https://doi.org/10.31449/inf.v46i1.3375 Informatica46 (2022) 49–55 49
AGlobalCOVID-19Observatory,MonitoringthePandemicsThroughText
MiningandVisualization
M. Besher Massri
1;2
, Joao Pita Costa, Marko Grobelnik, Janez Brank, Luka Stopar and Andrej Bauer
3
E-mail: besher.massri@ijs.si, andrej.bauer@andrej.com
1
Jožef Stefan Institute, Slovenia
2
Jožef Stefan International Postgraduate School, Ljubljana, Slovenia
3
University of Ljubljana, Slovenia
Keywords: text mining, data analytics, data visualisation, public health, Coronavirus, COVID-19, epidemic intelligence
Received: November 23, 2020
The global health situation due to the SARS-COV-2 pandemic motivated an unprecedented contribution of
science and technology from companies and communities all over the world to fight COVID-19. In this
paper, we present the impactful role of text mining and data analytics, exposed publicly through IRCAI’s
Coronavirus Watch portal. We will discuss the available technology and methodology, as well as the on-
going research based on the collected data.
Povzetek: Opisana je vloga rudarjenja besedil in podatkovne analitike na primeru portala IRCAI’s Coro-
navirus Watch.
Figure 1: Coronavirus Watch portal
1 Introduction
When the World Health Organization (WHO) announced
the global COVID-19 pandemic on March 11th 2020 [36],
following the rising incidence of the SARS-COV-2 in Eu-
rope, the world started reading and talking about the new
Coronavirus. The arrival of the epidemic to Europe scaled
out the news published about the topic, while public health
institutions and governmental agencies had to look for ex-
isting reliable solutions that could help them plan their ac-
tions and the consequences of these. Technological compa-
nies and scientific communities invested efforts in making
available tools (e.g. the GIS [2] later adopted by the World
Health Organisation (WHO)), challenges (e.g. the Kag-
gle COVID-19 competition [22]), and scientific reports and
data (e.g. the repositories medRxiv [24] and Zenodo [38]).
In this paper we discuss the Coronavirus Watch portal [21],
made available by the UNESCO AI Research Institute (IR-
CAI), comprehending several data exploration dashboards
related to the SARS-COV-2 worldwide pandemic (see the
main portal in Figure 1). This platform aims to expose the
different perspectives on the data generated and trigger ac-
tions that can contribute to a better understanding of the
behavior of the disease.
2 Relatedwork
With with growing dimension of the public health problem
caused by the SARS-COV-2, the global effort multiplied
worldwide: from the publicly shared research in, e.g., Zen-
odo [38] medRxiv [24] and IEEE Xplore [20]; to a diver-
sity of datasets made available by the global agencies (as
e.g., the European Commission [3] and the ECDC [6]), the
tech giants (as, e.g., Google [7] and Facebook [15]), insti-
tutional initiatives (as, e.g., the OxCOVID19 project [34]),
and private initiatives acting globally (e.g. [37]) and lo-
cally (e.g., the slovenian initiative [32]). With similar aim
to contribute for the global effort, many platforms have
been made publicly available over the internet to monitor
aspects of the COVID-19 pandemics are mostly focusing
on data visualization based on the incidence of the disease
and the death rate worldwide (e.g., the CoronaTracker [4]).
The limitations of the available tools are potentially due to
the lack of resolution of the data in aspects like the geo-
graphic location of reported cases, the commodities (i.e.,
other diseases that also influence the death of the patient),
the frequency of the data, etc. On the other hand, it was
not common to monitor the epidemic through the world-
wide news (with some exceptions as, e.g., the Ravenpack
Coronavirus News Monitor [31]). This is a rather impor-
tant perspective mostly due to the great importance that the
related misinformation - known as infodemia[1] - has been
playing a role in the context of this pandemics.

50 Informatica46 (2022) 49–55 M.B. Massri et al.
The Coronavirus Watch portal suggests the association
of reported incidence with worldwide published news per
country, which allows for real-time analysis of the epi-
demic situation and its impact on public health (in which
specific topics like mental health and diabetes are impor-
tant related matters) but also in other domains (such as
economy, social inequalities, etc.). This news monitoring
is based on state-of-the-art text mining technology aligned
with the validation of domain experts that ensures the rele-
vance of the customized stream of collected news.
