This article was downloaded by: [University of New Mexico]On: 22 November 2014, At: 10:55Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

Quality EngineeringPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/lqen20

Application of Experton Theory in the Sensory Analysisof Cell Phone FlapsZyed Zalila a b , Anne Guenand c & Joan Martin Lopez ca Université de Technologie de Compiègne, Costech Groupe Logique Floue , Compiègne,Franceb IntelliTech (Intelligent Technologies) , Compiègne, Francec Université de Technologie de Compiègne, Design Industriel. Conception et Qualité desProduits et Processus , Compiègne, FrancePublished online: 15 Feb 2007.

To cite this article: Zyed Zalila , Anne Guenand & Joan Martin Lopez (2005) Application of Experton Theory in the SensoryAnalysis of Cell Phone Flaps, Quality Engineering, 17:4, 727-734

To link to this article: http://dx.doi.org/10.1080/08982110500251337

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Application of Experton Theory in the Sensory Analysis of Cell Phone Flaps

Zyed ZalilaUniversite de Technologie de Compiegne, Costech Groupe Logique Floue, Compiegne, France; IntelliTech (IntelligentTechnologies), Compiegne, France

Anne Guenand and Joan Martin LopezUniversite de Technologie de Compiegne, Design Industriel. Conception et Qualite des Produits et Processus,Compiegne, France

Sensory metrology, which aims to define the organolepticcharacteristics of products, is an increasingly important methodin product design. The results obtained from sensory analysisindicate which features stimulate the senses and preferencesof customers. This information enables designers to createcompetitive products (Bassereau, 2001; Lageat et al., 1999).

Expert judgment is one of the most significant featuresof sensory analysis, but it is very delicate to process. Expertsuse calibrated subjective judgments to value the sensory char-acteristics of products. They assign marks, which are gener-ally aggregated to produce the model of an ‘‘averageexpert.’’ This model does not take into account either thereliability or the confidence of each expert in his or herown judgment. To solve this problem, we have applied theExperton theory, which allows each expert to express a levelof confidence in his or her own judgment and enables allexperts’ answers to be retained.

Keywords Expert group decision; Interval-valued opinion;Subjective evaluation; Objective evaluation;Sensory metrology.

1. INTRODUCTION

Expert evaluation is essential for the sensory ana-lysis of products. Experts are people who have beensensitized and trained to analyze, determine, and quan-tify the perceived quality of a product in an objectiveway (Lageat and Durandau, 1999; McEwan et al.,2002; Schlich, 1997). Their mission is to objectivelyanalyze those features that cannot be directly measuredby sensors.

One stage of sensory analysis is to find the expertwho can objectively define product characteristics. Infact, this is almost impossible because experts arehumans who cannot be expected to be totally sure oftheir answers. For this reason, it is usually preferredto use an aggregation of all answers given by the certi-fied experts. This overall answer must be reproducible,faithful, and reliable (Piggott, 1995; SSHA, 1998).

One aim of subjective evaluation is to study howto make the experts’ judgments more objective.Kaufmann (1987) developed the theory of Expertonsfor decision making in expert groups. This theory wasenhanced by Zalila (1997) and formally extended withFatene (Fatene, 2001; Zalila and Fatene, 1998a, 1998b).

2. PROJECT

This research project focused on the use of Exper-tons to carry out a sensory analysis of various cellphone flaps. The experts had to evaluate 13 character-istics:

. 6 Acoustic characteristics—sound quality

. 6 Ergonomic characteristics—ease of single-handmanipulation of the phone and flap

. 1 Visual characteristic—visual homogeneity ofphone and flap

This project studied 14 different phone devices with aflap. The group of experts comprised one womanand five men ages 20 to 45, all of whom were frequentcell phone users.

3. METHOD

The main characteristics (descriptors) of a cellphone were determined using practical and hedonistic

Address correspondence to Zyed Zalila, IntelliTech,14 rue du Fonds Pernant, 60200 Compiegne, France. Fax:33 3 44 23 48 99. E-mail: zyed.zalila@intellitech.fr

Quality Engineering, 17:727–734, 2005

Copyright # Taylor & Francis Inc.

