Knowing the past to Plan for the Future

An In-depth Analysis of the First 10 Editions of the WEBIST Conference

Giseli Rabello Lopes

, Bernardo Pereira Nunes

, Luiz Andr

e P. Paes Leme

Terhi Nurmikko-Fuller

and Marco A. Casanova

Computer Science Department, Federal University of Rio de Janeiro, Rio de Janeiro, RJ, Brazil

Department of Informatics, Pontiﬁcal Catholic University of Rio de Janeiro, Rio de Janeiro, RJ, Brazil

Computer Science Institute, Fluminense Federal University, Niter

oi, RJ, Brazil

Oxford e-Research Centre, Oxford University, Oxford/OX1 3QG, U.K.

Keywords:

Conference Analysis, Statistical Analysis, Bibliometrics, Social Network Analysis, WEBIST Analysis,

Linked Data.

Abstract:

Over the last ten years, members of the WEBIST community have dedicated their time and efforts to face

a number of research challenges, making signiﬁcant advances in Information Systems and pointing to new

directions for innovation and learning. After ten successful WEBIST conferences and several scientiﬁc pub-

lications, an extensive analysis of the WEBIST conferences has been carried out (involving authors, publi-

cations, conference impact, topics coverage, community analysis and other aspects) to possibly assist us to

further advance Information Systems. Thus, in this paper, we present an in-depth analysis of the last ten

WEBIST conferences based on social network analysis, bibliometrics and statistical measures and describe a

Web-based application built on top of tripliﬁed datasets to interactively explore the ﬁndings and possibly assist

the Information Systems community to reveal new directions.

1 INTRODUCTION

Data analysis has shown its utility for conferences,

detecting related research groups, topics of interest,

impact of authors and publications in a given ﬁeld,

among others. An example of this is the analy-

sis of a group of four conferences in the ﬁeld of

Human-Computer Interaction (HCI), conducted by

Henry et al. (2007). It was based solely on meta-

data information of publications (such as authors and

keywords) and was capable of providing valuable in-

sights into authors’ behaviours and research topics

investigated in HCI in the last two decades. Blan-

chard (2012) presented a ten-year longitudinal study

over Intelligent Tutoring Systems (ITS) and Artiﬁ-

cial Intelligence in Education (AIED) ﬁelds. He fo-

cused on the analysis of potential cultural biases of

the American Psychology Association (APA) in the

ITS and AIED ﬁelds. Chen et al. (2009) presented

a visual analytic approach to the study of scientiﬁc

discoveries and knowledge diffusion. Their analysis

focused on the identiﬁcation of co-citations clusters

where they were classiﬁed and used to understand

how astronomical research evolved between 1994 and

1998. Another example following the same line was

conducted by Gasparini et al. (2013). In their study,

they were able to identify central authors and institu-

tions in the HCI ﬁeld, as well as important trends and

topics. As for the Information System (IS), Posada

and Baranauskas (2014) analysed a sister event called

International Conference on Enterprise Information

Systems (ICEIS). They built a roadmap of the IS ﬁeld

based on paper titles and authors from the last three

years in ICEIS, and for the last eight years of selected

papers published in a Springer series on IS. Chen et al.

(2007) performed citation analysis of all papers pub-

lished in the International Conference on Conceptual

Modeling (ER) between 1979 and 2005. The anal-

ysis conducted by the aforementioned communities

opened up a wide range of opportunities for research

agendas and trends as well as supporting the domains

introspective analysis.

This paper aims at extending previous analyti-

cal methods and providing a comprehensive social

analysis of the community of WEBIST. Recently,

Zervas et al. (2014) presented a study on research

collaboration patterns via co-authorship analysis re-

garding Technology-enhanced Learning ﬁelds. Sim-

431

Rabello Lopes G., Pereira Nunes B., P. Paes Leme L., Nurmikko-Fuller T. and A. Casanova M..

