An Integrated QoE and QoS Based Approach for Web

Service Selection
Shripad S. Joshi1, O.B.V. Ramanaiah2
Department of Computer Science and Engineering,
JNT University Hyderabad, Kukatpally, Hyderabad, India
1joshi@jntuh.ac.in
2obvramanaiah@jntuh.ac.in

Abstract - An exponential increase in the number of web value of the performance of the whole system according to
services over the last few years increases the importance of the user’s perspective.
service selection task for choosing the best among a group of web
services with similar functionalities. Most of the web service Most of the research work for web service selection has
selection approaches are service provider perspective based on been carried out using QoS based approaches in which only
non-functional properties such as performance, reliability etc. provider based QoS values are used for selection. But in real
(such attributes known as Quality of Service or QoS). But it has world scenario, users’ perspective in terms of feedback or
been observed that in any decision support issues in selection, experience of using the service plays a crucial factor in web
users’ feedback (known as Quality of Experience or QoE) plays a service selection. These approaches are based on the concept of
crucial role. In this paper, an integrated model has been proposed, QoE.
based on both QoE and QoS, where the best service selection is
made based not only on current QoS values of the services but In this paper, An approach has been proposed for web
also users’ past experience of using them. Further, the case study service selection based on both QoS and QoE. The approach
has been provided and the results have been analyzed by inducing selects the service not only based on current quality attributes
users’ ratings as a QoE factor along with QoS parameters. The of services but also considers the historical usage experience
results show that the proposed approach, augmented by user’s and performance perceived by the user. In this approach, online
feedback, improves the quality of selection. ratings reflecting user experience feedback on web services’
QoE attributes are considered. Further, the hypothetical
Keywords - Service selection, user review, quality of service, quality example of selection of Book Buy Service with our approach is
of experience, block nested loops, TOPSIS given. The results show that when users’ feedback is
considered, the ranking of web services gets changes. The
I. INTRODUCTION work in this paper shows the ranking order at different
scenarios like considering only QoS, only QoE, and
A. Overview considering different weightage to QoE and QoS.
Web Service became a popular technology that describes a
standardized way of integrating Web-based applications using The remainder of this paper is organized as follows: Section
the XML, SOAP, WSDL, and UDDI open standards over an II reviews the referred literature in related area. Section III
Internet protocol backbone. XML is used to tag the data, SOAP gives an overview of quality in web services. Section IV
is used to transfer the data, WSDL is used for describing the presents an overview of our approach along with a case study
services available and UDDI is used for listing what services and finally Section V concludes the paper and explores the
are available. A web service is a software component which future work.
supports interoperability machine-to-machine interaction over
a network. So, a Web service is used to make the application B. Contributions
platform and technology independent. • An integrated web service selection approach using
In the recent years, it has observed that there is a both QoE (users’ feedback) and QoS (provider
tremendous growth in the number of published web services. perspective) has been proposed.
The effect of this rapid growth put forth a requirement of • BNL (Block Nestd Loops) algorithm has been used to
choosing the best service among the discovered services, compute skyline services to reduce the search space for
having a mix of some low quality services. The functional selection mechanism.
requirements in user request may match with huge number of
services providing similar functionality. In such scenario, there • The TOPSIS method of multi-criteria decision making
is requirement of a different criteria for service selection. Non- (MCDM) approach is used for ranking.
functional attributes [1] (QoS attributes) can be exploited for
• The experimental results show that the proposed
further service selection among similar functionality services.
approach improves the quality of the selection process
Competition from providers is based mainly on Quality of by exploiting users’ feedback in the form of user
Service (QoS) values. There is a need of another differentiating ratings.
factor rather than QoS alone under the user’s perception. The
factor is known as Quality of Experience (QoE), a quantified