ISSN: 0898-2112 print=1532-4222 online

DOI: 10.1080/08982110500251337

727

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ew M

exic

o] a

t 10:

55 2

2 N

ovem

ber

2014

semantic tests carried out by consumers. These des-criptors are precise (devoid of any ambiguity), discri-minative (for a given descriptor, the range of theexperts’ marks must reflect real differences betweenthe products), exhaustive (the product must be comple-tely, not partially, described), and independent (maxi-mum information obtained using a minimum numberof words) (SSHA, 1998).

The Expertons were calculated by the method putforward by Kaufmann (1987), and the complementarycumulative curve chart was based on Zalila (1997).

The result is a Multiple Experton that defines thesensory profile of each telephone. This Experton isrepresented by complementary cumulative curves forsix experts (Figure 1).

The classic sensory profile can be obtained by theusual method of sensory analysis as shown in Figure 2(three experts).

4. DATA PROCESSING AND RESULTS

4.1 Data Processing and Results Obtained Using a

Classic Approach

The usual sensory profile method is widelyemployed to evaluate the perceived quality of a rangeof products (Lepage et al., 2000). Significant resultsare obtained, but this method results in information lossand skews the representation of the expert’s opinion.This approach does not take into account the degreeof the expert’s certainty (Langhoff et al., 2001); inaddition, the experts’ judgments are aggregated into asingle scalar number obtained by projection (average).

The information obtained from a usual sensoryprofile is determined by the position of the points cor-responding to the opinion of one expert for eachdescriptor. The experts evaluate the descriptors, andthe average of their marks is calculated for eachdescriptor to obtain a synthesis of their evaluation.To facilitate the interpretation, a profile is drawn,representing the opinion given by one expert for oneproduct on the reference judgment framework.

4.2 Data Processing and Results Obtained with the

Expertons

For each phone and each descriptor, the SingletonExperton is represented graphically by two curves:

. The upper line of the Experton (upper bounds)

. The lower line of the Experton (lower bounds)

A convex curve means that the experts have expressedan opinion that tends toward the answer ‘‘true.’’ If thecurve is concave, nearly all the experts agree to theanswer ‘‘false.’’ If the curve nears the diagonal line,each expert has made a different assessment, and thereis no consensus.

The results for phone A (Figure 1) are similar tothose obtained using the classic method. However,the Expertons provide additional information aboutthe degree of confidence and the level of agreementbetween the experts:

. The more the lines diverge, the greater the experts’uncertainty about their evaluation.

. The closer the curves are to the diagonal, the lowerthe consensus between experts.

. The more the curves drop off discontinuously, thegreater the agreement between experts (same marks).

There are several advantages of using this kind ofrepresentation:

. Easy observation of the evaluations

. Simplified data processing to compare the cellphones

. Interval-valued judgment if the expert is not sure ofhis or her valuation

. Visual representation of the level of uncertainty

In this study, a number of applications of subjec-tive evaluation were examined, including the use ofExpertons in the mono=multicriteria selection of pro-ducts evaluated by a group of experts.

4.2.1 Selection of the Best Phone According to SingleDescriptor

The results can be represented by confidence pro-files either by considering all descriptors for eachphone (Figure 1) or one descriptor for all phones(Figure 3). In the second case, all sensory characteris-tics can be viewed individually for easy qualitativecomparison of the phones based on

. The value given to the descriptor (mostly tending to‘‘true,’’ mostly tending to ‘‘false,’’ opinion disper-sion)

. The confidence attributed by the experts to thisdescriptor

This method allows the certainty of the judgments ofa group of experts in evaluating a descriptor to be

728 Z. Zalila, A. Guenand, and J. Martin Lopez

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ew M

exic

o] a

t 10:

55 2

2 N

ovem

ber

2014

Figure 1. Cell phone A. Sensory profile with Expertons.

European Edition: Application of Experton Theory 729

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ew M

exic

o] a

t 10:

55 2

2 N

ovem

ber

2014

examined, an advantage compared with more tradi-tional methods. It should be noted that when allexperts gave an identical assessment for the descriptor,the two cumulative complementary curves are super-imposed. The opposite is true when the lower andupper curve cohabit on the graph. This graphical char-acterization gives some very valuable information onthe quality of the reference judgment framework (haveall descriptors been well interpreted by the experts?)and=or enables evaluation of the degree of the experts’confidence in their own judgment (do they systemati-cally doubt their own judgment?).

