADS methods support the

modelling of knowledge-intensive tasks which

divided into analytic and synthetic tasks. The task

structures are captured generically as task templates.

The template of each task type is flexible to addition

and modification of its inference in order to fit a

particular application task. Therefore, this method

gives a flexibility to control the reasoning process.

It allowed us to capture the expert reasoning

strategies especially in sequencing the reasoning

steps.

Among the task types, we found the Assessment

task structure as a suitable to adapt for engineering

the triage decision-support task knowledge. The goal

of Assessment task is to determine a decision for a

set of case (condition or event) with domain-specific

norms as a rule. In the context of triage decision

making, a set of case will be presented as a patient’s

condition while the norms consist of established

triage guidelines or scale. CommonKADS has

emerged as an industrial strength knowledge

engineering method, and have been used in many

cases (Lindow et al., 2013).

3 TRIAGE DECISION-SUPPORT

TASK MODELLING

The explanation about this section will begin with

structuring the triage decision-support task model

and followed by inference models.

3.1 Triage Decision-Support Task

Model

We engineered the triage decision-support task

model by adapting the generic CommonKADS

Assessment task. Figure 1 shows the resulting triage

decision-support task model. In the process, we

have made several modifications to the Assessment

task structure. The Sort and Verify inferences are

new. The Sort inference is introduce since the triage

decision deals with a set of case which consist of

more than one elements that need to be prioritized

according to a particular preference. The Select

inference is replaced by Fetch, which is a non-

cognitive action. This action will fetch the element

that has been prioritized by the Sort inference. The

decision-support process flow, as a whole, has been

revised and restructured to reflect the flow and

processing of knowledge in triage decision making.

In Table 2, we describe the key knowledge or

information resources utilised in the triage decision-

support task.

The dotted flow line A in Figure 1 points to the

iterative addition of a modifier under consideration.

If there are no norms to be considered for that

modifier, the evaluation continues using other

modifiers (indicated by dotted flow line B).

As shown in Figure 1, there are eight inferences

in our triage decision-support task model: Specify

modifier, Abstract modifier value, Sort modifier,

Evaluate abstracted modifiers value, Verify if a

TaskKnowledgeModelforTriageDecision-Support

263

Figure 2: Triage Decision-support Task, Inference model and the Specific Inferences.

Table 2: The key knowledge/ information resources

utilised in the Triage decision-support task.

Term Description

Case Knowledge or information gathered from the

patient, which consists of oral history, vital signs

and observation.

Chief

Complaint

(CC)

The most significant illness inferred from Case

(based on Habboushe’s guide (Habboushe, 2012))

Modifiers A determinant to determine a triage level. Eleven

modifiers are involved (Murray, Bullard and

Grafstein, 2004): Glasgow Coma Score (GCS),

Respiratory distress, Hemodynamic stability,

Dedicated presenting complaint, Mental health,

Bleeding severity, Hypertension, Temperature,

Pain, Mechanism of injury and Blood glucose.

Modifier

value

The value assign to a modifier. Ex. Modifier value

is 38C for modifier Temperature.

orms of

modifier /

Modifier’s

norm

A conditional rule of a modifier. Each modifier

has more than one rule. The following is example

of a norm.

IF level of distress is severe

THEN triage level is I

Triage level The outcome of a norm’s evaluation. Triage levels

consist of level one to level five in a robust triage

scale (Gerdtz et al, 2009). The triage level

indicates the severity of a patient’s clinical

condition.

Critical

condition

The critical condition refers to triage level I and

II, i.e., when the patient must be given immediate

treatment.

triage value indicates a critical condition, Verify if

all norms of modifier X have been considered, Verify

if all modifiers have been considered and Match

modifier. Another two non-cognitive actions are

Fetch the set of norms of modifier X and Fetch a

norm from a set of norms of modifier X. These

inferences will be described in section 3.2.1, 3.2.2,

3.2.3, 3.2.4, 3.2.5, 3.2.6 and 3.2.7, respectively.

3.2 Triage Inference Models

Based on the triage decision-support task model, we

have decomposed six intermediate inference

subtasks, namely Specification, Abstraction, Sorting,

Verification, Evaluation and Match. The

decomposition of the task describes the control of

sequence of task design and help in determining the

inference models. Figure 2 shows the task

decomposition of the triage decision-support task.

Triage decision-support is the most general task.

The inference knowledge details the reasoning

mechanism for the triage decision-support solution.

This type of knowledge is described by specifying

the performed function and their input and output.

The six inference models that will guide the

representation of the specific inference knowledge

are: Specification model, Abstraction model,

Selection model, Verification model, Evaluation

model and Matching model. These models are

described in the CommonKADS reference

(Schreiber et al., 2000).