Knowing the past to Plan for the Future - An In-depth Analysis of the First 10 Editions of the WEBIST Conference.

DOI: 10.5220/0005447704310442

In Proceedings of the 11th International Conference on Web Information Systems and Technologies (WEBIST-2015), pages 431-442

ISBN: 978-989-758-106-9

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

ilar analyses were conducted by Procopio Jr. et al.

(2011) regarding Databases ﬁelds and by Cheong and

Corbitt (2009) regarding Information Systems ﬁelds

(analysing the Paciﬁc Asia Conference on Informa-

tion Systems). The analysis of co-authorships in re-

search communities can reveal strong research groups

in the area and also enable the creation of links over-

time between different groups. Apart from presenting

the analysis of the last ten editions of WEBIST, we are

also concerned with the publication of the results in a

format where they can be replicated and reused in fur-

ther analysis. For this, we have borrowed Batista and

Loscio’s approach (Batista and Loscio, 2013) where

they used Linked Data (LD) principles to publish con-

ference data.

In this paper we present an in-depth and thor-

ough analysis of the ﬁrst ten editions (2005-2014)

of the WEBIST conference. To brieﬂy summarise,

WEBIST has brought together researchers and pro-

fessionals to develop and advance the IS ﬁeld. So

far, it has already attracted 2,867 researchers and pro-

fessionals from several institutions as well as pub-

lished 1,449 papers, which in turn are being cited

by other researchers in IS and other ﬁelds. More-

over, the conference currently has ﬁve main tracks

that cover a wide range of aspects involving IS: Inter-

net Technology, Web Interfaces and Applications, So-

ciety, e-Business and e-Government, Web Intelligence

and Mobile Information Systems.

The analysis presented in this paper relies on

techniques borrowed from social network analysis

(Wasserman and Faust, 1994), bibliometrics and tra-

ditional statistical measures. We also created a Web-

based application that enables users to interactively

explore WEBIST data. As the WEBIST data is pub-

lished following LD principles (Berners-Lee, 2006),

we also provide a SPARQL endpoint where other re-

searchers can extend our analysis. The importance

of such analysis goes beyond the analysis and possi-

ble interpretation of data and represents a milestone

achieved by the IS and WEBIST community so far.

The remainder of this paper is organised as fol-

lows. Section 2 overviews metrics and measures used

in the analysis of the last ten WEBIST conferences.

Section 3 details the extraction, enrichment and pub-

lication process of raw WEBIST data into RDF data

and presents a visualisation tool speciﬁcally created to

manipulate and possibly assist users in ﬁnding new re-

search groups, topics and insights. Section 4 presents

several analysis conducted with the WEBIST tool. Fi-

nally, Section 5 concludes the work with remarks and

future directions.

2 BACKGROUND

This section provides the necessary background infor-

mation required to understand the analysis conducted

with the data produced over the last decade by the

WEBIST community. We review metrics and meth-

ods of statistical analysis, social network analysis and

bibliometric indices.

2.1 Classical Statistical Measures

Pearson’s Correlation Coefﬁcient (Rodgers and

Nicewander, 1988), often denoted by the letter r, mea-

sures the strength and direction of the linear correla-

tion between two variables X and Y . Pearson’s coef-

ﬁcient (see Equation 1) can be deﬁned as the covari-

ance of the variables divided by the product of their

standard deviations to measure their dependence:

r =

∑

i=1

−

X)(y

−

Y )

∑

i=1

−

∑

i=1

−

Y )

(1)

An r value between +1 and -1 indicates the degree

of linear dependence between X and Y , r=1 indicates

a total positive correlation between the two variables

and, ﬁnally, r=-1 indicates a total negative (inverse)

correlation. For instance, as X values increase, Y val-

ues linearly decrease.

