=Paper=
{{Paper
|id=Vol-1279/iesd14_5
|storemode=property
|title=A Visual Exploration Workflow as Enabler for the Exploitation of Linked Open Data
|pdfUrl=https://ceur-ws.org/Vol-1279/iesd14_5.pdf
|volume=Vol-1279
|dblpUrl=https://dblp.org/rec/conf/semweb/VochtDBCVMMW14
}}
==A Visual Exploration Workflow as Enabler for the Exploitation of Linked Open Data==
https://ceur-ws.org/Vol-1279/iesd14_5.pdf
A Visual Exploration Workflow as Enabler
for the Exploitation of Linked Open Data
Laurens De Vocht1 , Anastasia Dimou1 , Jonas Breuer2 , Mathias Van Compernolle3 ,
Ruben Verborgh1 , Erik Mannens1 , Peter Mechant3 , and Rik Van de Walle1
1
iMinds - Ghent University, Multimedia Lab,
Ghent, Belgium
{firstname.lastname}@ugent.be
2
iMinds - Vrije Universiteit Brussel, SMIT,
Brussels, Belgium
{firstname.lastname}@iminds.be
3
iMinds - Ghent University, MICT,
Ghent, Belgium
{firstname.lastname}@ugent.be
Abstract. Semantically annotating and interlinking Open Data results in Linked
Open Data which concisely and unambiguously describes a knowledge domain.
However, the uptake of the Linked Data depends on its usefulness to non-Semantic
Web experts. Failing to support data consumers to understand the added-value of
Linked Data and possible exploitation opportunities could inhibit its diffusion.
In this paper, we propose an interactive visual workflow for discovering and ex-
ploring Linked Open Data. We implemented the workflow considering academic
library metadata and carried out a qualitative evaluation. We assessed the work-
flow’s potential impact on data consumers which bridges the offer: published
Linked Open Data; and the demand as requests for: (i) higher quality data; and
(ii) more applications that re-use data. More than 70% of the 34 test users agreed
that the workflow fulfills its goal: it facilitates non-Semantic Web experts to un-
derstand the potential of Linked Open Data.
1 Introduction
A lack of in-depth understanding of the underlying semantic technology limits most
Web users to interpret and query Linked Open Data (LOD). Data consumers do not
straightforwardly perceive that the enriched and machine-understandable representation
of the information, the offered cognitive prosthetics, could help in solving complex
problems [21]. As Dadzie and Rowe [3] mentioned, the uptake of the Linked Data by
a mainstream audience depends on its utility to non-Semantic Web experts; failing to
support them in understanding the LOD potential could inhibit its swift adoption and
positive impact on society. Therefore, non-Semantic Web experts need the means to
discover and to explore LOD more intelligently.
We expect that data consumers will form their demands for more LOD of better
quality and for targeted applications benefiting from LOD, if they are aware of the po-
tential value. Until now, observation of data was sufficient as data is often fragmented
by default in different sources, even if sets of these data combined describe a certain
knowledge domain. In the case of LOD though, their linked nature should be revealed.
Exploring LOD implies that data consumers become aware of the available content and
the links between them to really appreciate the “potential” they have at their disposal.
Furthermore, Janssen et al. [12] argued that information may appear to be irrelevant
when viewed in isolation, but when linked and analyzed collectively can result in new
insights. Goedertier et al.4 confirmed that there is a positive trend to open up public data.
The Web accords to consumers the infrastructure and means for interaction with the
data. However, the study also confirms that the re-use of provided LOGD is still limited.
Finally, Genie Stowers5 outlines the following reasons to make use of data visualization:
(i) for storytelling and making information more visible; (ii) for simplifying, clarifying
and analyzing data and making plans, within a more data-driven policy-cycle.
As LOD are typically represented as graphs, exploring its visualization is a way to
allow Web users to implicitly compose queries, identify links between resources and
intuitively discover new relevant pieces of information [3]. This addresses the need to
reveal links, made available in the LOD, which were not explicitly defined in the original
data.
From the perspective of the initiating government, there are clear benefits of such an
initiative; in the particular case of research information data, public institutions make
investments by funding research and they need to assess the return on such investment.