The following sub-sections will explain the

inferences models.

3.2.1 Specification Inference

The Specification inference in the context of triage

decision-support will identify CC as an input and

determines a list of modifier as output. The output is

produced by inferring over the Specify inferencing

knowledge.

The list of modifiers consists of two groups:

Specified and Optional modifier. A specified

modifier is a modifier which has been identified by a

domain expert. The selection of modifier is primarily

determined by the degree of severity of the

symptom. For example headache is a symptom and

increase intra cranial pressure, migraine and stress

are causes of headache. The level of triage is

determined by the Pain modifier value not by the

cause of headache and together with other modifiers:

GCS, Hypertension, Presenting complaint, Bleeding

and Mechanism of injury. The optional modifier is

compliment handle for the triage officer, in case

more modifiers are needed. The following pseudo

code broadly describes the Specify inference

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

264

knowledge. Note Variable X represents a particular

CC.

READ CC from chief complaint list

IF CC is X

THEN DISPLAY list of Specified

modifiers for X

3.2.2 Abstraction Inference

The Abstraction inference has case description as

input and an abstracted case description as output. A

case description can be a description about certain

conditions, situations or any attributes that explain

the entity. For example, a value of Respiratory

modifier “cyanosis” is abstracted into severe level of

distress.

In the triage decision-support task, out of the

eleven modifiers, four modifier values are abstracted

indirectly (computed) from the case, while other

values can be extracted directly. The following

pseudo code describes part of the Abstraction

inference knowledge.

READ modifier value for GCS from

list of Specified modifier for GCS

READ modifier value for Presenting

Complaint from list of specified

modifier for Presenting Complaint

...

IF modifier value for GCS is not

null

THEN CALL function abstraction GCS

IF modifier value for Temperature is

not null

THEN abstracted value Temperature =

modifier value for Temperature

...

In the above example, the GCS modifier value

has to be abstracted indirectly, while the

Temperature can be extracted directly from the case.

The following pseudo code explains the identified

abstraction of the GCS modifier value.

READ eye opening response from list

of specified modifier for GCS

READ verbal response from list of

specified modifier for GCS

READ motor response from list of

specified modifier for GCS

IF eye opening response is

spontaneous

THEN eye point = 4

IF eye opening response is verbal

stimuli

THEN eye point = 3

...

IF verbal response is oriented

THEN verbal point = 5

IF verbal response is confused

THEN verbal point = 4

...

IF motor response is obeys commands

THEN motor point = 5

IF motor response is withdraws in

response to pain

THEN motor point = 4

...

COMPUTE abstracted value GCS = eye

point + verbal point + motor point

3.2.3 Sorting Inference

A sorting inference has a set of elements as input

and a sorted list which contains the same elements as

an output. This inference decides the relative order

of two or more elements. In the triage decision-

support task, the Sorting inference prioritizes the

specified modifiers (input) to be considered based on

CC. The example of sorted modifier (output) for

Abdominal pain CC is Pain, Hemodynamic,

Respiratory, Hypertension and Presenting complaint.

The following pseudo code describes a piece of

the Sorting inference knowledge which will show

the order of priority Specified modifiers for a CC

when their order is known.

READ CC from chief complaint list

READ list of specified modifier

IF CC is X

THEN OUTPUT list of specified sorted

modifiers for X

However, in triage decision making, the

modifiers of certain CC are not specific. In such

situation, the principle of emergency care will be

applied to sort the modifiers. The principle of

emergency care highlights the element of the

modifiers represented by the acronym of ABCD

(A=airway; B=breathing, C= circulation;

D=dysfunction of central nervous system) to relieve

suffering and to prevent further deterioration of the

illness. This knowledge has been used in practice to

prioritise the modifiers.

3.2.4 Evaluation Inference

The inputs of Evaluation inference consist of two

components: a set of data and a norm. Data is

evaluated with a norm based on evaluation criteria.

The truth value is derived, which indicates whether

TaskKnowledgeModelforTriageDecision-Support

265

or not the data complies with the norm.

In the triage decision-support task, the

Abstracted modifiers values and the Fetched norm of

modifier X are data and norm, respectively. The

evaluation knowledge consists of rules that examine

whether the abstracted values comply with the norm

in hand. A truth value of 1 indicates that the Fetched

norm of modifier X fulfils the Abstracted modifiers

values. In case of failure (0), the next norm of

modifier X is considered. The following pseudo

code describes a piece of the evaluation inference

knowledge.