Lorenz Curve (Gini, 1912) represents the cumulative

distribution of a probability density function. Such

a function is built as a ranking of the members of

the population disposed in ascending order of the

amount being studied. The percentage of individu-

als is plotted on the x-axis and the percentage of the

variable values on the y-axis. The distribution is per-

fectly equalitarian when every individual has the same

variable value; a 45-degree line represents the per-

fect equality. On the other hand, the perfectly un-

equal distribution is the one in which only one in-

dividual has all the variable value, the curve is y=0

for all x<100%, and y=100% when x=100%, known

as the perfect inequality line. This curve was ini-

tially created to study the social inequality of wealth

and income distributions for a population, but can be

applied to analyse other distributions (Lopes et al.,

2012). We used the Lorenz curve (Section 4) to study

the distribution of papers by author.

Gini Coefﬁcient (Gini, 1912) is a measure of statisti-

cal dispersion indicating the inequality among values

of a frequency distribution. It is graphically repre-

sented as the area between the perfect equality line

and the observed Lorenz curve.

WEBIST2015-11thInternationalConferenceonWebInformationSystemsandTechnologies

432

Robin Hood Index (Hoover, 1941), also called

Hoover index, is used to measure the fraction of the

total variable value that must be redistributed over

the population to become a uniform distribution. It

is graphically represented as the longest vertical dis-

tance between the Lorenz curve and the perfect equal-

ity line.

2.2 Social Network Analysis

Before introducing social network metrics and con-

cepts (Wasserman and Faust, 1994; Freeman, 1979;

Hoser et al., 2006; Marsden, 2002; Newman, 2001,

2003), it is convenient to represent a social network

as a graph structure G = (N, E), where N is the set

of nodes, where n

∈ N represents an actor of the net-

work, and E is the set of edges, where e

∈ E repre-

sents a relational tie between a pair of actors.

Density is calculated as the number of the actual ex-

isting edges of a graph, divided by the maximum

number of edges the graph can have. A density value

equal to 1 indicates an entirely connected network

while 0 indicates a disconnected network. Consider-

ing an undirected graph where the possible number of

connections between each two nodes is 1, the density

can be calculated as:

D =

2|E|

|N|(|N| − 1)

(2)

where |E| is the cardinality of the set of edges and |N|

is the cardinality of the set of nodes.

Modularity is a measure of the structure of networks

and estimates the strength of division of a network

into communities (groups). It is often used in optimi-

sation methods for detecting community structure in

networks. A high modularity value indicates a net-

work having dense connections between the nodes

within the communities, but sparse connections be-

tween nodes in different communities. Modularity

can be calculated as (Newman and Girvan, 2004):

Q =

∑

− a

) (3)

where e

i j

is the portion of edges connecting nodes

from the community i to nodes from the community

j; a

∑

i j

is the portion of edges with at least one

node from the community i. Each edge contributes

once in the count (the contribution must be divided

by half, each halve for e

i j

and the other for e

Giant Component (also named main component) is

the connected component which contains most of the

nodes in the graph.

Giant Coefﬁcient is based on the size of the giant

component G

of a graph G. It is calculated as the

number of nodes N

in the giant component divided

by the total number of nodes N in the entire graph:

GC =

|N|

, where N

⊆ N (4)

Diameter is associated with graph distance. It is

calculated as the maximum value among all shortest

paths between two nodes of the graph (i.e., the longest

distance between any pair of nodes belonging to the

graph).

Average Clustering Coefﬁcient is a measure of the

degree to which nodes in a graph tend to cluster to-

gether (connectivity of neighbours). It is calculated

as the average of the clustering coefﬁcients of all the

nodes in the graph:

C =

|N|

∑

i=1

(5)

where C

is the clustering coefﬁcient of a node n

and

is calculated as the number of existing edges between

the direct neighbours of n

divided by the total number

of possible edges directly connecting all neighbours

of n

2.3 Bibliometric Indices

This section introduces two common bibliometric in-

dices often used to measure the impact, in terms

of popularity, of researchers, scientiﬁc publications,

conferences and journals.