Further employing the example of academic library metadata illustrates that researchers
and organizations undertaking research (universities etc.) benefit foremost themselves
from having qualitative data; this data they can then provide to a common LOD envi-
ronment provided by a public sector body, where the public and (especially) the gov-
ernment can access and explore the data, thereby generating value.
In this paper, we present an interactive visual graph-based exploration workflow
over published LOD and we assess how this workflow supports data consumers to dis-
cover and explore the data: how useful it is for data consumers to discover the data and
explore the links between them. Our uppermost goal is to assess the users’ satisfaction
and thus the potential of such a visual workflow to act as enabler for LOD exploita-
tion. The remainder of the paper is structured as follows: in section 2, we describe the
current state of the art regarding LOD visualizations, information exploration and what
LOD means in the societal context. In Section 3, we describe the exploration workflow,
in Section 4 we describe the system architecture that implements the workflow and in
Section 5 we present the evaluation. In Section 6 we summarize our conclusions and
we suggest future work.
2 State of the Art
As our proposed solution is multi-component and multi-dimensional, we outline the
state of the art of the different relevant fields below. Because our workflow relies on
visualizations, we review the state of the art regarding LOD visualizations and academic
metadata visualizations, as this is the domain of our use case. We outline the information
exploration techniques applied to our workflow and summarize some insights already
formed relevant to the LOD socio-economic potential.
2.1 Information Exploration
The resolution of vague or complex information problems requires exploratory be-
haviours, for instance: multiple publishers providing resources. We examine how ex-
ploratory search and exploratory data analysis applied to LOD can fruitfully solve such
problems. Exploratory Data Analysis (EDA) [18] allows the data itself to reveal its un-
derlying model and its relationships without requiring any formal statistical modeling
and inference (non hypothesis-driven). Graphical EDA employs a variety of techniques
4
https://joinup.ec.europa.eu/community/semic/document/
study-business-models-linked-open-government-data-bm4logd
5
http://www.businessofgovernment.org/report/use-data-visualization-government
to present the underlying data, maximizes the insight into a dataset and uncovers the
underlying data patterns, allowing the users to discover the resources in the dataset.
Exploratory search, on the other hand, describes either the problem context that
motivates the search or the process by which the search is conducted [14]. The users
start from a vague but still goal-oriented defined information need and are able to refine
their need upon the availability of new information to address it, with a mix of keyword
look-up, expanding or rearranging the search context, filtering and analysis. During
exploratory searches and analysis, it is likely that the problem context becomes better
understood, allowing the searchers to make more informed decisions about interaction
or information use [21].
2.2 Linked Open Data Visualizations.
The “Linked Data Visualization Model” (LDVM) allows to connect different datasets,
data extractions and visualizations in a dynamic way [1] rather than focusing on a single
platform. Furthermore, the use of coordinated views to facilitate integration of visual-
izations [16], is a complex process to decide on visualization methods to successfully
aid seeking and discovery of information [15]. Coordinated views are provided to align
multiple perspectives on a dataset. In “VisLink”, “coordinated views” link existing vi-
sualizations [2], while we use a coordinated view to align the narrowing and broadening
views in the workflow. Tvarozek et al. [19] empower users with access to semantic in-
formation spaces via an exploratory browser. At the end of an exploration session users
need to start a new search, a history view allows users to step back.
Dadzie and Rowe [3] concluded in their study on Linked Data interfaces and vi-
sualizations that only a limited number was available at the time of writing and each
of them focuses on a separate aspect to support users. They highlighted the issue, an
important motivation for our workflow as well, that without good quality LOD there is
little motivation to build such interfaces for end-users while these interfaces are needed
to locate and retrieve LOD in the first place. To address this issue, we aim to have visu-
alizations for diverse domain specific solutions. However, these visualizations remain
generic because they are built for any case specific queries against published LOD.
2.3 Academic Data Visualizations
In the past there were attempts to visualize research networks but most of them did
not rely on LOD. The below mentioned works based on research LOD consider visu-
alizations as a supportive mean to the presented information. None of them focus on
the visualizations and, therefore they do not take into consideration a methodological
approach for the data exploration as such.