READ norm i of modifier X

READ abstracted value modifier X

IF (norm i of modifier X ==

abstracted value modifier X) is

true

THEN truth value = 1

GET triage level value

GET abstracted modifier X value

ELSE truth value = 0

...

3.2.5 Verification Inference

The Verification inference is used to test a

description of the system based on certain

hypothesis. A system description represents a

condition or event that has to be tested and the

output for this task is a truth value, which indicates

whether the system has passed the test. The violation

is also an output.

The Verification inferences in triage decision-

support verify three different events: Verify if a

norm value indicates a critical condition, Verify if

all norms of modifier X have been considered and

Verify if all modifiers have been considered. The

following paragraph discusses the first event.

This event is to verify whether the determined

triage level indicates the critical condition which is

triage level I and II. The truth value from this

verification also determines whether there is a need

to consider the other norms if the verification fails. If

the verification succeeds, the following Match

inference takes over. The following pseudo code

describes part of the Verify critical condition

inference knowledge.

READ triage level value

IF triage level value = 1

THEN GO TO Match modifier

ELSE GO TO Verify all norms

3.2.6 Matching Inference

The input for the Matching inference is an abstracted

case description, which describes a particular event

or entity and a set of norms represents the rules that

indicate whether the description leads to decision.

The purpose of this inference in triage decision-

support is to provide justified explanation for the

determined triage level (decision). The explanation

is based on norms that meet the lowest triage level.

The following pseudo code broadly describes the

matching inference knowledge.

FOR (norm i of modifier X ==

abstracted value modifier X) is

true

READ all norm i of modifier X

END FOR

DETERMINE lowest triage level value

READ norm i of modifier X with

lowest triage level value

PRINT norm i of modifier X with

lowest triage level value

PRINT triage level value

3.2.7 Fetch Action

Fetch is a non-cognitive action that appears in the

triage decision-support task. This action fetches a

set of norms for modifier X. The following pseudo

code describes a piece of the Fetch inference

knowledge.

READ list of specified sorted

modifiers for CC

OBTAIN number of modifiers from

specified sorted modifier for CC

...

IF specific sorted modifier i is X

THEN FETCH set norms of modifier X

4 VALIDATION OF THE TASK

KNOWLEDGE MODEL

Once the understanding of the triage decision-

support task is clear, the key resources utilised in the

triage decision-support task will serve the basis of

validation of the task knowledge model.

Subsequently, the modelling of the purposive

domain knowledge will be based on these identified

resources (Annamalai, 2006; Annamalai and

Sterling, 2003). In the ensuring paragraphs, we will

discuss these knowledge resources. Due to page

limitation, we only provide a brief description of the

key knowledge resources stated in Table 2.

Case is a composition of Patient, Clinical

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

266

judgment, History. Patient consists of patient

identity such as his name, age and gender. Clinical

judgment aggregates the objective and subjective

observations and oral history taken from patient.

The Case is abstracted to identify the Chief

Complaint (CC). A list of Chief Complaint (CC) is

provided in Habboushe’s guide (2012).

Each CC has a set of Specified and Optional

modifier associated with it. Table 3 shows the the

modifiers for CC: Altered mental status.

Table 4 shows the modifier values and the

abstracted modifier values for the modifier:

Respiratory.

Table 3: The associated of the modifiers for CC: Altered

mental status.

CC Specified modifier Optional modifier

Altered

mental

status

GCS

Presenting complaint

Bleeding

Hypertension

Pain

Mechanism of injury

Hemodynamic

Respiratory

Temperature

Mental health

Blood glucose

Table 4: Modifier values for modifier: Respiratory.

Modifier Modifier value

Abstracted

modifier value

Respiratory

Cyanosis

Single word speech

Stridor

Lethargic

Confused

Severe

Short of breath with mild

exertion rest

Speaking phrases

Significant strider with

airway protected

Moderate

Mild short of breath

Rapid breathing

No obvious work of breathing

Able to speak in sentences

Stridor without obstruction

Mild

Table 5: Set of norms for modifiers GCS, Respiratory and

Hemodynamic.

Modifier

Set of norms

Value Level

GCS

3 to 9

10 to 13 II

14 to 15 III

Respiratory

Severe I

Moderate II

Mild III

Hemodynamic

Severe end organ hypo-

perfusion

Borderline perfusion II

Upper and lower end VS III

Normal VS IV

Table 5 shows the set of norms and the triggered

triage levels for three example modifiers: GCS,

Respiratory and Hemodynamic. The terms that

feature in the modifier values such as Severe, Mild,

Hypo-perfusion, Borderline perfusion and so on will

be structured and defined in the domain knowledge

model.