The clear graphic representation makes it veryeasy to identify which Experton gathers the more‘‘true’’ or ‘‘false’’ evaluations and which phone charac-teristics appear to be the most obvious for the experts.

For example, according to the ‘‘hard, clear’’descriptor, P and J appeared to be the best phones,and B, E, and T, the worst. There was a high level ofconfidence for P and H. However, there was significantuncertainty in the judgment for D with no consensus inthe opinions of the experts.

This same qualitative conclusion can be drawnfrom quantitative analysis. The expected value is

calculated for each Singleton Experton, using either aMultiple Experton or the probability law table(Kaufmann, 1987):

For a continuous judgment scale (probabilitydensity),

8x 2 X ; E FðlxÞð Þ ¼Z 1

l¼0

l � f ðlÞdl

For a discrete judgment scale (probability distribution),

8x 2 X ; E FðlxÞð Þ ¼Xl

l � prðlÞ

To illustrate the procedure, the lower bound ofthe expected value for the Experton for phone A iscalculated for the ‘‘hard, clear’’ descriptor.

Given that the statistics for the phone A evalua-tions lead to the probability law in Table 1, the lowerbound can be calculated as follows:

0� 0:00þ 0:1666� 0:17þ 0:3333� 0

þ 0:5� 0þ 0:6666� 0:50

þ 0:8333� 0:33þ 1� 0:00 ¼ 0:64

Figure 2. Cell phone A. Classic sensory profile.

730 Z. Zalila, A. Guenand, and J. Martin Lopez

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ew M

exic

o] a

t 10:

55 2

2 N

ovem

ber

2014

Figure 3. Confidence and judgment profile for the ‘‘hard, clear’’ descriptor. The letters correspond to the 14 phones studied.

European Edition: Application of Experton Theory 731

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ew M

exic

o] a

t 10:

55 2

2 N

ovem

ber

2014

Table 2 shows the expected values calculated for allphones.

Using these values and the Pareto order, a partialorder can be defined to compare the phones against thejudgment attributed by the group of experts to the‘‘hard, clear’’ descriptor. The resulting inferior semilat-tice is given in Figure 4a. For example, it can be notedthat the phones with the highest value are P (0.75) andJ [0.72, 0.86]. This means that according to the experts,of all the cell phones, P and J have the sound closest to‘‘hard, clear.’’ However, phones B and T have anexpected value of [0.08, 0.11], which is the lowest value.In other words, their sound is not judged to be very‘‘hard, clear.’’

The expected values are often intervals instead ofscalars. Therefore, it is not always correct to state thatone phone is better than another. For example, phoneP has a more ‘‘hard, clear’’ sound than phone A: theexpected values can be compared using the definedpartial order. However, phone D [0.33, 0.61] andphone R [0.36, 0.44] cannot be compared; in otherwords, it is not possible to say which has the most

‘‘hard, clear’’ sound. The expected value for F [0.58,0.67] is often higher than that for D, but not always.So, even though it cannot be strictly stated that Fhas a sound that is more ‘‘hard, clear’’ than D, it mightbe judged as such. However, as the lower bound of H(0.67) is greater than the upper bound of D (0.61), it iscertain that H is always more ‘‘hard, clear’’ than D.

It is possible to define a total order for the phonesby giving preference to the most certain expectedvalues (i.e., the narrowest intervals). The resultingchain is given in Figure 4b.

4.2.2 Selection of the Best Phone Using n Descriptors

The same technique is employed as for the selec-tion of the ‘‘best item’’ defined by a single descriptor;however, this time more than one descriptor is used.A sample of three phones {A, B, N} was selected fromthe 14 used in the study, and three descriptors(‘‘muffled’’, ‘‘composite’’, and ‘‘solid’’) were selected.We examined which phone of {A, B, N} was the‘‘best’’ according to these descriptors.

Using a Multiple Experton describing each cellphone on the reference judgment framework, welooked for an aggregative Singleton Experton synthe-sizing a conjunctive request for the three descriptors:‘‘muffled AND composite AND solid.’’ To modelthe AND operator, we can choose for example Zadeh’sT–norm (Zadeh, 1965): xTy ¼ min (x,y), where x andy are the Singleton Expertons.