The Semantic Web Journal published its own Drupal-based journal management
system [11] focusing on providing a novel user interface. Among others, they pro-
vide graph-based research networks that visualize the emerging research networks as
researchers author papers together or they review the different submissions. “Arnet-
Miner” [17], on the other hand, distinguishes between the networks (star graph of co-
authors) and the communities of researchers (simple graphs). Finally, “TalkExplorer” [20]
takes into consideration bookmarks and tags for the visualizations of the research groups
and puts the focus on providing recommendations rather than exploring the underlying
dataset. In our workflow we make abstraction of the query creation process and use of
pre-defined query templates to facilitate the creation of the visualizations.
3 Exploration Workflow
Data consumers need Information Analysis and Synthesis (IAS) to acquire knowledge
and form inquiries for a domain unknown to them, consisting of complex and disparate
data. Corresponding Information Exploration techniques can facilitate the exploration
of the available datasets as mentioned in Section 2.1. To this end, we considered for
our workflow exploratory data analysis to assist data consumers to analyze the avail-
able dataset and exploratory search to facilitate them synthesizing complex queries. We
applied each technique to visualisations and implemented them.
3.1 Defining the Workflow
The visual workflow is streamlined through a coordinated view of the two different
parts centralizing the link focused on a specific resource that binds them. This way,
our workflow enables users to discover, search and analyze LOD. Figure 1 shows how
users start with an overview of the dataset (Figure 1a) through which the users “dive”
in more narrow perspectives (Figure 1b) by selecting a group to find out details and see
the internal relations of the subdivisions (Figure 1b). A coordinated view (Figure 1c)
of selected resources leads them through a broadened view (Figure 1d) by exploring
relations of these resources.
Narrowing Broadening
a. Overview of Groups b. Group Details c. Coordinated View d. Broadened View
Fig. 1: Narrowing views (a, b) allow data consumers to analyse the dataset. The coordinated view (c) allows
perspective switching in the workflow. Broadening views (d) allow data consumers to explore the interlinked
information beyond the dataset’s boundaries.
a,b. Narrowing Views. The narrowing views aim to familiarize the data consumer with
a certain dataset, as they are not aware of its context. Since there is no explicit assump-
tion regarding its content, as EDA ordains, the dataset itself reveals its underlying model
and the relationships between its resources. Given the “unlimited” extent of a dataset,
the initial view is focused on this certain dataset and its broader concepts are demon-
strated. Deeper exploration, following the links until reaching the resources that can not
be further decomposed, helps to better comprehend the available data.
c. Coordinated View. As the users, supported by the visualizations, narrow down to
more detailed resources (a certain resource or the links between two resources), they
reach the resources that cannot be further decomposed and thus act as the coordinated
view. Starting from this view, data consumers, being aware of the underlying dataset,
start exploring the dataset. The coordinated view form a “bridge” between the narrow
view and the broader view, which exploits existing links amongst resources across dif-
ferent datasets.
d. Broadening Views. In the case of broadening views, data consumers find novel
relations between existing and known resources interacting with the visualizations of
the LOD. The possible views are not limited to the data of the narrowing view but the
links to other datasets are also revealed and visualized if considered relevant. It is a
new way to search and explore the information. This way, users get an overview by
using an approach that visualizes interactively search process in an aligned linked data
knowledge base of related resources.
3.2 Applying Information Exploration Techniques to the Workflow
The narrowing view is achieved based on exploratory analysis techniques [18] applied
to the dataset. Without any formal modeling or assumption about the underlying dataset,
the main concepts and their relationships are gradually revealed. Subsequent views
narrow the broader concepts and reveal more details about the relations among the
concepts. The broadening view is achieved using exploratory search techniques [14]
over individual instances of the LOD. Users iterate over individual concepts, their direct
neighbors and their relationships. Iteratively expanding and focusing the visualization
leads to more insight into selected concepts in the datasets.
4 Implementation
We developed a demo where we implemented each step of the workflow: the narrowing,
coordinated and broadening views; as described in Section 4.1. The visualizations used
complementary datasets (see Section 4.2) whose intersection serves for the coordinated
view. In Figure 2 we apply the workflow to a use case of academic library metadata
which is generalizable to other types of LOD and domains.