Moreover, the Coronavirus Watch portal offers the user
other perspectives of the epidemic monitoring, such as the
insights from the published biomedical research that will
help the user to better understand the disease and its im-
pact on other health conditions. While related work was
promoted in [22] in relation with the COVID-19, and is of-
fered in general by MEDLINE mining tools (e.g., MeSH
Now [25]), there seems to be no dedicated tool to the mon-
itoring and mining of COVID-19 - related research as that
presented here.
3 Descriptionofdata
The nature of the proposed global observatory system en-
tails a range of heterogeneous data sources that entangle to
provide a perspective as complete as possible on the mat-
ters of the COVID-19 pandemics. In the following section
we discuss these data sources and their potential contribu-
tions.
3.1 HistoricalCOVID-19data
To perform an analysis of the growth of the coronavirus,
we use the historical data on the daily number of cases
of infections and deaths. This data is retrieved from a
GitHub repository made available by the John Hopkins
University [5]. The data source is based mainly on the offi-
cial data from the World Health Organization (WHO)[28]
along with some other sources provided by, e.g., the Center
for Disease and Control[17], and Worldometer[37], among
others. This data provides the basis for all the system’s
functionality that depended on the statistical information
about SARS-COV-2 incidence.
3.2 Livedatafromworldometer
Apart from historical data, live data about the COVID-19
number of infection cases, deaths, recovered individuals,
and tests made is retrieved from the Worldometer web por-
tal [37]. Although the information might not be taken as
official as the one provided by John Hopkins University
(which is based on WHO data), this source is updated many
times per day providing the latest up-to-date data about
COVID-19 statistics at all times. Thus its usefulness in
providing the system with an almost real-time perspective
on the pandemics worldwide.
3.3 Livenewsaboutcoronavirus
The live news is retrieved from the global news engine
Event Registry [16], which is a media-intelligence plat-
form that collects news media from around the world in
many languages. The service analyzes news from more
than 30 thousand news, blog posts, and press releases daily,
over around 75 thousand news sources in more than 60 lan-
guages. The multilingual capabilities of the system allow
for the identification of topics across languages, ensuring a
global coverage with local granularity [18]. With this data
we are able to access what is the awareness of the media on
aspects and key figures in the pandemics.
3.4 GoogleCOVID-19communitymobility
data
Google’s Community Mobility [19] data compares mobil-
ity patterns that date from before the COVID-19 crisis, de-
scribing the situation on a weekly basis. Mobility patterns
are measured as changes in the frequency of visits to six lo-
cation types: retail and recreation; grocery and pharmacy;
parks; transit stations; workplaces; and residential areas.
The data is provided on a country level as well as on a
province level. These reports are available for a limited
amount of time, limited to their usefulness in supporting
the work of public health officials. This data provides us
with a wide perspective on the exchanges with potential of
contamination based on mobility.
3.5 MEDLINE:medicalresearchopen
dataset
In 2020 the MEDLINE dataset [23] contains more than 30
million citations and abstracts of the biomedical literature
dating back to 1966. Over the past ten years, an average
of a million articles were added each year. Around 5%
of MEDLINE is on published research on infections, with
cancer research being the most prevalent occuping 12% of
this body of knowledge. Most scientific articles in this
dataset are hand-annotated by health experts using 16 ma-
jor categories and a maximum of 13 levels of deepness.
The labeled articles are hand-annotated by humans based
on their main and complementary topics, and on the chem-
ical substances that they relate to. It is widely used by the
biomedical research community through the well-accepted
search engine PubMed [29]. The richness of this data al-
lows us for insight on aspects of the disease as well as ap-
proaches and best practices from the published research.
4 Coronaviruswatchdashboard
The main layout of the dashboard displayed in figure 1 con-
sists of two sides. On the left side, the dashboard is split
into the summarized information on the incidence across
countries, where a simple table of statistics is provided
about countries along with the total numbers of cases, death

A Global COVID-19 Observatory, Monitoring the. . . Informatica46 (2022) 49–55 51
cases, and recovered cases. On the right side, there is a
navigation panel with tabs, each representing a functional-
ity. Each of these tabs is a view on the pandemics, based on
its own specific data sources, and answers some questions
and provides insights about a certain type of data focusing
a specific aspect.
