Content uploaded by Milos Jovanovik
Author content
All content in this area was uploaded by Milos Jovanovik on May 22, 2021
Content may be subject to copyright.
Semantic Web and Data Science Integration
Using Computational Books
Dimitar Mileski, Milos Jovanovik and Dimitar Trajanov
Faculty of Computer Science and Engineering
Ss. Cyril and Methodius University in Skopje, North Macedonia
E-mail: dimitar.mileski@ieee.org, {milos.jovanovik, dimitar.trajanov}@finki.ukim.mk
Abstract—This paper presents the architecture for the devel-
opment of web applications for exploring semantic knowledge
graphs through parameterized interactive visualizations. The
web interface and the interactive parameterized visualizations,
in the form of a computational book, provide a way in which
knowledge graphs can be explored. An important part of using
this approach for building interactive web visualizations is that
we can substitute the knowledge graph entities with other entities
within the existing interactive visualizations, execute commands
in a web-based environment, and get the same visualization for
the new entities. With this architecture, various applications for
interactive visualization of knowledge graphs can be developed,
which can also stimulate the interest to explore the graph and its
entities. We also present a publicly available open source use-case
that is built using the concepts discussed in this paper.
Index Terms—Knowledge Graphs, Linked Data, Semantic
Web, Visualisation, RDF, SPARQL
I. INTRODUCTION
The Semantic Web, along with the concepts of Linked Data
and Knowledge Graphs, are built on mature technologies.
However, the development of applications in this domain
requires a certain learning curve, for people which are not
familiar with the basic concepts. Changing the paradigm and
thinking about data as tables to data as graphs, can deter
some people from this set of technologies. Many advanced
tools developed for knowledge graphs do not have a web
interface, which further deters people who are not developers
or scientists. In this paper, we present a new way in which a
user can explore knowledge graphs and their entities, and can
discover new data that can be used for machine learning and
data science.
The steps that users need to take are as follows.
1) Select the entities and parameters they want to visualize.
2) Select an existing visualization of the computational
book whose data types of parameters will match the
entities and parameters they want to visualize.
3) Execute code with the changed parameters, with an in-
line interactive web environment.
Such web interactive parameterized visualizations can en-
courage scientists, engineers, and other users who have no
prior knowledge in the field of knowledge graphs, to become
familiar with the principles of knowledge graphs, to explore
new data, to change entities in the interactive visualization
and execute commands, to write and execute SPARQL queries
directly on the computational book and update interactive
visualizations with data from new entities. After the entities
are changed, we have interactive visualizations but with data
from the new entities. This can encourage the creation of more
such web applications (computational books).
The proposed architecture for building such interactive
applications (computational books) uses technologies such as
Jupyter Book, and Binder which are open-source, and will
allow for the creation of new such applications that will be
built with the technologies and principles of this architecture,
or can enable modifications of existing applications which are
built in the same way. The interactive in-line way of executing
the code and the web interface will allow small changes in
the code that users make when browsing the computational
book to return new data that had not been seen before and
to be displayed in the existing interactive visualization. It is
also possible to work with a completely different knowledge
graph, but in that case the parameters should be changed, or
if that part is not parameterized, the users need to change the
SPARQL queries. Since the changes are made inline in the
interactive web environment, the users have the opportunity
to execute the code and get new data and new visualizations.
From a developement perspective, one of the most important
parts is setting up a reproducible environment.
The proposed architecture uses Jupyter Book [1], which
represents a collection of Jupyter Notebooks [2]. Jupyter Book
uses Binder [3], which is a code repository that contains code
or content that you would like other people to run. This might
be a Jupyter Notebook, or an script in a different language,
such as R or Julia. Jupyter Book uses Binder to provide a
reproducible environment, but also implements an in-line code
execution feature directly on the web page.
The rest of the paper is organized as follows: the related
work is presented in Section II. In Section III we provide
details on these technologies and how they take part in this
architecture. Section IV presents a use case that was build
with this proposed architecture, where users can see the pub-
licly hosted computational book, and explore new data from
knowledge graphs with the interactive visualisations. Section
V contains a discussion about the potential of building such
computational books for exploring new data from knowledge
graphs.