Published discover
Data Consumers switch
Open Data explore Narrowing perspective
Overview
views
analyse
Contract
Data Owners/Publisher Demand Coordinated
View
interpret
Expanded explore Broadening
View Views
LOD
reuse Expand
Published LOD
Linked Open Data Exploration Workflow
Fig. 2: The visualizations are part of the LOD exploration workflow which can be applied for the exploration
of different datasets, in this case academic library metadata.
4.1 Visualization Tools
We implemented the workflow using two tools we developed for visualizing LOD.
This results in the graph based exploration interface supporting narrowing views by
LOD/VizSuite and broadening views by ResXplorer over a coordinated view.
LOD/VizSuite The goal of the LOD Visualization Suite (LOD/VizSuite)6 is to create
an easily customizable visualization framework on top of LOD. LOD/ VizSuite aims to
be data and schema agnostic, therefore it can be easily transferable to visualize different
datasets. Its functionality is based on SPARQL queries which are published as SPARQL
templates. Parameters can be passed to the SPARQL template at request time, which
replace placeholders to construct a valid SPARQL query. The SPARQL templates are
published at a DataTank7 instance, a RESTful (Linked) Open Data management system
6
http://ewi.mmlab.be/academic
7
http://thedatatank.com
which publishes data on the Web. LOD/VizSuite exposes research and collaboration
networks, communities of practice in a certain discipline and timelines to monitor a
discipline’s evolution over time [7].
ResXplorer The goal of ResXplorer8 is that researchers can find novel relations be-
tween existing known items such as authors, publications or conferences. Users in-
teract with a visualization of resources [5] using an interface combining an optimized
pathfinding algorithm (the Everything is Connected Engine) [4] with Web 2.0 technolo-
gies (such as JQuery and Django). The result is a semantic search tool providing both a
technical demonstration and a visualization that is applicable to many other applications
beyond academic library metadata.
4.2 Datasets
LOD/VizSuite provides visualizations based on the LOD provided by the “Research In-
formation Linked Open Data” (RILOD)9 data-set. RILOD is the result of the integration
of heterogeneous sources related to research in Flanders, ending up in a rich and diverse
dataset. The datasets contains resources of researchers from the region of Flanders, their
publications and projects, which are associated with the corresponding research groups
and institutes, and classified under the IWETO Discipline classification10 . ResXplorer
makes use of the “Digital Bibliography and Library Project”11 (DBLP), an on-line ref-
erence for bibliographic information on major computer science publications [13]. The
binding between RILOD and DBLP is their content’s intersection: the same researchers
and publications appear in both datasets.
4.3 Embedding Visualizations in the Workflow
In this section, we explain how we embedded the visualizations to implement the three
types of views of our exploration workflow. Figure 3 shows the visualizations embedded
in the workflow.
a c
d
Narrowing Coordinated View
b
Broadening
Fig. 3: Corresponding with steps a, b, c, d in Figure 1, users narrow down from disciplines (a) to research
groups and further to the individual researchers in this group (b). To find out relations between researchers
they select two researchers and, using the coordinated view (c), shift to the broadening view and expand to
resources beyond their research community (d).
8
http://www.ResXplorer.org
9
http://ewilod.be/ewilod/html/sparql-test.html
10
http://ewilod.be/ewilod/lod/0.1/ontology/taxonomies/iwetoDisciplineCodes
11
http://dblp.l3s.de/
The broadest concepts, which cover all the dataset, are chosen for the overview
view. The overview view serves the users to discover the main concepts of the dataset,
the strength of the relations between them (Figure 3a) and the diversification of the to-
tal number of the instances that constitute the broader concept. From the overview, the
users discover the narrower entities (Figure 3b). Broader views are achieved by aggre-
gating narrower entities using SPARQL queries that select and group them. In our use
case, visualizations that provide an overview view of a research discipline are achieved
by aggregating the researchers under their research groups and providing their collabo-
ration links considering their co-publications. The research groups are demonstrated as
graph nodes that diversify in size depending on the total number of projects they have,
while the strength of the links depends on their researchers’ co-publications.
As a research group is the aggregation of individual researchers, an end user can fur-
ther narrow down and view the researchers and their publications (decomposed views).
This is the narrowest view which acts as the coordinated view (Figure 3c). In our
use case such a resource can be an individual researcher or the links between two
researchers whose extensive collaboration network is demonstrated. As the end users
view the network formed around a researcher or the exhaustive list of paths between
two researchers, they can be transposed to the corresponding view of the broadening
part of the workflow.