h-index was proposed to measure both the number

of publications and the number of citations per pub-

lication of a scientist. According to Hirsch (2005), a

scientist has index h if h of his/her N

papers have

at least h citations each, and the other (N

− h) pa-

pers have no more than h citations each. This index

is also applied to estimate the productivity and impact

of conferences.

i10-index indicates the number of publications of a

scientist having at least ten citations

3 WEBIST WORKFLOW - FROM

RAW TO RDF DATA

3.1 Overview of the Process

This section overviews the process of data acquisi-

tion involving extraction, enrichment, preparation and

consolidation to create the WEBIST Dataset and to

http://googlescholar.blogspot.com.br/2011/

KnowingthepasttoPlanfortheFuture-AnIn-depthAnalysisoftheFirst10EditionsoftheWEBISTConference

433

use it by the WEBIST Analytics. Figure 1 depicts the

whole process.

WEBIST'

Dataset'

…

RDF'

dump'

(1)'

(2)'

(3)'

(5)'

(4)'

Figure 1: WEBIST workﬂow.

Initially, we created an interlinked open dataset,

named WEBIST Dataset, available in RDF, follow-

ing the Linked Data principles (Berners-Lee, 2006),

about the 10 editions of WEBIST conference. This

dataset was created by aggregating data extracted

from different data sources. The initial core of the

data about WEBIST was extracted from DBLP (Dig-

ital Bibliography & Library Project)

(Step 1). Then,

the data was enriched using data crawled from differ-

ent Web sources such as Google Scholar Citations

(Step 2).

Based on the information loaded in the WEBIST

Dataset (Step 3), the proposed Web application pro-

vides different functionalities as exploratory search

and several analysis over the data presented through

different graphical visualisations (Step 4).

Moreover, through the WEBIST Analytics inter-

face, the RDF dump of the WEBIST Dataset is avail-

able for download (Step 5). WEBIST Dataset creation

and WEBIST Analytics functionalities are detailed in

the next subsections.

3.2 WEBIST Dataset

Data Acquisition. Over the last ten years a huge

amount of data has been generated on the Web in

different formats. This also happened with WEBIST

conferences, where information about the conference,

such as paper acceptance or organisation committee

has been published. Thus, to create a tool to seam-

lessly make sense of the data, we aggregated data ex-

tracted from different data sources, being aware of the

http://www.informatik.uni-trier.de/˜ley/db/

http://scholar.google.com/citations

possible necessity of initially preparing the data for

deduplication (Elmagarmid et al., 2007) techniques.

The initial core of the data about WEBIST was

extracted, in December 2014, from DBLP a digital li-

brary about computer science publications. We were

not able to ﬁnd an updated source of DBLP data in

RDF format (containing all editions of WEBIST con-

ference). Thus, we had to extract the data directly

from the XML version of DBLP available. This XML

data also contained information about the name dis-

ambiguation of the authors (different spellings of the

name representing the same author in XML version

of DBLP). Thus, the authors name disambiguation

(Borges et al., 2011) was facilitated in this initial core.

In summary, we collected information about the pub-

lished papers and authors of WEBIST, reaching a total

of 1,449 papers and 2,867 authors.

Data Enrichment. Data enrichment serves as a

means to extending the initial data from additional

data sources. For this, we developed a focused

crawler to obtain this complementary information.

In this step, information from Google Scholar Cita-

tions

and Google Scholar were used to obtain bib-

liometric indices of WEBIST authors. Speciﬁcally,

the key of authors in Google Scholar Citations and

the authors indices (h-index, i10-index and number of

citations) were extracted from Google Scholar

and

Google Scholar Citations, respectively. The crawl-

ing process used the name of the authors to perform

the searches. Using this strategy, 748 authors pro-

ﬁles were found in Google Scholar Citations, repre-

senting 26.09% of the total WEBIST authors. Other

complementary information about some publications

citations was crawled from Google Scholar. We col-

lected the number of citations for the assumed most

cited papers: the candidates to be most cited papers

were obtained by the topmost ranked WEBIST pa-

pers presented in SHINE (Simple H-INdex Estima-

tor)