II. RE LATE D WOR K
The uptake and consumption of Linked Data is currently
restricted almost entirely to the Semantic Web community.
While the utility of Linked Data to non-tech savvy web users is
evident, the lack of technical knowledge and an understanding
of the intricacies of the semantic technology stack limit such
users in their ability to interpret and make use of the Web
of Data [4]. A key solution in overcoming this hurdle is
to visualize Linked Data in a coherent and legible manner,
allowing non-domain and non-technical audiences to obtain
a good understanding of its structure, and therefore implic-
itly compose queries, identify links between resources and
intuitively discover new pieces of information [4]. Extensive
survey was provided in [4] of current efforts in the Semantic
Web community with respect to our requirements, and identify
the potential for visual support to lead to more effective,
intuitive interaction of the end user with Linked Data.
Casual users find it difficult to explore and use Semantic
Web data due to the prevalence of specialized browsers that
require complex queries to be formed and intimate knowledge
on the structure of datasets [5]. New approach makes it possi-
ble to obtain an overview of the dataset being explored using
techniques, such as navigation menus, treemaps or sitemaps,
which are usually not available in text-based Semantic Web
browsers. From there, users can interactively explore the data
using facets. Moreover, facets also feature a pivoting opera-
tion, motivated during tests with lay users, that removes the
main constraint of most faceted browsers, i.e. the inability to
combine filters for differently faceted views to build complex
queries [5].
In [6], a novel interactive visualization platform is pre-
sented, called ALOHA. Their study showed that graph-based
interactive visualization is a novel and acceptable approach to
end-users who are interested in seeking online health informa-
tion of various domains. They use user-centered design process
to create a user-friendly web-based application, ALOHA, for
the general public, to explore knowledge relevant to their
needs, through an iterative development process. Moreover,
their study showed that graph-based interactive visualization
is promising in helping consumers explore complex health
concepts quickly and can potentially lead to a new way of
finding and consuming health information online [6]. A novel
knowledge exploration mechanism such as ALOHA maybe
appealing to the general public and can serve as a template for
developing consumer-facing, evidence-based (i.e., supported
by scientific literature) knowledge graphs in many other health
and disease domains [6].
In the paper [7] modules were designed with an emphasis
on ease of use, query flexibility, and interactive visualization
of results. They presented LDlink, a web-based collection of
bioinformatic modules that query single nucleotide polymor-
phisms (SNPs) in population groups of interest to generate
haplotype tables and interactive plots. They have developed
different modules, fore example LDhap calculates population
specific haplotype frequencies of all haplotypes observed for a
list of query SNPs. They use predefined modules, which in our
architecture are predefined interactive visualisations. Users can
change entities in the interactive visualisations and can explore
new data from the knowledge graph.
Two applications were presented in [8], Thinkbase and
Thinkpedia, which aim to make web content more accessible
and usable by utilizing visualizations of the semantic graph as
a means to navigate and explore large knowledge repositories.
Both of the applications implement a similar concept: they
extract semantically enriched contents from a large knowl-
edge spaces (Freebase and Wikipedia, respectively), create an
interactive graph-based representation out of it, and combine
them into one interface together with the original text based
content.
There is a interactive tool which was presented in [9], for
exploring and visualizing large RDF ontologies. Primary goal
is to present the information to the user in a simple and
intuitive way. It provides an environment where users can
select a small set of objects to examine dynamically in real-
time, providing better contextual information.
Exploring and visualizing very large datasets has become
a major research challenge, of which scalability is a vital
requirement. In the survey [10], they describe the major
prerequisites and challenges that should be addressed by the
modern exploration and visualization systems. Considering
these challenges, they present how state-of-the-art approaches
from the database and information visualization communities
attempt to handle them. Finally, they survey the systems
developed by the Semantic Web community in the context
of the Web of Linked Data, and discuss to which extent these
satisfy the contemporary requirements.
III. ARCHITECTURE
Figure 1 presents the system architecture. Starting from
the left side of the figure, we will store the code in a
repository with a code versioning system. Hosting the en-
tire computational book will be achieved with the help of
the next component of the architecture. In our repository
we will have the files needed to configure and create an
executable, reproducible environment. These configuration
files are used by Repo2Docker, which is a tool to build,
run, and push Docker images from source code repositories.