While exploring the broadening view, data consumers are not limited to the data
of the dataset but their exploration is enhanced with links to other datasets of the LOD
cloud that might be relevant to their exploration (e.g., DBLP in our use case). We show
that visualizing resources, such as conferences, publications and proceedings, expose
affinities between users and those resources (Figure 3d). We characterize each affinity,
between users and resources, by the amount of shared interests and other commonali-
ties.
5 Evaluation
We evaluated the workflow at “iMinds The Conference”12 , a yearly gathering for re-
searchers active in several aspects of digital innovation. The conference’s audience was
identical to the use cases’ target groups, namely researchers and R&D policy makers.
Data was gathered using a multi-method approach. We used the think-aloud protocol
during the experiments to collect feedback from the participants and we recorded the
screen actions of participants using QTrace13 .
Test users received no information about the tool in advance. They were asked to
execute some assignments and to fill in a questionnaire afterwards. During the user test,
users were asked to think aloud. The observer recorded comments and took notes. In
the additional framework, we asked people coming over at the booth, but who did not
participate in the user test, to fill in the same questionnaire after receiving an explana-
tion about the workflow and the tools. The survey included (i) perceived aims of each
step in the workflow; (ii) a series of statements on a five-point Likert scale measuring
the usefulness, learnability, complexity, explorability, transparency as perceived by the
respondents; and (iii) open auestions on insight in quality of the used data and sugges-
tions. Each test took about 30 to 45 minutes. The questionnaire took an additional 5 to
10 minutes to be completed. Apart from the 17 users who participated in the evaluation,
17 additional users participated in the evaluation by filling out the same questionnaire
after receiving information about the visualizations, giving us richer data for the survey
analysis. In total, 34 users participated in the survey.
12
http://conference2013.iminds.be/
13
http://www.qasymphony.com/qtrace.html
We kept the audience of the assessment broad by conducting also semi-structured
interviews with various stakeholders. All of them are likely to be affected by the impact
and value of accessible and explorable LOD. The use case was situated in the context
of research information. Thus interviewees are active for the Flemish government de-
partment of Research and Innovation, the Department of Research policy from Ghent
University, and, from a commercial point of view, in the domain of Business Develop-
ment & Academic Relations.
In our previous work [6], we evaluated how appealing our visual workflow and the
visualization tools are to the end users by assessing the productivity and precision of
the narrowing part and the complexity and searchability of the broadening view. Our
evaluation showed that the implemented visualisations were capable of assisting the
end-users to interpret the visualisations, thus adequate for the scope they were designed.
In this work, we evaluate the workflow in regard to the scope it was implemented for.
5.1 Visualizations in the Workflow
We observed how the test users executed the assignments and we asked them to think
aloud. The test users were asked (i) to start from their preferred research discipline
(overview view), (ii) to go on towards their preferred research group and researchers
and, explore their collaborations (explore the links of the narrower views) and (iii) to
explore the links of one of the researchers that they concluded at while they navigated
to broader views (broadening view).
Observations. The think-aloud analysis gave us information regarding the perception
of the visualisations by the participants, during the executions of the assignments. The
observers asked test users to search for their own name; or when it did not appear in
the dataset; to searh for a research subject they are familiar with (for example related
to their own research). Users had to indicate differences in the size of nodes and they
received an explanation on the meaning of the node sizes. Afterterwards they had to
indicate if the visualization appeared to them as a good representation of the real-world
situation. Via this direct feedback, we concluded that test users are able to reason via
the tools: for example or by appointing missing research groups in the visualizations,
by putting the size of the nodes into discussion in LOD/VizSuite.
The observations give us further insights regarding how the users expect the ex-
ploration to happen: clicking on the broad views, e.g., clicking research groups within
disciplines, they expect they get an intermediate overview and each step forward in the
workflow can give additional input to explore. Once they realize the fact of the narrowed
view and the effect of the coordinated view, data consumers are able to fully compre-
hend the workflow and start with simple reasoning that supports the intention of the
exploratory search tool. We observed that, once test users comprehend the exploration
workflow, they better accept the visualized data and become more able to form their
exploration path. This affects their exploration behavior: they use the different features
to get further insights (search queries, top affinity suggestions, or expanding via node
clicking). Although complexity rises within the visualizations during the explorations
(earlier explored data stay visualized and taken into account), test users understand the
potential of visualizing academic data and can name how they are related to another
researcher via conferences or publications from intermediate researchers. Test users
declared that further input could bring in additional points of interests.