4.1 Coronavirusdatatable
The data table functionality is a straightforward data visu-
alisation that shows the basic statistics about the new coro-
navirus. It’s sourced from Worldometer as it’s the most
frequently updated source for coronavirus. This data table
extends the one that is consistently shown in most views, a
summarized version, to the table on the left containing the
full information details. These include the new infection
cases and the new death cases, in the day of the visit to the
portal, as well as the total number of recovered cases, the
active cases and the critical cases. The cumulative counts
include the total number of infected, dead and tests. The
latter are provided in absolute and per one million capita,
taking into consideration the different population per coun-
try.
4.2 Coronaviruslivenews
The news monitoring functionality offers a live news feed
about coronavirus-related topics from sources around the
world. The feed is provided by the multilingual news en-
gine Event Registry over an API. The ingested news arti-
cle sample is generated by querying the system for articles
that are annotated with concepts and keywords related to
the coronavirus. The user can check for a country’s spe-
cific news (based on the news source in that country) by
clicking on the country name on the left table, as seen in
figure 1.
This feature is a differentiator to most COVID-19 obser-
vatories, allowing the visitor to access real-time informa-
tion about the pandemics, either at a global scale, in the
European context or at country level. It also allows for a
perspective on the public awareness on aspects of the pan-
demics.
4.3 Exploratoryanalytics
The following set of data visualization modules aim at
displaying the statistical data about COVID-19 cases and
deaths. While they all provide countries comparison, each
one focus on a different perspective. Some are more com-
plex and focused on the big picture (5D evolution), and
some others are simple and focus on a unique aspect (Pro-
gression and Trajectory). Moreover, all of them have con-
figuration options to tweak the visualization, like the ability
to change the scale of the axes to focus on the top countries,
or to focus on the long tale. Some offer a slider to manu-
ally move through the days for further inspection. Further-
more, the default view compares all the countries or the
Figure 2: A snapshot of the 5D Visualization on March
23rd. Countries that were at the peak in terms of growth
are shown high up.
top N countries, depending on the visualization selected.
However, it is possible to track a single country or a set
of countries and compare them together for a more spe-
cific view. This is done when selecting the main country
by clicking on it on the left table and proceeding to select
more countries by pressing the ctrl key while clicking on
the additional country.
4.3.1 5Devolution
5D Evolution is a visualization that displays the incidence
of the virus in the population through time. It is named 5D
since it encompasses five dimensions: x-axis, y-axis, bub-
ble size, bubble color, and time, as seen in figure 2. By
default, it illustrates the evolution of the pandemic in coun-
tries based on N cases (x-axis). The growth factor of N
Cases (y-axis), N Deaths (bubble size), and country region
(bubble color) through time. In addition, a red ring around
the country bubble is drawn whenever the first death ap-
pears. The growth rate represents how likely the numbers
are increasing with respect to the day before. A growth
rate of 2 means that the numbers are likely to double in the
next day. The growth rate is calculated using the exponen-
tial regression model. At each day the growth rate is based
on the N cases from the previous seven days. The goal of
this visualization is to show how countries relate to each
other, which are "exploding in numbers" and which ones
managed to "flatten the curve" (since flattening the curve
means less growth rate). It is intended to be one visualiza-
tion that gives the user a big picture of the situation.
4.3.2 Progression
The progression visualization displays the line graph com-
paring Date vs N cases/deaths. It helps to provide a sim-
plistic view of the situation and compare countries based
on the raw numbers only. The user can display the cumula-
tive numbers where each day represents the current counts,
or daily where at each date we count the cases/deaths on
that day only.

52 Informatica46 (2022) 49–55 M.B. Massri et al.
4.3.3 Trajectory
While the progress visualization displays the normal date
vs N cases/deaths, this visualization seeks to compare how
the trajectory of the countries differ starting from the point
where they detect cases. This visualization helps to com-
pare countries’ situations if they all start having cases on
the same date. The starting point has been set to the day
the country reaches 100 cases, so we would compare coun-
tries when they started gaining momentum.
4.4 Timegap
The time gap functionality tries to estimate how the coun-
tries are aligned and how many days each country is behind
the other, whether that is in the number of cases or deaths.
This assumes that the trajectory of the country will con-
tinue as is without taking much more strict/loose measure-
ments, which is a rough assumption. It helps to estimate
how bad or good is the situation in terms of the number
of days. To see the comparison, a country has to be se-
lected from the table on the left. However, not all countries
are comparable, having very different trajectories or growth
rates.