978-1-5090-5515-9/16/$31.00 ©2016 IEEE

 • It gives a requester the best set of services as per the III. QOS AND QOE FOR WEB SERVICES
requirements which reduces chances of unavailability of In general, Qos and QoE are described as the two main
the required service. research trends for approaching decision support issues. There
are two perspectives to look at the quality of web services,
II. LITERATURE SURVEY service provider perspective and user perspective. Service
The numbers of approaches have been proposed to solve provider describes the quality attributes in terms of non-
the web service selection problem. The major classes [2] based functional attributes while user’s view will be in terms of their
on different criteria are the multi-objective optimization, mono- feedback or experience with a service.
objective optimization and hybrid optimization. The problem
of web service selection based on QoS [3, 4 and 5] has been A. Quality of Service (QoS)
addressed in many research works which proposed a QoS deals with performance aspects of physical systems. It
comprehensive representation of service request and refers to non-functional properties web services such as
implemented a selection algorithm to achieve an optimized performance, reliability, availability etc. QoS parameters are
composition. The modal based on mono-objective class using used for selecting web services, which do not necessarily
global selection[6,7] and local selection[8,9] have been reflect the user’s satisfaction towards a particular web service.
discussed in which former, better for global constraint Table I shows a set of common QoS attributes for web services
satisfaction, is of exponential complexity while later which can [26].
not manage global constraints, is of linear complexity
There are many works based on skyline computation. A set TABLE I. QOS OF WEB SERVICES
of database techniques [2, 11] in skyline computation for
QoS Definition
optimization have been discussed. One of the methods for
skyline computation is BNL (Block Nested Loops) [10]. Many Response time execution time(s) +waiting time(s)+network latency
techniques for getting the top-k dominant skylines [12, 13] Reliability failure rate(s)
based on user’s preferences and the dominance relationship
have been proposed. Index and Bitmap- based algorithms Availability uptime(s) / (uptime(s) + downtime(s))
based model [14] for returning skyline result progressively Price Execution fee for a request
have been proposed. [15] used NN( Nearest Neighbour) and Usage limits Number of requests per day
BBS ( Branch and Bound Skyline) based on R-trees indexing
structure. Security Authentication model, SSL support

Various approaches of MCDM (Multi-Criteria Decision

Making) are AHP (Analytical Hierarchy Process)[16], ANP B. Quality of Experience (QoE)
(Analytical Network Process), Promethee, TOPSIS(Technique
QoE is a subjective measurement which reflects user’s
for Order Preference by Similarity to an Ideal Solution)[17]
experience with a service. In contrary to QoS, QoE reflects
have been used for ranking web services. The TOPSIS based
quality from the users’ point of view. The QoE concept has
web service selection has been given by [18] in which an
emerged in this field mainly with basic motivation that QoS is
extension of TOPSIS i.e. WSCA- Topsis for better QoS
not powerful enough to fully express everything. Nowadays,
parameter specification and aggregation has been proposed.
most of the web resources are being reviewed and rated by the
The AHP approach [16] has been used in [19] to build
users. Mostly, users express their experience via online reviews
hierarchical structuring for sorting the services.
or ratings to reveal their satisfactions and disappointments
Many have used hybrid method of ranking like ANP with about services. The reviews are written in natural language and
Promethee [20], ANP for calculating weights of each QoS the user ratings are expressed in Mean Opinion Score (MOS)
attributes and Promethee for ranking the services. Similar for rating overall quality. The following five-level scale for
hybrid approach [21] which is combination of three methods rating overall quality is used: 5 for excellent, 4 for good, 3 for
like Skyline method for eliminating non dominant services, fair, 2 for poor and 1 for bad.
AHP to Weight QoS attributes and Promethee for ranking. In
A set of common QoE parameters for web services that can
[22], author combined Vikor with AHP. The authors of [23]
be considered are likelihood to recommend, likelihood to
proposed a model that uses Skyline approach followed by
renew, usability, availability, support, online training and
BWM (Best Worst Method) for weight assignment to QoS
implementation.
attributes and Vikor for ranking web services.
In the literature, there are QoE based approaches like [24, IV. THE INTEGRATED APPROACH
25] where authors used NLP technique for extracting QoE
parameters from users’ online reviews and used for service A. The Approach
selection.
The basic objective of the approach is to select the best
service among the similar functionality services based on past
users experience and non-functional attributes. Fig. 1 shows the
 Discovered Web Services