Survey. To evaluate the exploration of the LOD, we asked the test users and twenty
extra respondents their impression of the views using a questionnaire. To determine
the impact and quality of the workflow considering their use for data consumers, we
analyzed how the users explored and perceived the visualizations in the corresponding
views. We especially measured the perceived usefulness and learnability and how the
participants estimate the potential of the visualizations.
100%
90%
80%
70%
60%
50%
40%
Usefulness Learnability Innovativeness Transparency
Explorability Complexity Potential LOD Quality
Narrowing Broadening Overall
Fig. 4: User perceived goals of the views (left) indicate that the narrowing view is perceived to be more
suitable for analysis while the broadening view scores slightly better for exploration. User satisfaction for
the workflow (right) shows that overall the views don’t seem to expose innovation. In terms of uselness and
complexity the users are very satisfied with the narrowing, they need some time to learn how the broadening
works.
Visual workflow’s goals. To understand how the users perceive this visual workflow
and its goals, we asked them to score possible purposes of use. As displayed in Fig-
ure 4, the respondents indeed perceived both the narrowing and broadening views as
adequate tools to explore, discover and search. The broadening view is considered by
the respondents as being a tool for exploration in the first place and discovery in the
second place. The narrowing view is considered as a means to explore and to search.
Usefulness and Explorability. Test users agree that the visualizations is useful in terms
of what it exposes 22 out of 34 (65%) agree for the broadening views and 28 out of 34
(82%) for the narrowing views. 28 out of the 34 respondents (74%) agree or strongly
agree that they understand the displayed relations of the broadening view are presented
as an optimized selection of all results. Although, respondents stay rather undecided
when it comes to the limitations: 16 out of 34 respondents (47%) agree or strongly
agree that it is useful that the number of visualized resources and relations are fixed to
the 7 most relevant resources, whereas 11 out of 34 (32%) disagree or strongly disagree
on this. Finally, the respondents strongly agree that both the broadening view and the
narrowing views support them gaining insights into the published data, but they were
less confident in the case of the narrowing view.
Complexity and Learnability. The majority of the respondents agreed that they can
interpret the complexity of the visualizations both for the narrowing views, 27 out of 34
(79%), and the broadening views, 23 out of 34 (68%). Most of the respondents agree
with the statement that after a learning period (we did not measure this) they get familiar
with the visualizations in the narrowing view and they can get benefit out of it 30 out
of 34 (88%) and even more of them agree for the broadening exploration 31 out of 34
(91%).
Workflow Potential and Innovativeness. The respondents were asked how they perceive
the potential of the workflow for LOD exploration. 23 of 34 test users (68%) state that
the visual exploration workflow clearly helps them to understand the potential of Open
Data. 16 out of 34 (47%) respondents agreed or strongly agreed that the visualizations
help to get insight into innovation. However, 11 out of 34 (32%) respondents remained
undecided.
Linked Open Data Quality and Transparency. 23 out 34 (68%) respondents agreed
that the workflow can serve as an encouraging factor to provide more open data and
of better quality. The workflow imposes the data consumers to behave in a different
and unfamiliar way, but at the end, they get more familiar with the underlying LOD
which can bring in new unexplored information. This can explain the mixed response
on the visualizations as an encouraging factor to gain data of better quality. 22 out of 34
respondents (65%) agreed that the government becomes more transparent if the public
is able to explore the pubished LOD with such visualizations. Once they get familiar
with the narrowing and coordinated view, our respondents state it helps them in the
discovery and exploration of the data. The broadening part helps them to gain new
insights in the dataset and explore links with data published to the broader LOD cloud.
5.2 Workflow applied to Linked Open Data and Academic Metadata
The observations lead us to the conclusion that wider ecosystems need to be taken into
account to analyze and facilitate value creation around Linked Open Data. The term
ecosystem is applicable here, because it emphasizes interconnections and interdepen-
dencies between the different actors and roles.
