=Paper=
{{Paper
|id=Vol-2088/paper9
|storemode=property
|title=Natural Disaster Database Design and Development for Himalaya Using Social Media
|pdfUrl=https://ceur-ws.org/Vol-2088/paper9.pdf
|volume=Vol-2088
|authors=Kiran Zahra
|dblpUrl=https://dblp.org/rec/conf/agile/Zahra17
}}
==Natural Disaster Database Design and Development for Himalaya Using Social Media==
https://ceur-ws.org/Vol-2088/paper9.pdf
Natural Disaster Database Design and Development for Himalaya Using
Social Media
Kiran Zahra
Department of Geography, UZH
Winterthurerstr. 190, 8057 Zurich, Switzerland
kiran.zahra@geo.uzh.ch
Abstract
Ever increasing population of social media and increasing trend of sharing information related to several different topics has the great potential
to use this information for different purposes such as disaster management. The use of Volunteered Geographic Information (VGI) has its own
challenges because of the unstructured nature of social media such as extracting information out of noise and credibility related issues. This
project focusses on the core issues of using user generated content for disaster management in developing countries.
Keywords: VGI, Twitter, Credibility, toponym granularity.
1 Motivation and Project Objectives not only usable, but also very valuable in terms of
its contents (Kavanaugh & Yang, 2011). In case of
Frequency and impact of natural disasters is on disaster in a region the most important information
the rise worldwide (Kellenberg & Mobarak, 2008 ; for disaster response organizations are location,
Whybark, 2007). For more than fifty years, timing and impact (Jongman et al., 2015). Social
researchers have focused on a series of vital media has the potential to serve in all these aspects.
questions such as human occupancy of hazard The increasing trend of using smart phones and
zones, response of people and societies, and how to accessibility to fast internet connections are the
lessen the effects of environmental hazards (Cutter, main causes of rapid growth of social media
2012). A well-structured database about natural population (Kwak et al., 2010). Social media users
disasters can better answer these questions and can such as Twitter community tend to share bulk of
be an efficient source of analysing causes, effects, tweets publicly available and these tweets have
capacity, vulnerability and further planning of great potential to extract meaningful information
disaster management such as EM-DAT1. and use this information during emergency events.
The Himalayas are the youngest mountain range But it is often very difficult to filter only “useful
in the world. This mountain range is still information” from the Twitter stream by making
geologically active (Lavé & Avouac, 2000), which keyword based queries and thus it is recommended
causes a high occurrence of natural disasters in this to use machine leaning algorithms to filter
region. Documenting natural disasters is important information (Sakaki, Okazaki, & Matsuo, 2013).
in order to understand current issues and future Credibility of information shared on social media
predictions. Although this kind of information is is another challenge to using this type of data
available, it is technically incomplete and difficult during disasters. Mendoza, Poblete, & Castillo
to maintain. (2010) state in their research that rumours tend to
People share a bulk of information on social be questioned more than authentic news and
media sites such as Twitter during mass propose a framework to deal with credibility
emergencies (Hossmann et al., 2011 ; Terpstra et issues.
al., 2012). Information shared on social media is
1
D. Guha-Sapir, R. Below, Ph. Hoyois - EM-DAT:
International Disaster Database – www.emdat.be –
Université Catholique de Louvain – Brussels – Belgium
AGILE 2016 – Helsinki, June 14-17, 2016
Gelernter & Balaji (2013) discuss the increased retweets question the certainty and reliability of
use of geographical content in tweets during Twitter in disaster situations because during a
disasters. This geographic location information can disaster credibility of information is very
be used to manage rescue operations during and important. Semantic analysis of tweets and
after disasters and also for generating maps. keyword ambiguities have very little or no research
Although disaster relief is very important issue, but about tweets originating from Himalayan region
no research has done on utilizing VGI to design and and its surrounding countries, moreover the role of
create natural disaster database. geographic features has never been studied in
Overall aim of this study is to analyse different Naïve Bayes classification efficiency.
aspects of disaster related information shared on Twitter streaming Application Program Interface
Twitter about any disaster happening in Himalayan (API) downloads many fields called metadata of
region and its surrounding countries. This social tweet with tweet text field. This metadata can be
media data has the potential to contain helpful used to effectively assess tweet credibility (Canini,
geographic information and situational updates Suh, & Pirolli, 2011). Credibility assessment issues
during mass emergency events. These can be used are expected to be more complex due to cultural
by disaster response agencies for efficient and language diversity in the Himalayan region.
management. This research will help to investigate When there is a disaster, precise location
the issues brought about by natural hazard identification is a very important element in many
phenomena. situations. Disaster/emergency response
The objectives of this research are to: organizations can only respond if they know
• Apply machine learning algorithms to precise locations. People tend to add location
resolve keyword ambiguities and optimize tweet information when they share something about
classification disasters on social media. Lingad, Karimi, & Yin
• Assess credibility of tweets based on data (2013) discuss a tool “Named Entity Recognizers”.