The aggregation results are shown in Figure 5.Graphically, phone B best satisfies the requirements,compared with A, which least conforms to the request.

However, this qualitative analysis can be validatedby calculating the expected values, as shown in Table 3.

There is no intersection between the resultingintervals, so a total order can be defined. B ranks first,

Table 1Probability laws for cell phone A

0 0.1666 0.3333 0.5 0.6666 0.8333 1

Lower bound 0.00 0.17 0.00 0.00 0.50 0.33 0.00Upper bound 0.00 0.00 0.17 0.00 0.33 0.50 0.00

Table 2Interval-valued expected values for the ‘‘hard, clear’’ descriptor

A B D E F G H J N O P R T Y

Lower bound 0.64 0.08 0.33 0.14 0.58 0.53 0.67 0.72 0.53 0.67 0.75 0.36 0.08 0.67Upper bound 0.69 0.11 0.61 0.19 0.67 0.58 0.67 0.86 0.61 0.72 0.75 0.44 0.11 0.75

Figure 4. Partial and total orders.

732 Z. Zalila, A. Guenand, and J. Martin Lopez

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ew M

exic

o] a

t 10:

55 2

2 N

ovem

ber

2014

followed by N and finally A. This means that thegroup of experts considered B as the most ‘‘muffledAND composite AND solid.’’ Therefore, if lookingfor a phone that simultaneously best satisfies all threefeatures, B would be chosen.

5. CONCLUSIONS

In sensory metrology, the judgments given byexperts are expected to be precise, to be reliable, andto display as little variation as possible. In general,expert judgments are dealt with in the same way asobjective measurements. The theory of Expertonstakes into account the fact that experts are humansand thus are bound to have doubts concerning theirevaluations: their judgments are considered ‘‘subjec-tively objectivated’’ (Zalila, 2002). When applying theExperton theory to sensory metrology, the experts willhave more confidence and freedom when making theirown judgments, and the results will more accuratelyand reliably fit their perceptions.

Cumulative complementary curves can be easilycalculated and provide a representation of the sensoryprofile, with the advantage of being clear, easy to use,and involve no data loss. The technique enables all theexperts’ judgments to be included with less data distor-tion. Each Experton graph may be seen as an opinionmorphology.

This study has shown that the results obtainedusing Expertons not only agree with those usually

calculated, but also contain additional information onthe experts, such as the evaluation of their confidence,an estimation of the dispersion in their judgments, ortheir understanding of the different descriptors.

This article describes the creation of Expertonprofiles and two innovative applications: the first asa guide to assist in the choice of the best product, basedon a single descriptor, and the second (an extensionof the first) to choose the product that best satisfiesseveral descriptors.

These two applications of the Expertons theorycan help satisfy the needs of companies to solve theinherent difficulties involved in the subjective evalua-tion of products by a group of experts. The advantagesof this approach are the ease of calculation, the simpli-city of interpreting complementary cumulative curves(due to their visual representation), and the rapid con-clusions that can be drawn.

Consequently, the Experton concept contributesto the definition of an optimized evaluation methodfor multicriteria group decision in the field of sensoryanalysis.

ABOUT THE AUTHORS

Dr. Zyed Zalila is the CEO-Founder of the R&Dcompany IntelliTech (Intelligent Technologies), specia-lizing in the design and edition of soft computing-based decision software for intelligent systems. Heholds an engineering degree in artificial intelligenceand a Ph.D. in fuzzy mathematics from the Universitede Technologie de Compiegne. He is the co-inventor of10 French and international patents about fuzzy auto-motive systems. Professor in fuzzy theory and its appli-cations, he has been the research adviser of 15 DEAs(postgraduate degree) and 9 Ph.Ds.

Figure 5. Conjunctive request Expertons. The letters correspond to the three phones of the request.

Table 3Expected values for each request Experton

A B N

Expected values [0.11, 0.14] [0.67, 0.72] [0.31, 0.39]

European Edition: Application of Experton Theory 733

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ew M

exic

o] a

t 10:

55 2

2 N

ovem

ber

2014

Anne Guenand is a Professor at the Universite deTechnologie de Compiegne. She focuses her researchon the integration of the perceived aspects of the pro-ducts and the representation of consumer perception inthe product design process; methods and tools for thecontrol and improvement of product developmentprocess; and semiology of the products: methods forcreativity and decision support tools.