The growth of each country is represented as an expo-
nential function, the base is calculated using linear regres-
sion on the log of the historical values (that is, exponential
regression). Based on that, the duplication N days, or the N
days representing the number of cases/deaths will double
is determined. Two countries are comparable if they have a
reasonable difference in the base or doubling factor. If they
are comparable, we see where the country with the smaller
value fits in the historical values of the country with the
larger numbers. We use linear interpolation if the number
is not exact, hence the decimal values.
4.5 Mobility
The mobility visualization is based on google community
mobility data that describe how communities in each coun-
try are moving based on 6 parameters: retail and recre-
ation, grocery and pharmacy, parks, transit stations, work-
places, and residential areas. The data is then reduced to
2-dimensional data while keeping the Euclidean proxim-
ity nearly the same. The visualization can indicate that
the closer the countries are on the visualization, the sim-
ilar the mobility patterns they have. The visualization uses
the T-SNE algorithm for dimensionality reduction [35],
which reduces high dimensional data to low dimensional
one while keeping the distance proximity between them
proportionally the same as possible. The algorithm works
in the form of iterations, at each iteration, the bubbles rep-
resenting the country are drawn. We used those iterations
to provide animation to the visualization.
Figure 3: A snapshot of the Social Distancing Simulator.
The canvas show a representation of the population. with
red dots representing sick people, yellow dots representing
immunized people, and grey dots represent deceased peo-
ple.
4.6 Socialdistancingsimulator
The Social Distancing simulator is displayed in figure 3.
Each circle represents a person who can be either healthy
(white), immune (yellow), infected (red), or deceased
(gray). A healthy person is infected when they collide with
an infected person. After a period of infection, a person
either dies or becomes permanently immune. Thus the
simulation follows the Susceptible-Infectious-Recovered-
Deceased (SIRD) compartmental epidemiological model.
The simulator is controlled by three parameters. Firstly,
the Social distancing that controls to what extent the popu-
lation enforces social distancing. At 0% there is no social
distancing and persons move with maximum speed so that
there is a great deal of contact between them. At 100%
everyone remains still and there is no contact at all. Sec-
ondly, mortality is the probability that a sick person dies. If
you set mortality to 0% nobody dies, while the mortality of
100% means that anybody who catches the infection will
die. Finally, the infection duration determines how long a
person is infected. A longer time gives an infected person
more opportunities to spread the infection. Since the simu-
lation runs at high speed, time is measured in seconds.
4.7 Biomedicalresearchexplorer
To better understand the disease, the published biomedical
science is the source that provides accurate and validated
information. Taking into consideration a large amount
of published science and the obstacles to access scien-
tific information, we made available a MEDLINE explorer
where the user can query the system and interact with a

A Global COVID-19 Observatory, Monitoring the. . . Informatica46 (2022) 49–55 53
Figure 4: The interactive MEDLINE data exploration
showing a scientific article originally positioned as 182nd,
now in the 4th place.
pointer to specify the search results (e.g., obtaining results
on biomarkers when searching for articles hand-annotated
with the MeSH class "Coronavirus").
To allow for the exploration of any health-related texts
(such as scientific reports or news) we developed an au-
tomated classifier [8] that assigns to the input text the
MeSH classes it relates to. The annotated text is then
stored in Elasticsearch [27], from where it can be accessed
through Lucene language queries, visualized over easy-to-
build dashboards, and connected through an API to the ear-
lier described explorer (see figure 4 and read [12], [30] and
[26] for more detail).
The integration of the MeSH classifier with the world-
wide news explorer Event Registry allows us to use MeSH
classes in the queries over worldwide news promoting an
integrated health news monitoring [13] and trying to avoid
bias in this context [11]. An obvious limitation is a fact that
the annotation is only available for news written in the En-
glish language, being the unique language in MEDLINE.