As illustrated in Figure 2, consumers can discover published LOD and its mean-
ing through the exploration workflow. As this adds value to their own activities, they
are likely to increase the demand, and address data publishers to improve respective
infrastructure. To assess the potential of LOD, the process of linking open data in a
broader multi-stakeholder context needs also to be taken into account. For this assess-
ment, Linked Open Government Data (LOGD) [8] and the significance of visualization
and exploration for governments and policy-making plays a decisive role. Geiger and
Von Lucke [9] evaluated opportunities for government as a positive paradigm shift, for
instance, new ways of political legitimation or innovation and modernization of admin-
istration in an increasingly open world. Higher scalability and quality in the production
of LOGD entails that governments need to be seen both as data supplier and as data
consumer [10].
Here, the much discussed impact of LOD as a driver for economic growth and inno-
vation is most applicable14 . Thus, each actor must have different incentives for partic-
ipating in LOD initiatives: Governments might aim for efficiency; companies for gen-
erating revenue; universities for better research etc. However, the interaction and inter-
dependence of all involved parties can finally impact on more data, with better quality,
and benefits for all affected parties. This potential is not always revealed because of var-
ious challenges which form roadblocks to the realization of that potential. The problem
14
http://www.slideshare.net/DeloitteAnalytics/open-data-13509015
is that, even when data is published adequately, there will neither be value nor impact
arising from the development, as long as citizens, businesses, public servants etc. do
not use it. This means that the challenges of achieving the potential, which LOD clearly
holds, go beyond its mere publication.
An interview, undertaken with the responsible functionary of the initiating govern-
ment department, clarified their motivations in relation to Open Government Data. To
achieve the desired objectives, a Linked Open Data environment for research informa-
tion was created, based on the spirit that data, which is publicly financed, should be
properly accessible to the public. However, its capacities were still limited and infor-
mation was not as complete as it could be. Data was published in an infrastructure to
facilitate exploration and ultimately to add value, but still accessibility remains cur-
tailed, in particular for non-expert users. Therefore, the development of visual explo-
ration workflows on top of the LOD was encouraged, providing accessibility to complex
data, and understanding of the meaningful correlations emerging from the links.
6 Conclusions and Future Work
Interactive user interfaces based on enhanced visualizations allow users to have a unique,
multifaceted experience when combined with techniques for information exploration.
Such interface is demonstrated in our graph-based workflow for the exploration of a
dataset. According to our respondents, they, as data consumers, become acquainted with
the underlying dataset and the workflow can bring in new unexplored information and
knowledge as soon they familiarized themselves with the workflow. The respondents,
all active in the occupation of academics (namely domain experts) remained satisfied,
and thus we verify that the workflow is applicable to academic library metadata and
generalizable to the exploration of any other LOD. Overall, offering such visualizations
on top of LOD: turns the potential of expressing a demand for LOD exploitation more
likely; and it increases the demand for data of better quality, both in terms of context
and semantics.
Considering the results of the demo’s evaluation, we are planning an integrated
solution that aligns with the users’ workflow as they explore the information. Addi-
tionally, we would like to apply the proposed exploration workflow to other datasets
and evaluate its applicability to other domains such as cases of exploring other type of
actors and their actions’ output but also even broader. Considering the test users’ feed-
back, we will improve several aspects of the implementation. A task based evaluation,
could show that the proposed workflow indeed facilitates tasks that can not be carried
out otherwise, therefore reinforcing the “potential of linked data”. We plan to support
multigraphs introducing multiple relations between displayed resources and subgraphs
and direct expansion of the aggregated resources in order to achieve a more consistent
exploration over the available data. We aim to keep the exploration workflow as data
and schema agnostic so it can be easily transferable to other datasets. Finally, we plan
a thorough evaluation of the integrated workflow applied to diverse datasets.
7 Acknowledgement
The research activities described in this paper were funded by Ghent University, the
Flemish Department of Economy, Science and Innovation (EWI), the Institute for the
Promotion of Innovation by Science and Technology in Flanders (IWT), the Fund for
Scientific Research-Flanders (FWO-Flanders), and the European Union.