With a DEA (postgraduate degree) from theUniversite de Technologie de Compiegne, Joan MartinLopez is an engineering consultant in rail reliability.He is currently on assignment as a safety engineer inSpain (working on a new high-speed rail betweenMadrid and Barcelona) and in Portugal (working onthe subway in Lisbon).

REFERENCES

Bassereau, J. F. (2001). Metrologie sensorielle du toucher,Association des Ingenieurs Diplomes de l’ESIM, pp.10–16.

Fatene, M. (2001). Contribution a la theorie des expertons—Sous-ensembles flous de type 2 et 3. Ph.D. thesis, D1374;Groupe Logique Floue, Universite de Technologie deCompiegne: Compiegne, France.

Kaufmann, A. (1987). Les Expertons. Paris: coll. Traite desNouvelles Technologies, Mathematiques appliqueesseries, Hermes.

Lageat, T., Durandau, C. (1999). Marketing sensoriel: laPolysensorialite des emballages. Proceedings of the 2ndMeeting on Sensorial Marketing; Paris; Eurosyn.

Lageat, T., Montet, A., Lecoq, M. (1999). Marketing sensor-iel: ou comment integrer les preferences des consommateursdans un processus de conception de produits. Proceedings ofthe 6th CONFERE ISTIA workshop; Angers, France.

Langhoff, Guiset, Guenand, A. (2001). Application de lamethode d’analyse sensorielle pour la maıtrise de la qualitepercue du telephone portable a clapet. Proceedings of the8th CONFERE, Marrakech, Morocco.

Lepage, B., Perot, A., Guenand, A. (2000). Comment satis-faire les attentes difficilement exprimables des consomma-teurs. Proceedings of the 7th CONFERE; Marseille,France.

McEwan, J. A., Hunterb, E. A., Gemertc, L. J. (2002). Profi-ciency Testing for Sensory Profile Panels: Measuring PanelPerformance. Food Quality and Preference 13(3):181–190.

Piggott, J. R. (1995). Design Questions in Sensory andConsumer Science. Food Quality and Preference6(4):217–220.

Schlich, P. (1997). CAP: Une methode et un outil de controlerapide et synthetique des performances des sujets en evalua-tion sensorielle descriptive. Proceedings of the 5th Eur-opean Days of Agro-Industry and Statistical Methods,Versailles, France, ASU.

SSHA. (1998). Sciences et Techniques Agroalimentaires ser-ies. In Depledt, F., Strigler, F., eds. Evaluation sensorielle:manuel methodologique. Paris: Lavoisier TEC et DOC.

Zadeh, L. A. (1965). Fuzzy Sets. Information and Control8:338–353.

Zalila, Z. Contribution des mathematiques du flou a l’evalua-tion de produits et a la conception centree sur l’homme:testeur virtuel, pilote virtuel & systemes intelligentsd’aide a la conduite automobile. Proceedings of theSIA conference R-2002-07, Le Style, un defi pour laTechnique, Compiegne, France, Dec 10, 2002; SIA; pp.13–40.

Zalila, Z. (1997). Les Expertons. Evaluation Subjective:Methodes, Applications et Enjeux, Paris: Les Cahiers desclubs CRIN; Club CRIN ‘‘Logique Floue’’, AssociationECRIN, pp. 52–62.

Zalila, Z., Fatene, M. (1998a). Operateurs flous en theorie desexpertons: semantique d’une decision de groupe en evalua-tion subjective, Vol. II. Proceedings of the 7th Interna-tional Conference IPMU Information Processing andManagement of Uncertainty in Knowledge-BasedSystems, July 6–10; Paris, pp. 1691–1699.

Zalila, Z., Fatene, M. (1998b). Resolution d’equations flouesen intervalles—Application aux Expertons. Proceedingsof the Fuzzy Logic and Applications Conference LFA;November 18–19; Rennes, France; Cepadues-Editions;pp. 217–228.

734 Z. Zalila, A. Guenand, and J. Martin Lopez

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ew M

exic

o] a

t 10:

55 2

2 N

ovem

ber

2014