s1 s3 s5 s6 s11 s12 s15 s16

Step 1
Discovered Web Services
s4 s2 s10 s9 s7 s8 s13 s14

Skyline User Constraints Dominant Web Services

Approach
Step 2 s16
s3 s9 s6 s11 s8

Dominant Web Services

QoE Ranking QoE Ranking QoS Ranking

QoS Ranking
Step 3 by by
TOPSIS TOPSIS s8 s11 s16 s9 s3 s6 s8 s9 s16 s6 s11 s3

Wqoe Wqos
QoES Ranking Wqoe = Wqos = 0.5
Step 4 QoES Ranking

s8 s16 s9 s11 s6 s3

Best Web Services Best Web Services’ order

s8 s16 s9 s11 s6 s3

Fig. 1. The integrated approach based on QoE and QoS for Web Service Selection

integrated approach based on QoS and QoE for web service mentioned in this step is to give opportunity to user whether to
selection. give importance to the historical perspective i.e. QoE ranking
or current QoS by service provider i.e. QoS ranking. If user
The approach has the following steps as shown in the Fig. 1. mentions Wqoe = 0.5 and Wqos = 0.5 means the user wants to
give same importance to QoE and QoS ranking. The rank of
Step 1: Service Discovery - The discovered web services services in QoES ranking i.e. Rqoes (Wi), is calculated by the
are the selected services from service registry that meet the following formula:
user requirements by functional parameters. There may be
more number of services selected in this step.
Step 2: Filtering with Skyline Approach - The skyline Rqoes (WSi ) = Wqoe * Rqoe (Wi ) +Wqos * Rqos (Wi )
method is used on the selected services in step 1 to reduce the where Wqoe + Wqos = 1
number of candidate web services to be considered during the
selection process based on user constraints. It eliminates the
dominated web services and keep only dominant ones. where Rqoe (Wi) and Rqos (Wi) are ranks calculated by
QoE ranking and QoS ranking. Wqoe and Wqos , ( 0<= Wqoe
Step 3: QoE and QoS Ranking - The dominant web services <= 1 ) and ( 0<= Wqos <= 1 ), are the weightage given to the
obtained in step 2 are applied with the TOPSIS ranking historical perspective i.e. QoE ranking and current QoS by
algorithm based on QoE and QoS of the services. The QoE service provider i.e. QoS ranking respectively.
ranking gives the best services considering the user feedback or
experience and the QoS ranking gives the best services
exploiting the current QoS values of the services.
Step 4: The Integrated Approach - Depending on the
weights given to QoE ranking and QoS ranking, the QoES
ranking gives the best among all the services. The weights
 TABLE II. DISCOVERED BOOK BUY SERVICES WITH QOE AND QOS ATTRIBUTES

QoS QoE
Service Response Price Reliability Availability Likelihood Usability Support
ID Time ($) (%) (%) to
(ms) Recommend
s1 9 60 11 67 5 3.2 4
s2 28 40 93 67 4.6 4.9 3.4
s3 95 19 59 68 3 4.2 3.2
s4 15 21 54 26 3.8 2.4 4
s5 19 40 53 61 2.6 2 5
s6 47 96 83 96 4 1 5
s7 7 63 16 17 5 1.4 4
s8 84 24 22 98 4.6 4.3 4
s9 94 75 12 19 3.4 4.5 5
s10 89 28 20 40 3.2 4.6 5
s11 74 86 19 77 4 3 4.6
s12 18 41 46 11 5 3.8 4.4
s13 45 36 64 38 4.3 2.6 5
s14 82 65 71 86 4.6 4 5
s15 38 47 55 85 4.4 4.2 4.6
s16 64 32 75 75 3.8 4.6 3
s17 21 35 12 84 3.9 3.4 3.8
s18 31 91 78 26 3.2 3.2 2.6
s19 41 52 43 16 4 4 4
s20 97 56 40 66 3 5 5
s21 16 52 98 67 4.7 5 4.2
s22 66 96 28 40 3.5 4.6 2.4
s23 73 77 53 71 4.9 3 2
s24 79 25 73 59 4.2 3.8 1
s25 35 59 71 91 2.4 2.6 3.5
s26 38 41 53 43 2 4 4.9
s27 74 76 77 74 1 2 4.2
s28 40 69 62 95 1.4 1.2 2.4
s29 17 40 60 80 4.3 3.6 4
S30 32 54 88 86 4 3.2 4.7