Repo2Docker will create a Docker container which will then
be used by Binder to host Jupyter Notebooks. The config-
uration files that Repo2Docker uses to create a container
can be: requirements.txt,Dockerfile,apt.txt,
environment.yaml,setup.py, etc. So far we have a
code repository, in which we have a configuration file or a
combination of multiple files, with which Binder creates a
reproducible environment to run and host Jupyter Notebooks.
That environment will include all dependencies that we need
to execute the code in the Jupyter Notebook cells. In the
configuration files, we will set the dependencies for SPARQL,
RDF, RDFLib, visualization libraries and any other depen-
dencies that can enable executable environment. With this, we
have a reproducible environment where users can run Jupyter
Binder
JupyterBook
Repository
Repo2Docker
Executable
Reproducible
Environment
Parametarized
Knowledge
Graph
Visualisation
Jupyter
Notebook
RDF
SPARQL
Data
Visualisation
Library
Jupyter
Notebook
Hosting
Fig. 1. Architecture overview.
Notebook cells without installing dependencies and we have
a hosted Jupyter Notebook, i.e. users will not run Jupyter
Notebook locally. The next component of the architecture is
Jupyter Book, which is a collection of Jupyter Notebooks that
are displayed in the form of a web site. For code execution
you have two options in each Jupyter Notebooks. The first
option is to create a Binder environment and then directly
execute the cells, while changing the visualization parameters
with new entities so that we can discover new data from
the knowledge graphs, and the second option is to create
a Binder session that will be opened in a new browser tab
and we will have standard Jupyter Notebook environment or
JupyterLab. When building such computational books, with
this architecture on each of the pages (Jupyter Notebooks)
we will have predefined and parameterized visualizations, for
example TreeMap, where users will be able to change the
entities and populate the visualizations with new data from
the knowledge graph. Libraries and data visualization tools
are used to create parameterized visualizations. The data from
the knowledge graphs are retrieved and manipulated with
SPARQL, the RDF query language.
IV. USE CA SE : EXPLORING MEDICAL DATA FRO M
DBPEDIA WITH A COM PU TATIONAL BO OK
With the help of the architecture proposed in this paper,
a use-case was developed for exploring DBpedia data related
to medicine. The code is hosted on GitHub. In addition to
the code versioning system, GitHub provides hosting to static
sites through Github Pages. The advantage of hosting Jupyter
Book with GitHub Pages is that you only need to enable
GitHub Pages in repository settings. In the repository there
is a Dockerfile file which is one of the configuration
file options, which will be used by Repo2Docker to create a
reproducible environment in Binder, which will allow users to
execute the code in Jupyter Book. Jupyter Book is a collection
of Jupyter Notebooks. Each of the Jupyter Notebooks has
Plotly [11] visualizations. Figure 3 presents only one of
the ready-made parameterized visualizations, and that is the
TreeMap of the entities whose activity sector is medicine:
?Thing dbp:activitySector dbr:Medicine.
The visualizations are parameterized and they use data
from the DBPedia knowledge graph. In this use case the
data are taken from the DBPedia SPARQL endpoint. In the
specific visualization with TreeMap, the parameters can be
other entities instead of ?Thing dbp:activitySector
dbr:Medicine. You can completely change the SPARQL
endpoint to have a new knowledge graph and new entities,
which will be displayed in the existing visualization. That
change of entities or SPARQL endpoints can be done in the
cells that are directly in the computational book. Figure 2
shows the cells where the code can be changed, and the
parameters of the visualization, Figure 3 shows the TreeMap
which is one of the visualisations in this use case. This way,
users can discover new data from knowledge graphs. The code
from this use case is open source1, and the hosted version is
available online2.
V. DISCUSSION
Discovering new data is important for creating and devel-
oping new applications for machine learning, data science
or creating Semantic Web applications where data will be
retrieved from knowledge graphs. The combination of tech-
nologies used in the proposed architecture provides ways
to discover new data and provide view of the data from a
particular perspective through predefined visualizations. First,
technologies that allow us to look at data from a different angle
rather than tables or spreadsheet are Knowledge Graphs which
are part of the Semantic Web and Linked Data technologies.