5 SUMMARY AND DISCUSSION

The paper discusses the construction of the

knowledge model to support the development of a

triage decision-support system. We adapted and

extended the generic CommonKADS Assessment

task to structure the triage decision-support task

knowledge model, which consists of six inference

models and one non-cognitive action. They are:

Specification, Abstraction, Sorting, Evaluation,

Verification, Matching and Fetching. The Sorting

and Verification are new inferences that are

introduced in this task, which do not exist in the

generic Assessment task. The introduction of these

two inferences is to support the nature of triage

decision making.

Since we adapted a top-down modelling

approach to engineer our task model, the validation

of the task model involved checking with the domain

experts on the concretisation of the abstract terms,

the inference methods and the ordering, and their

input and output data and/ or knowledge resources.

In our future work, we will construct the triage

purposive domain knowledge model informed by

this task knowledge model.

ACKNOWLEDGEMENTS

This work was supported by Ministry of Education

Exploratory Research Grant Scheme, Malaysia. We

would like to thank Mohammad Hafidz bin Rahmat

for his cooperation during our initial study.

REFERENCES

Angele. J., Fensel. D., Landes. D., Studer. R., 1998.

Developing Knowledge-Based Systems with MIKE.

Automated Software Engineering, October.5(4).

Annamalai, M., Sterling, L., 2003. Guidelines for

Constructing Reusable Domain Ontologies.

Proceeding of the AAMAS'03 Conference Workshop

on Ontologies in Agent Systems, CEUR Series, Vol.

73.

TaskKnowledgeModelforTriageDecision-Support

267

Annamalai, 2006. Modelling Knowledge for Scientific

Collaboration on the Semantic Web. Ph.D Thesis,

University of Melbourne.

Chandrasekaran, B., 1986. Generic Tasks in Knowledge-

based Reasoning: High-level Building Blocks for

Expert System Design. IEEE Expert 1, Volume 3, pp.

23-30.

Considine, J., Botti, M., Thomas, S., 2007. Do Knowledge

and Experince Have Specific Roles in Triage

Decision-making?. The Society for Academic

Emergency Medicine, pp. 722-726.

Gennari, J. H., Musen, M. A., Fergerson, R. W., grosso,

W. E., Crubezy, M., Eriksson, H., Noy, N. F., Tu, S.

T., 2003. The Evolution of Protégé: An Environment

for Knowledge-based Systems Development.

International Journal of Human-Computer Studies,

January, 58(1), p. 89–123.

Gerdtz, M. F., Considine, J., Sands, N., Stewart, C. J.,

Crellin, D., Pollock, W. E., et al, 2009. Emergency

Triage Education Kit, Canberra City.

Habboushe, J., 2012. The Basics of Emergency Medicine,

A Chief Complaint Based Guide. s.l.:EMRA

Publications.

Halim, S., Annamalai, M., Ahmad, M. S., Ahmad, R.,

2011. A Conceptualisation of an Agent-Oriented

Triage Decision Support System. In: Knowledge

Technology. s.l.:Springer Berlin Heidelberg, pp.36-46.

Kingston, J., 2007. Multi-perspective Modeling for

Knowledge Management and Knowledge Engineering.

Ph. D Thesis. University of Edinburgh.

LeVasseur, S., Considine, J., Charles, A., Castle, C.,

Villanueva, E., 2001. Consistency of Triage in

Victoria's Emergency Departments: Consistency of

Triage Report. Monash Institute of Health Services

Research.

Lindow, K., Heimann, O., Adolphy, S., Hayka, H., Stark,

R., 2013. Decision-Making Support for Sustainable

Product Development. In: M. S. R. Abramovici, ed.

Smart Product Engineering. Bochum, Germany:

Springer Berlin Heidelberg, pp. 979-988.

Marcus, S. e., 1988. Automatic Knowledge Acquisition for

Expert Systems. Boston: Kluwer.

Murray, M., Bullard, M., Grafstein, E. , 2004. Revisions to

the Canadian Emergency Department Triage and

Acuity Scale Implementation Guidelines. CJEM.

Schreiber. A. T., Wielinga. B., Breuker. J., eds., 1993.

KADS: A Principled Approach to Knowledge-based

System Development. 11 ed. London: Academic Press.

Schreiber. G., Akkermans. H., Anjewierden. A., de Hoog.

R., Shadbolt. N., de Velve. W. V., Wielinga. B., 2000.

Knowledg Engineering and Management: The

CommonKADS Methodology. Cambridge: The MIT

Press.

Tarta, A. M., 2004. Task Modeling in Systems Design.

Informatica, XLIX(2), pp. 37-44.

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

268