, Arnetminer

and Microsoft Academic Search

Additional information about the main research areas

of each edition of WEBIST were extracted from each

conference Web site

Data Transformation. Another crucial step is data

transformation, carried out after data acquisition in-

volving the preparation and enrichment steps, requir-

ing a common format for the data. For this, we fol-

lowed the Linked Data principles (Berners-Lee, 2006)

http://scholar.google.com/citations

http://scholar.google.com

http://shine.icomp.ufam.edu.br

http://arnetminer.org/

http://academic.research.microsoft.com/

2005-2011: http://www.webist.org/WEBIST$year$ ;

2012-2014: http://www.webist.org/?y=$year$

WEBIST2015-11thInternationalConferenceonWebInformationSystemsandTechnologies

434

that encourage data publishers to expose their data

through HTTP mechanism and to use RDF as the

data description language. According to this guide

lines, the publishers should name things using HTTP

URIs and provide appropriate clipping of data in RDF

when users follow the URIs. We used a relational-to-

RDF framework (D2RQ) (Bizer and Seaborne, 2004)

that dinamically transforms relational data into RDF

graphs. It provides an HTML browser for relational

databases as well as a SPARQL interface to query the

database. This framework also provides a mapping

language to deﬁne rules for transforming relational

data and schema into RDF graphs.

Data Publication. The successful completion of

these previous steps ensured that the dataset was

available to others (both in terms of users and/or ap-

plications) that want to use it for a myriad of different

purposes. The RDF dump of the WEBIST dataset is

available for download from the WEBIST Analytics

interface.

3.3 WEBIST Analytics Application

WEBIST Analytics, a Web-based application, was cre-

ated to provide multiple perspectives of the data pro-

duced by WEBIST conferences over the 10 editions.

The proposed application is composed of analytics

tools, graphical visualisations and a simple search en-

gine that assists users in ﬁnding, uncovering and mak-

ing sense of the information available. WEBIST Ana-

lytics application can be accessed at:

http://lab.ccead.puc-rio.br/webist analytics/.

Based on the information loaded in the WEBIST

Dataset, the proposed Web application provides dif-

ferent functionalities as both exploratory search and

several analyses over the data, presented through dif-

ferent graphical visualisations. Free text search is

available over two different WEBIST graphs, the co-

authorships graph (among authors) and a more com-

plete graph composed by co-authorships and author-

ing relations (among authors and publications). It al-

lows users to search and retrieve related information

about WEBIST conferences, including an interactive

visualisation of networks. Other exploratory search is

allowed via tag cloud visualisations. In this case, the

terms in the tag cloud can be selected and the associ-

ated publications retrieved, this in turn assisting users

in ﬁnding papers related to each research topic. De-

tails about the different analyses available and their

results discussions are presented in Section 4.

4 ANALYSIS AND RESULTS

This section presents and discusses the results of the

analysis available in WEBIST Analytics. Note that the

outcomes obtained in this section were computed us-

ing the methods and metrics presented in Section 2.

4.1 General Analysis

An initial analysis of all WEBIST conferences was

conducted with regard to its authors and publications.

In this analysis we gathered 1,449 publications, which

included all full papers, short papers, posters and se-

lected papers. Figure 2 depicts the distribution of the

papers over the conference editions. The number of

accepted papers reaches its peak in 2007, where 270

papers were accepted to a single conference, a ﬁg-

ure almost twice the average number of papers ac-

cepted to other editions. This peak number of pub-

lications may be an indication of the rapid increase in

the popularity of WEBIST, and its reaching a certain

level of maturity over the years, settling on a stable

conference-size and community.

Number of papers

Conference year

Figure 2: Number of papers published per year.