recommend, usability, availability, support. The static QoS by

B. Case Study provider are considered and the QoE values for all the QoE
In this case study, a hypothetical example of selection of a attributes for each Book Buy Service (WSi) can be calculated
Book Buy Service is considered. The numbers of Book Buy by
Services discovered are considered to be 30. TABLE II shows
the services along with its QoS and QoE values. The QoS 1 n
attributes are Response Time, Price, Reliability and Availability QoEi (WSj ) = *  qi
and the QoE attributes considered are likelihood to n i=1
where qi (i = 1...m) is value of ith quality rating in the MOS TABLE V. SELECTED SERVICES BY BNL SKYLINE APPROACH
ON QOE
scale of 1-5 and m is number of quality attributes; WSj is jth
web service (j = 1...n) and n is number of web services.
Service ID Likelihood to Usability Support
1. Applying Skyline Approach: The user specified Recommend
constrains on QoS and QoE are as given in TABLE III for s1 5 3.2 4
Skyline approach which extracts only dominant services from s7 5 1.4 4
the discovered services. TABLE IV gives the dominant
services extracted by applying BNL skyline approach on QoS s12 5 3.8 4.4
data of TABLE II. Similarly, the BNL skyline approach on s20 3 5 5
QoE data of TABLE II extracts the dominant services as given
in TABLE V. s21 4.7 5 4.2

2. Ranking the Services using TOPSIS: The set of services

TABLE III. USER SPECIFIED CONSTRAINTS ON QOS AND QOE selected by BNL Skyline approach on QoS and QoE given in
FOR SKYLINE APPROACH TABLE VI forms the input for the next stage. The selected
services and the weight of each QoS attributes given in TABLE
Constraint Description VII, are supplied as input to TOPSIS method of ranking (QoS
QoS Constraint Price should be less and reliability should be more ranking) . The ranks calculated after QoS Ranking, for the
QoE Constraint Likelihood to recommend and usability should be given set is shown in TABLE IX.
more
Similarly, The TOPSIS approach of ranking is applied to
the selected services on QoE (QoE Ranking) with the weight of
each QoE attributes as given in TABLE 8. The final QoE
TABLE IV. SELECTED SERVICES BY BNL SKYLINE APPROACH Ranking values are as shown in TABLE 10.
ON QOS

3. QoES Ranking: Using the ranks calculated by QoS

Service ID Response Price Reliability Availability Ranking and QoE Ranking, depending on the weightage given
Time ($) (%) (%) to QoS (Wqos) and QoE (Wqoe), the final QoES ranks are
(ms)
calculated by using formula given in Step 4. TABLE 11 shows
s2 28 40 93 67 the QoES ranks for different values of (Wqoe, Wqos). From
s3 95 19 59 68 the Fig. 2, the conclusion can be made that the ranking is
affected when users’ historical experience is given more
s21 16 52 98 67
weightage than the provider specific QoS. Service s2 is ranked
highest considering only provider perspective (QoS) while
service s21 is ranked highest in users’ feedback perspective.
Similar change of ranking can be seen in services s3 and s12.

TABLE VI. SELECTED MINIMAL SET OF SERVICES FOR RANKING BY QOS AND QOE

QoS QoE
Service ID Response Price Reliability Availability Likelihood Usability Support
Time ($) (%) (%) to
(ms) Recommend
s1 9 60 11 67 5 3.2 4
s2 28 40 93 67 4.6 4.9 3.4
s3 95 19 59 68 3 4.2 3.2
s7 7 63 16 17 5 1.4 4
s12 18 41 46 11 5 3.8 4.4
s20 97 56 40 66 3 5 5
s21 16 52 98 67 4.7 5 4.2
 4. QoES Ranking: Using the ranks calculated by QoS TABLE VIII. THE WEIGHTS GIVEN TO EACH ATTRIBUTES
OF QOE
Ranking and QoE Ranking, depending on the weightage given
to QoS (Wqos) and QoE (Wqoe), the final QoES ranks are QoE Likelihood to Usability Support
calculated by using formula given in Step 4. TABLE 11 shows Recommend
the QoES ranks for different values of (Wqoe, Wqos). From Weight 0.5 0.4 0.1
the Fig. 2, the conclusion can be made that the ranking is
affected when users’ historical experience is given more
weightage than the provider specific QoS. Service s2 is ranked
highest considering only provider perspective (QoS) while TABLE IX. RANKS CALCULATED BY QOS RANKING
service s21 is ranked highest in users’ feedback perspective. Service ID s1 s2 s3 s7 s12 s20 s21
Similar change of ranking can be seen in services s3 and s12.
Rank 0.23 0.77 0.61 0.21 0.45 0.30 0.69