5 Conclusionandfuturework
In this paper, we presented the coronavirus watch dash-
board as a use-case of observing pandemic. However,
this methodology can be applied to other kinds of diseases
given the availability of similar data. Having this global
system in place we are collecting data on the usual num-
bers accounted on the pandemics by other platforms (as in
e.g., Worldometer [37] or ECDC [6]) but within an inte-
grated approach (as e.g. in the OxCOVID19 project [34])
aiming for data interoperability enhancing the business in-
telligence generated by the opservatory For further devel-
opment, we plan to implement a local dashboard for other
countries as well, which would provide local data in the lo-
cal language. This would provide the local coverage (as in,
e.g., Slednik [32]) that can bring the global problems ad-
dressed to a specific and more usefu context (as was initi-
ated in the context of the MIDAS project [12]). In addition,
given the existence of more than seven months of histori-
cal data, we would like to build some predictive models to
predict the number of cases/deaths in the next days.
Moreover, we are using the StreamStory technology [33]
[14] in order to: (i) compare the evolution of the dis-
ease between countries by comparing their time-series of
incidence; (ii) investigate the correlation between the in-
cidence of the disease with weather conditions and other
impact factors; and (iii) analyze the dynamics of the evo-
lution of the disease based on incidence, morbidity, and
recovery. This technology allows for the analysis of dy-
namical Markov processes, analyzing simultaneous time-
series through transitions between states, offering several
customization options and data visualization modules.
Furthermore, following the work done in the context of
the Influenza epidemic in [9], we are using Topological
Data Analysis methods in [10] to understand the behav-
ior of COVID-19 in the European context [10] through
the behaviour of the data it is reflected in. This is done
analysing simultaneously a variety of time-series represent-
ing the several aspects of the pandemics, in a high dimen-
sion space. In it, we examine the structure of data through
its topological structure, which allows for comparison of
the evolution of the epidemics within countries through the
encoded topology of their incidence time series.
Acknowledgement
The first author has been supported by the Knowledge 4 All
foundation and the H2020 Humane AI project under the
European research and innovation programme under GA
No. 761758), while the second author was funded by the
European Union research fund ’Big Data Supporting Pub-
lic Health Policies’, under GA No. 727721. The third au-
thor acknowledges that this material is based upon work
supported by the Air Force Office of Scientific Research
under award number FA9550-17-1-0326.
References
[1] H. Allahverdipour. 2020. Global challenge of health
communication: infodemia in the coronavirus dis-
ease (covid-19) pandemic. J Educ Community
Health, 7, 2, 65.–67.
[2] ArcGIS. 2020. ArcGIS who covid-19 dashboard.
https://covid19.who.int/. (2020).
[3] European Commission. [n. d.] The european
covid-19 data portal. https : / / www .
covid19dataportal . org/. Accessed in:
April 2021. ().
[4] CoronaTracker. 2020. CoronaTracker. https://
www.coronatracker.com/analytics/.
(2020).

54 Informatica46 (2022) 49–55 M.B. Massri et al.
[5] CSSE. 2020. Covid-19 data repository by the center
for systems science and engineering (csse) at johns
hopkins university. https://github.com/
CSSEGISandData/COVID-19. (2020).
[6] ECDC. [n. d.] Ecdc covid-19 datasets. https://
www.ecdc.europa.eu/en/covid-19/
data. Accessed in: April 2021. ().
[7] B. Xu et al. 2020. Epidemiological data from the
covid-19 outbreak, real-time case information. Sci-
entific data, 7, 1, 1–6. DOI: 10.1038/s41597-
020-0448-0.
[8] J. Pita Costa et al. 2021. A new classifier designed
to annotate health-related news with mesh headings.
Artificial Intelligence in Medicine, 114, 102053.
DOI:10.1016/j.artmed.2021.102053.
[9] J. Pita Costa et al. 2019. A topological data analysis
approach to the epidemiology of influenza. In Pro-
ceedings of the Slovenian KDD conference.
[10] J. Pita Costa et al. 2020. A topological data analysis
perspective on the covid-19 pandemics. In In prepa-
ration.
[11] J. Pita Costa et al. 2019. Health news bias and its
impact in public health. In Proceedings of the Slove-
nian KDD conference.
[12] J. Pita Costa et al. 2020. Meaningful big data inte-
gration for a global covid-19 strategy. Computer In-
telligence Magazine, 15, 4, 51–61. DOI:10.1109/
MCI.2020.3019898.
[13] J. Pita Costa et al. 2017. Text mining open datasets
to support public health. In WITS 2017 Conference
Proceedings.