A rough analysis of the community can be carried

out based on the number of authors of a scientiﬁc pub-

lication. The number of authors of a paper gives us a

hint of the average size of the community and research

groups. Across the 10 editions of WEBIST, there have

been contributions from 2,867 authors, which gives an

average of 2.91 authors per publication (with a stan-

dard deviation (σ) of 1.35, the maximum number of

authors being 14 per paper and the minimum 1). Fig-

ure 3 shows the distribution of the average number of

authors per year.

The list of topmost authors of WEBIST may re-

veal not only proliﬁc authors, but possible experts and

supporters for future editions of the conference. The

engagement of researchers in a speciﬁc community

could be initially measured with the number of pa-

pers they have had accepted in the earlier editions of

the conference. The assumption is that if they had

over a speciﬁc number of papers, they might be eli-

KnowingthepasttoPlanfortheFuture-AnIn-depthAnalysisoftheFirst10EditionsoftheWEBISTConference

435

Number of (co)authors

Conference year

Figure 3: Average number of (co)authors per paper over the

conference years.

gible to make part of the program committee. After

10 editions, a total of 29 authors had more than six

papers. The most active researcher had 15 published

papers and the second had 12 papers. Figure 4 shows

the top authors as a tag cloud

. The size of the names

represents how active a research is in the WEBIST

conference.

08/01/15 17:24TagCrowd: make your own tag cloud from any text

Página 1 de 1http://tagcrowd.com/

Here is your PDF download link.

« Go back to edit your cloud

Christophe Cruz (15)

Christophe Nicolle (12)

Anne Boyer (10)

Christos Makris (10)

Frans A Henskens (10)

Luc Martens (9)

Toon De Pessemier (9)

Claudia Bauzer Medeiros (8) Daniel Krause (8)

Fabian Abel (8)

Li Li (8)

Maria Jose Escalona Cuaresma (8)

Monique Janneck (8)

Simon Dooms (8) Thomas Ze!erer (8) Tim A Majchrzak (8)

Wu Chou (8)

Antonia Bertolino (7)

David Paul (7)

Jose Palazzo Moreira de Oliveira (7) Juha Puustjarvi (7) Kazunori Sugahara (7)

Michael Hannaford (7)

Nicola Henze (7) Petri Vuorimaa (7) Ricardo Kawase (7) Shinichi Motomura (7)

Takao Kawamura (7)

Yamine Ait Ameur (7)

Figure 4: Top authors with more than 6 papers.

Figure 5 presents the Lorenz curve

along with an

analysis based on the Gini coefﬁcient and the Robin

Hood Index (see Section 2). The Gini coefﬁcient re-

sulted in 25.99% of inequality, while the Robin Hood

Index was 23.06%. The results show that the Lorenz

Curve is closer to the equality than to the inequality

line. This is an expected result for peer-reviewed con-

ferences, where only high quality papers are accepted

for publication. Although a few authors have 6 or

more papers in WEBIST editions, the Lorenz Curve

and the Robin Hood Index show that no redistribution

is necessary, i.e., there is no bias in accepting papers

from a research group or another, but simply merit.

A high Robin Hood Index would indicate a possible

need for further analysis in some publications.

http://tagcrowd.com

http://www.peterrosenmai.com/lorenz-curve-

graphing-tool-and-gini-coefﬁcient-calculator

% papers

Perfect'Equality'

Lorenz'Curve'

Gini=25.99''Robin'Hood=23.06'

Perfect'Inequality'

100'

80'

60'

40'

20'

20' 40'

60' 80'

100'

% authors

Figure 5: Lorenz curve for the number of papers per author

distribution.

4.2 Co-authorships Network

Social Network Analysis (SNA) techniques were ap-

plied to obtain information about the co-authorships

in the WEBIST conference (see Figure 6). The anal-

ysis was conducted over an undirected graph G (de-

ﬁned in Section 2), where the nodes represent the au-

thors and the edges represent a co-authorship between

researchers. A fraction of the co-authorship network

is shown in Figure 6, where the size of the nodes de-

notes the co-authorship connectivity. The WEBIST

co-authorships network is comprised of 2,867 authors

and 4,235 pairs of authors (edges) having at least one

co-authored paper.