The graph structure allows us to discover new data, over and
over again, so that each subsequent execution of the code
will potentially return different data. As a second point, the
technologies that allow us to have a reproducible environment
and executable code in a web environment in the form of
a website are Jupyter Book, Jupyter Notebook and Binder.
Last, the various visualization libraries will enable us to
present data in a way that gives additional context to the
data. Parameterization of these visualizations will provide a
generic definition that can be reused with other entities or
other knowledge graphs.
This architecture aims to encourage users to instantly inter-
act with data and discover new data. We want to encourage the
development of applications with this proposed architecture,
so users can explore new data from knowledge graphs thought
interactive parameterized visualisations.
1https://github.com/dimitarmileski/Medical-KGs-WBS
2https://dimitarmileski.github.io/Medical-KGs-WBS/intro.html
Fig. 2. SPARQL query cell execution.
Fig. 3. Parameterized treemap visualisation.
REFERENCES
[1] E. B. Community, “Jupyter book,” Feb 2020.
[2] T. Kluyver, B. Ragan-Kelley, F. P´
erez, B. E. Granger, M. Bussonnier,
J. Frederic, K. Kelley, J. B. Hamrick, J. Grout, S. Corlay et al.,
Jupyter Notebooks-a publishing format for reproducible computational
workflows., 2016, vol. 2016.
[3] Project Jupyter, Matthias Bussonnier, Jessica Forde, Jeremy Freeman,
Brian Granger, Tim Head, Chris Holdgraf, Kyle Kelley, Gladys Nalvarte,
Andrew Osheroff, M. Pacer, Yuvi Panda, Fernando Perez, Benjamin Ra-
gan Kelley, and Carol Willing, “Binder 2.0 - Reproducible, interactive,
sharable environments for science at scale,” in Proceedings of the
17th Python in Science Conference, Fatih Akici, David Lippa, Dillon
Niederhut, and M. Pacer, Eds., 2018, pp. 113 – 120.
[4] A.-S. Dadzie and M. Rowe, “Approaches to Visualising Linked Data:
A Survey,” Semantic Web, vol. 2, no. 2, pp. 89–124, 2011.
[5] J. M. Brunetti, R. Garc´
ıa, and S. Auer, “From Overview to Facets and
Pivoting For Interactive Exploration of Semantic Web Data,” Interna-
tional Journal on Semantic Web and Information Systems (IJSWIS),
vol. 9, no. 1, pp. 1–20, 2013.
[6] X. He, R. Zhang, R. Rizvi, J. Vasilakes, X. Yang, Y. Guo, Z. He,
M. Prosperi, J. Huo, J. Alpert et al., “ALOHA: Developing an Inter-
active Graph-Based Visualization for Dietary Supplement Knowledge
Graph Through User-Centered Design,” BMC Medical Informatics and
Decision Making, vol. 19, no. 4, pp. 1–18, 2019.
[7] M. J. Machiela and S. J. Chanock, “LDlink: A Web-Based Application
for Exploring Population-Specific Haplotype Structure and Linking Cor-
related Alleles of Possible Functional Variants,” Bioinformatics, vol. 31,
no. 21, pp. 3555–3557, 2015.
[8] C. Hirsch, J. Hosking, and J. Grundy, “Interactive Visualization Tools
for Exploring the Semantic Graph of Large Knowledge Spaces,” in
Workshop on Visual Interfaces to the Social and the Semantic Web
(VISSW2009), vol. 443, 2009, pp. 11–16.
[9] L. Deligiannidis, K. J. Kochut, and A. P. Sheth, “RDF Data Exploration
and Visualization,” in Proceedings of the ACM First Workshop on
CyberInfrastructure: Information Management in eScience, 2007, pp.
39–46.
[10] N. Bikakis and T. Sellis, “Exploration and Visualization in the Web of
Big Linked Data: A Survey of the State of the Art,” arXiv preprint
arXiv:1601.08059, 2016.
[11] P. T. Inc. (2015) Collaborative data science. Montreal, QC. [Online].
Available: https://plot.ly