[14] L Stopar et al. 2018. Streamstory: exploring multi-
variate time series on multiple scales. IEEE trans-
actions on visualization and computer graphics, 25,
4, 1788–1802. DOI: 10.1109/TVCG.2018.
2825424.
[15] T. Kuchler et al. 2020. The geographic spread of
covid-19 correlates with the structure of social net-
works as measured by facebook. National Bureau of
Economic Research, w26990. DOI: 10.1016/j.
jue.2020.103314.
[16] EventRegistry. 2020. Event Registry. https://
eventregistry.org. (2020).
[17] Center for Disease Control. 2020. Center for Disease
Control and Prevention. https://www.cdc.
gov/coronavirus/2019-ncov/index.
html. (2020).
[18] J. Brank G. Leban B. Fortuna and M. Grobelnik.
2014. Event registry: learning about world events
from news. In In Proceedings of the 23rd Interna-
tional Conference on World Wide Web, 107–110.
DOI:10.1145/2567948.2577024.
[19] Google. 2020. Google COVID-19 Community Mo-
bility Report. https://www.google.com/
covid19/mobility/. (2020).
[20] IEEE. [n. d.] Ieee xplore covid-19 resources.
https : / / ieeexplore . ieee . org /
Xplore/home.jsp. Accessed in: April 2021. ().
[21] IRCAI. 2020. IRCAI coronavirus watch portal.
http://coronaviruswatch.ircai.org/.
(2020).
[22] Kaggle. 2020. Kaggle covid-19 open research
dataset challenge. https://www.kaggle.
com/allen-institute-for-ai/CORD-
19-research-challenge. (2020).
[23] MEDLINE. 2020. MEDLINE description of the
database.https://www.nlm.nih.gov/bsd/
medline.html. (2020).
[24] medRxiv. [n. d.] Covid-19 sars-cov-2 preprints from
medrxiv and biorxiv. https : / / connect .
medrxiv.org/relate/content/181. Ac-
cessed in: April 2021. ().
[25] MeSHNow. 2020. MeSHNow. https://www.
ncbi.nlm.nih.gov/CBBresearch/Lu/
Demo/MeSHNow/. (2020).
[26] MIDAS. 2020. MIDAS COVID-19 portal. http:
//www.midasproject.eu/covid-19/.
(2020).
[27] Elastic NV. 2020. Elasticsearch portal.https://
www.elastic.co/. (2020).
[28] World Health Organisation. 2020. WHO Coro-
navirus portal. https : / / www . who .
int / emergencies / diseases / novel -
coronavirus-2019. (2020).
[29] PubMed. 2020. PubMed biomedical search engine.
https://pubmed.ncbi.nlm.nih.gov/.
(2020).
[30] Quintelligence. 2020. Quintelligence COVID-19
portal. http://midas.quintelligence.
com/. (2020).
[31] Ravenpack. 2020. Ravenpack coronavirus
news monitor. https : / / coronavirus .
ravenpack.com/. (2020).
[32] Slednik. [n. d.] Covid-19 sledilnik slovenija.
https://www.sledilnik.org. Accessed in:
April 2021. ().
[33] Luka Stopar. 2020. StreamStory. http : / /
streamstory.ijs.si/. (2020).
[34] Oxford University. [n. d.] Oxford covid-19 (ox-
covid19) project.https://covid19.eng.ox.
ac.uk/. Accessed in: April 2021. ().
[35] Laurens van der Maaten and Geoffrey Hinton. 2008.
Viualizing data using t-sne. Journal of Machine
Learning Research, 9, (November 2008), 2579–
2605.

A Global COVID-19 Observatory, Monitoring the. . . Informatica46 (2022) 49–55 55
[36] WHO. 2020. World Health Organization who
director-general’s opening remarks at the media
briefing on covid-19 - 11 march 2020. https://
www.who.int/dg/speeches/detail/
who-director-general-s-opening-
remarks-at-the-media-briefing-on-
covid-19---11-march-2020. (2020).
[37] WorldoMeters. 2020. WorldoMeters. https://
www.worldometers.info/coronavirus/.
(2020).
[38] Zenodo. [n. d.] Zenodo coronavirus disease research
community - covid-19.https://zenodo.org/
communities/covid-19/. Accessed in: April
2021. ().

56 Informatica46 (2022) 49–55 M.B. Massri et al.