TABLE VII. THE WEIGHTS GIVEN TO EACH ATTRIBUTES

OF QOS
TABLE X. RANKS CALCULATED BY QOE RANKING
QoS Response Time Price Reliability Availability
(ms) ($) (%) (%) Service ID s1 s2 s3 s7 s12 s20 s21

Weight 0.1 0.4 0.4 0.1 Rank 0.62 0.86 0.53 0.39 0.73 0.61 0.91

TABLE XI. QOES RANKING FOR DIFFERENT VALUES OF WQOE AND WQOS

Weight (Wqoe,Wqos)
Service ID (0,1) (0.1,0.9) (0.2,0.8) (0.3,0.7) (0.4,0.6) (0.5,0.5) (0.6,0.4) (0.7,0.3) (0.8,0.2) (0.9,0.1) (1,0)
s1 0.23 0.269 0.308 0.347 0.386 0.425 0.464 0.503 0.542 0.581 0.62
s2 0.77 0.779 0.788 0.797 0.806 0.815 0.824 0.833 0.842 0.851 0.86
s3 0.61 0.602 0.594 0.586 0.578 0.57 0.562 0.554 0.546 0.538 0.53
s7 0.21 0.228 0.246 0.264 0.282 0.3 0.318 0.336 0.354 0.372 0.39
s12 0.45 0.478 0.506 0.534 0.562 0.59 0.618 0.646 0.674 0.702 0.73
s20 0.3 0.331 0.362 0.393 0.424 0.455 0.486 0.517 0.548 0.579 0.61

Weights (Wqoe,Wqos) Vs QoES

1
0.9
0.8
0.7 s1
0.6
s2
0.5
s3
0.4
QoES

0.3 s7
0.2 s12
0.1 s20
0 s21
)

)
)

)
.9

.8

.7

.6

.5

.4

.3

.2

.1
,1

,0
,0

,0

,0

,0

,0

,0

,0

,0

,0
(0

(1
.1

.2

.3

.4

.5

.6

.7

.8

.9
(0

(0

(0

(0

(0

(0

(0

(0

(0

(Wqoe,Wqos)