References
1. J. M. Brunetti, S. Auer, and R. Garcı́a. The linked data visualization model. In International
Semantic Web Conference (Posters & Demos), 2012.
2. C. Collins and S. Carpendale. VisLink: Revealing relationships amongst visualizations.
IEEE Trans. on Visualization and Computer Graphics (Proc. of the IEEE Conf. on Infor-
mation Visualization (InfoVis)), 13(6), 2007.
3. A.-S. Dadzie and M. Rowe. Approaches to visualising Linked Data: A survey. Semantic
Web, 2(2), 2011.
4. L. De Vocht, S. Coppens, R. Verborgh, M. Vander Sande, E. Mannens, and R. Van de Walle.
Discovering meaningful connections between resources in the Web of Data. In Proceedings
of the 6th Workshop on Linked Data on the Web. CEUR-WS, 2013.
5. L. De Vocht, E. Mannens, R. Van de Walle, S. Softic, and M. Ebner. A search interface
for researchers to explore affinities in a Linked Data knowledge base. In Proceedings of the
12th International Semantic Web Conference Posters and Demonstrations Track. CEUR-WS,
2013.
6. A. Dimou, L. De Vocht, M. Van Compernolle, E. Mannens, P. Mechant, and R. Van de
Walle. A Visual Workflow to Explore the Web of Data for Scholars. In Proceedings of the
1st Workshop on Big Scholar: Towards the Web of Scholars, 2014.
7. A. Dimou, L. De Vocht, G. Van Grootel, L. Van Campe, J. Latour, E. Mannens, P. Mechant,
and R. Van de Walle. Visualizing the information of a Linked Open Data enabled Research
Information System. In Proceedings of the Current Research Information Systems Confer-
ence, 2014.
8. L. Ding, V. Peristeras, and M. Hausenblas. Linked open government data. Intelligent Sys-
tems, IEEE, 27(3):11–15, 2012.
9. C. P. Geiger and J. von Lucke. Open Government and Linked Open Government Data.
Journal of eDemocracy and Open Government, 4(2), 2012.
10. A. Graves. Creation of visualizations based on Linked Data. In Proceedings of the 3rd
International Conference on Web Intelligence, Mining and Semantics. ACM, 2013.
11. Y. Hu, K. Janowicz, G. McKenzie, K. Sengupta, and P. Hitzler. A linked-data-driven and
semantically-enabled journal portal for scientometrics. In The Semantic Web - ISWC 2013,
volume 8219 of Lecture Notes in Computer Science. Springer Berlin Heidelberg, 2013.
12. M. Janssen, Y. Charalabidis, and A. Zuiderwijk. Benefits, adoption barriers and myths of
open data and open government. Information Systems Management, 29(4), 2012.
13. M. Ley. The DBLP computer science bibliography: Evolution, research issues, perspectives.
In String Processing and Information Retrieval. Springer, 2002.
14. G. Marchionini. Exploratory search: From finding to understanding. Communications of the
ACM, 49(4), 2006.
15. J. C. Roberts. State of the art: Coordinated & multiple views in exploratory visualization. In
Coordinated and Multiple Views in Exploratory Visualization, 2007. CMV’07. Fifth Interna-
tional Conference on. IEEE, 2007.
16. M. Scherr. Multiple and coordinated views in information visualization. Trends in Informa-
tion Visualization, 2008.
17. J. Tang, J. Zhang, L. Yao, J. Li, L. Zhang, and Z. Su. Arnetminer: Extraction and mining of
academic social networks. KDD ’08. ACM, 2008.
18. J. Tukey. Exploratory Data Analysis. Addison-Wesley series in behavioral sciences.
Addison-Wesley Publishing Company, 1977.
19. M. Tvarožek et al. Exploratory search in the adaptive social semantic web. Information
Sciences and Technologies Bulletin of the ACM Slovakia, 3(1), 2011.
20. K. Verbert, D. Parra, P. Brusilovsky, and E. Duval. Visualizing recommendations to sup-
port exploration, transparency and controllability. In Proceedings of the 2013 International
Conference on Intelligent User Interfaces, IUI ’13, 2013.
21. R. W. White and R. A. Roth. Exploratory search: Beyond the query-response paradigm.
Synthesis Lectures on Information Concepts, Retrieval, and Services, 1(1), 2009.