Table 1 shows an analysis of the co-authorship

network using SNA measures. The analysis consid-

ers all WEBIST authors in the last 10 years.

• Average Degree shows that the authors, on the

average, have co-authored papers with 2.9 other

authors.

• Density shows a low proportion of co-authorships

in the network relative to the total number pos-

sible (situation where all authors co-authored at

least one paper with all others), only 0.1%. It rep-

resents a weakly connected network. This shows

an expected result in a conference network, where

there are different groups of authors working in

different papers. The measured modularity and

the number of communities, as explained below,

can reinforced this result.

• Modularity shows a high value representing the

strength of division of the network into mod-

ules (also called groups, clusters or communities).

Thus, WEBIST co-authorships network has co-

authorships between the authors within the com-

munities but none between authors in different

WEBIST2015-11thInternationalConferenceonWebInformationSystemsandTechnologies

436

Figure 6: A fraction of the co-authorships network.

communities.

• Number of Communities detected based on the

modularity, was 803, being exactly the same as

the Number of Connected Components. This

shows that, in the analysed network, there are iso-

lated communities that have not co-authorships in

WEBIST with the authors of the other communi-

ties.

The following analysis takes into account only the

giant component of the WEBIST network:

• Giant Coefﬁcient represents the percentage of

authors in the Giant Component of the WE-

BIST co-authorships network, being approxi-

mately 1.57% (45 authors) of the total number

of authors that published in all WEBIST confer-

ences. These authors have 108 co-authorships

between them (2.55% of the total possible co-

authorships, i.e., if each of these authors co-

authored on at least one paper with all others).

• Diameter represents the longest of all the short-

est paths between two authors in the Giant Com-

ponent, being estimated as 8. This shows that

the farthest authors in the Giant Component have

more than six degrees of separation, based on

co-authorship in WEBIST papers. This reveals

that the Giant Component probably results from

a hierarchical structure, which is natural when

research groups of different institutions are in-

volved. The different research groups (subgroups)

are connected by “hub” authors (probably re-

search group leaders and/or professors) that col-

laborate in different research projects amongst the

subgroups, while some researches (probably stu-

dents) developed more speciﬁc tasks (sometimes

related to only one paper).

• Clustering Coefﬁcient measures the average de-

gree to which authors in the network tend to clus-

ter together, being approximately 93.4%. This

shows that many authors belonging the Giant

Component worked with other authors that also

worked together in at least one paper.

Table 1: Social Networks Analysis from the WEBIST co-

authorships network.

Measure Value

Average Degree 2.954

Density 0.001

Modularity 0.995

Number of Communities 803

Number of Connected Components 803

Giant Coefﬁcient

∗

0.0157

Diameter

∗

Average Clustering Coefﬁcient

∗

0.934

∗

Estimated considering the Giant Component.

4.3 Authors Indices

In this section, we considered different bibliometric

indices to analyse the proﬁles of WEBIST authors.

As stated previously (Section 3), we identiﬁed and ex-

tracted Google Scholar Citations proﬁles for 26.09%

of the WEBIST authors. Thus, the analysis presented

in this section is related only to this portion of authors.

The bibliometric indices from WEBIST authors

were ﬁrstly analysed in terms of the Average and the

Standard Deviation (σ) (see results in Table 2). The

bibliometric indices, obtained from Google Scholar

Citations data, were separated into global indices, es-

timated considering all the years of the citations, and

the same indices estimated, considering only the ci-

tations since 2009. On the average, the authors pre-

sented a considerable total number of citations and

i10-index values greater than their h-index. However,

the Standard Deviation was quite high, showing that

the community, as expected in good conferences, is

formed of both young and senior researchers.