Fig. 2. The graph (Wqoe, Wqos) Vs QoES

 [9] F. Li, F. Yang, K. Shuangand S. Su, “Q-Peer: A Decentralized QoS
V. CONCLUSION Registry Architecture for Web Services”, in Proceedings of the
International Conference on Service Computing, Pages 145-156, 2007.
In this paper, the integrated approach which exploits users’ [10] Stephan Börzsönyi, Donald Kossmann, and Konrad Stocker, “The
feedback as well as provider’s QoS values for selection of the Skyline Operator”,in Proceedings of the 17th International Conference on
best services according to specific criteria. The Skyline method Data Engineering, Pages 421-430 IEEE Computer Society Washington,
is applied on discovered services by applying user’s DC, USA ,2001.
preferences on QoE and QoS to reduce the search space. The [11] Q. Yu, A. Bouguettaya, “Foundations for Efficient Web Service
ranking based on QoS and QoE using TOPSIS has been carried Selection”, Springer Science+Business Media, 2010.
out on the obtained skylined services. Finally, the QoES [12] D. Skoutas, D. Sacharidis, A Simitsis, V. Kantere, and T. K. Sellis, “Top-
dominant Web Services under Multi-criteria Matching”, in EDBT, pages
ranking, a combined weight based ranking, is performed. The 898-909,2009
result shows the improvement in selection of the best service.
[13] D. Skoutas, D. Sacharidis, A Simitsis, and T. K. Sellis, “Ranking and
The high ranked service will be the best in terms of QoE and/or Clustering Web Services using Multicriteria Dominance Relationship”,
QoS based on user preference of selection. When users’ IEEE Transactions on Service Computation, 3(30, pages 163-177, 2010.
feedback is not considered, then the best service based on [14] Tan, Kian-Lee et.al. , “Efficient Progressive Skyline Computation”, in
provider’s QoS values will be selected. Further a study can be VLDB, pages 301-310, 2001.
carried out to see the effectiveness of the approach in service [15] D. Papadis, G. Fu. And B. Seeger, “Optimum Year and Progressive
composition. Algorithm for Skyline Queries”, in SIGMOD’03, pages 467-478, New
York, USA, ACM,2003.
[16] T.L. Saaty and L. Vargas, “ Fundamentals of Decision Making and
REFERENCES Priority Theory with the Analytic Hierarchy Process”, RWS Publications
[1] H. Q. Yu, and S. Reiff-Marganiec, “Non-functional Property based 2000
service selection: A survey and classification of approaches”, in Non [17] C.l. Hwang, Y.J. Lai, and T.Y. Liu, “ A New Approach for Multiple
Functional Properties and Service Level Agreements in Service Oriented Objective Decision Making”, Volume 20, Issue 8, Oct 1993, pages 889-
Computing Workshop co-located with The 6th IEEE European 899
Conference on Web Services, Pages 12 – 14, Nov 2008, Ireland, and [18] L.C. Zhang, H, Zou and F.C. Yang, “Web service Composition algorithm
Dublin. based on TOPSIS”, JCUPT, Elsevier, Volume 18, Issue 4, August 2011,
[2] M. Alrifai, D. Skoutas, T. Risse, “ Selecting Skyline Services for QoS- pages 89-97.
based Web Service Composition”, Pages 26-30, in ACM WWW [19] V.X.Tran, H, Tsuji and R. Masuda, “A New QoS Onto;ogy and its QoS-
2010,Raleigh,North Carolina, USA. based Ranking Algorithm for Web Services”, Simulation Modeling
[3] G. Canfora, M. Di Penta, R. Esposito, M.L. Villani, “An Approach for Practice and Theory, Volume 17, Issue 8, September 2009, pages 1378-
QoS-aware Service Composition based on Genetic Algorithms”, in 7th 1398.
annual conference on Genetic and evolutionary computation (GECCO), [20] R. Karim, C. Ding and C.H. Chi, “An Enhanced PROMETHEE Model
pages 1069-1075, 2005, ACM Press. for QoS-based Web Service Selection”, IEEE ISCC 2011,pages 536-543.
[4] M. Rathore and U. Suman, “A Quality of Service Broker Based Process [21] A. Ouadah, k. Benouaret, A. Hadjali and F. Nader, “ Combining Skyline
Model for Dynamic Web Service Composition” , Science Publications, and Multi-Criteria Decision Methods to Enhance Web Service Selection”
Journal of Computer Science, USA, Vol. 7, No. 8, pp. 1267-1274, 2011. IEEE ISPS 2015 Algiers.
[5] D.A. Menasce, “QoS Issues in Web Services”, IEEE Internet Computing [22] M. Khezrian, W.M.N. Wan Kadir, S. Ibrahim and A. Kalantari, “A Hybrid
6(6), 2002. Approach for Web Service Selection” IJCER, Volume 2, pages 190-198,
[6] D. Ardanga and B. Pernici, “Global and Local QoS Constraints 2012
Guarantee in Web Service Selection”, in Proceedings of the IEEE [23] Walid Serrai, Abdelkrim Abdelli, Lynda Mokdad and Youcef Hammal,
International Conference on Web Services, Pages 805-806, Washington, “An Efficient Approach for Web Service Selection” in ISCC
DC, USA,2005,IEEE Computer Socity. 2016:PEDISWESA, IEEE,2016.
[7] D. Ardanga and B. Pernici, “Adaptive Service Composition in Flexible [24] B. Upadhyaya, Y. Zou, I. Keivanloo, and J. Ng , “Quality of Experience:
Processes”, IEEE Transactions on Software Engineering, 33(60, pages What users Say About Web Services?”, 21th IEEE International
369-384,2007 Conference on Web Services (ICWS), June 27 - July 2, 2014, Alaska,
[8] B. Benatallah, Q.Z.Sheng, A.H.H. Ngu and M. Dumas, “ Declarative USA.
Composition and Peer-to-peer Provisioning of Dynamic Web Services”, [25] B. Upadhyaya, Y. Zou, I. Keivanloo, and J. Ng , “Quality of Experience:
in Proceedings of the International Conference on Data Engineering, User’s Perception about Web Services”, in IEEE Transactions on Services
Pages 297-308, Washington, DC, USA, 2002.IEEE Computer Society. Computing, 2013.
[26] W. Al-Masri and Q.H. Mahmoud, “Investigating Web Services on the
World Wide Web”, in International World Wide Web Conference, 2008.