To better understand the proﬁle of the WEBIST

authors, we performed further analyses by splitting

the authors into two groups, named A and B. We as-

signed to Group A those authors who had an overall

h-index greater than the h-index since 2009 and as-

signed to Group B those authors who had a overall

h-index equal to the h-index since 2009. This classiﬁ-

cation assumes that the authors whose overall h-index

KnowingthepasttoPlanfortheFuture-AnIn-depthAnalysisoftheFirst10EditionsoftheWEBISTConference

437

Table 2: Average and standard deviation of number of cita-

tions and bibliometric indices from authors.

Measure Average σ

overall citations 1,634.49 4,087.46

citations since 2009 988.95 2,565.17

overall h-index 14.30 12.17

h-index since 2009 11.54 8.98

overall i10-index 28.16 54.32

i10-index since 2009 19.94 42.03

consisted solely of citations made after 2009 were re-

searchers who had started their careers more recently

than those whose overall h-index included citations

from before 2009.

Table 3 presents the results using this classiﬁca-

tion. This table shows, for each conference year, the

percentage of authors and the respective average of

the h-index per class. The results evidence that, in all

conference editions, the number of authors in Group

A is greater than those in Group B. Also, the results

show that, in all conference editions, the average h-

index of authors in Group A is greater. Note that the

average of h-index is 18.35 for authors in Group A

considering all editions of WEBIST conference.

Table 3: Percentage and Average of h-index of scholars in

groups A and B.

year

Percentage Average of h-index

group A group B group A group B

2005 84.62% 15.38% 19.77 9.50

2006 78.65% 21.35% 18.03 8.84

2007 80.59% 19.41% 18.58 7.24

2008 63.73% 36.27% 18.38 6.95

2009 67.86% 32.14% 20.39 8.15

2010 67.03% 32.97% 18.64 7.00

2011 67.06% 32.94% 20.16 5.54

2012 57.02% 42.98% 15.65 5.39

2013 53.03% 46.97% 20.74 6.52

2014 55.56% 44.44% 19.72 4.90

All 65.64% 34.36% 18.35 6.58

4.4 Topics and Conference Areas

In this section, we analyse the topics of the papers

published over the 10 years of WEBIST conference

and their relation to the predeﬁned main conference

areas. Firstly, Figure 7 presents, in alphabetical or-

der, the main conference areas over the different con-

ference editions. Some areas appear in all confer-

ence editions, such as Society, E-Business and E-

Government and Web Interfaces and Applications.

The third most frequent area is Internet Technology,

which appeared from the second edition to the last

one, probably as an expansion of Internet Comput-

ing (which appears only in the ﬁrst conference edi-

tion). Web Intelligence and Mobile Information Sys-

tems appear more recently, in 2009 and 2012, respec-

tively. E-Learning appears only in the ﬁrst four edi-

tions of WEBIST conference. This phenomenon can

be explained by the fact that the WEBIST confer-

ence, from 2009 to 2014, was held in conjunction with

CSEDU (The International Conference on Computer

Supported Education), a conference focused in inno-

vative technology-based learning strategies and insti-

tutional policies on computer supported education (e-

learning). Web Security appears only in speciﬁc edi-

tions (2005 and 2011).

Another analysis was performed over the topics

covered by the papers published in WEBIST confer-

ences. Figure 8 shows a tag cloud generated from the

terms presented in the titles of the papers. This tag

cloud represents the terms followed by their total fre-

quencies in parentheses. Moreover, the term size in

the graphic is proportional to their frequency. Terms

as web, systems, services, applications, model and

information are the most frequent. These terms are

aligned with the research focuses of WEBIST con-

ference that are technological advances and business

applications of web-based information systems.

12/12/14 14:42WEBIST Analytics

Página 1 de 1http://localhost/dblp_webist/site/index.html#

Ten Years of WEBIST Conference