quality standards of tweets shared during disasters They state that location extraction from microblogs
• Analyse the granularity of geographic is becoming more accurate and that such tools have
features reported in tweets during disasters, great potential to find out locational information
geoparsing and geocoding those geographic more correctly.
features to display on the map for visual Social media’s application to disaster
representations management is getting more and more important
• Conceptualize and implement a VGI based because of its real-time nature such as
natural disaster database TweetTracker (Kumar et al., 2011), Artificial
Intelligence for Disaster Response (AIDR) (Imran
2 State of the art et al., 2014), Twitcident (Abel et al., 2012),
ScatterBlogs for situational awareness (Thom et
Sakaki, Okazaki, & Matsuo (2013) discussed al., 2015).
‘semantic analysis of tweets’ in their research. Since forecasting disaster is difficult, efficient
They did this analysis to detect target events from disaster management is very important to improve
tweets. They also mentioned that identifying disaster response. There are many national and
keywords alone from tweets about disasters is not international organizations maintaining records of
enough to gather valuable information about the natural disasters such as EM-DAT, a famous global
event. Verma et al. (2011) studied tweets natural and technological disaster database. Such
‘contributing to situational awareness’ and tweets databases are created and maintained by data
‘not contributing to situational awareness’. They collected from different primary and secondary
used examples of tweets discussing the disaster but data sources. Witham (2005) discussed in his
not contributing to situational awareness, as they research various issues about completeness and
do not give any actionable information. accuracy of maintaining databases about disasters.
Acar and Muraki (2011) argue that it is often very He also presented a new database about volcanic
difficult to filter correct information out of the disasters and analyse this information for risk
Twitter data stream. Uncontrollable informal management.
AGILE 2016 – Helsinki, June 14-17, 2016
The role of social media as a great potential features reported during the event. These
source of data during disasters has already been geographic features then later be geoparsed and
discussed, this research focusses on using geocoded by available online platforms such as
information shared on Twitter to design a VGI address-parser and gisgraphy. Geoparsing is the
based database. Designing and developing VGI process of automatically identifying geographical
based databases is a novel concept for the information (names of places) from text (Gelernter
Himalayan region. This research gap formulated & Balaji, 2013). Geocoding is the process of
the following research questions: converting text into geographical coordinates
RQ 1: Which data mining techniques are (Ratcliffe, 2004). The spatial information
successful in identifying relevant tweets identified from the tweets will be displayed on the
originating from developing countries? map. This map is expected to reveal complex
RQ 2: Does information shared on social media locations patterns about social media user’s way of
meet quality indicators for user generated content reporting locations during disasters.
(UGC) to use in case of disasters in the Himalaya Tweet text with its metadata will be analysed to
region? identify potential attributes and entities to design
RQ 3: What is geographic feature granularity in an Entity Relationship Diagram (ERD). This
UGC during disasters? analysis will be performed on adequate subset of
RQ 4: What are the potential contributions of acquired data. Designed ERD will be used for
social media data in designing and developing VGI development of the database.
based disaster database for the Himalaya region?
4 Conclusion
3 Methods
This research project has great potential to be
Twitter offers free, real-time data (tweets) implemented by disaster response agencies within
downloads through its streaming API. This API the Himalayan region and also by international
requires certain parameters to capture tweets such organizations by adapting during and post-disaster
as particular keywords, tweets sent from particular framework. The Himalayan region is important in
users, or tweets originating from a particular determining world climate, prone to frequent
region. For this research, an application has been hazards, and is generally less accessible due to a
designed in R to capture real-time tweets without lack of facilities and resources. This research can
any pause, based on disaster-related keywords such be a good platform to minimize the impact of
as flood, earthquake, and landslide. Starting from disasters in this region. This research project
25.4.2015 more than 25 Million tweets based on enriches the Geographic Information Science
disaster-related keywords have been downloaded methods and concepts by focusing on unique
so far. The supervised Naïve Bayes topic model is spatio-temporal phenomena of natural disasters.
used to classify tweets in two classes:
“information” and “not information”. 5 References
The Twitter streaming API downloads tweets
with their metadata. In addition to the tweet text • Acar, Adam and Muraki, Y. (2011).
field, forty-one associated fields are also Twitter for crisis communication : lessons
downloaded. Many researchers have identified learned from Japan’ s tsunami disaster.
different sets of features to assess the credibility of 7(3), 392–402.
tweets. Features such as the “number of followers, • Adam, N. R., Shafiq, B., & Staffin, R.
retweet count, number of URL’s, verified, time, (2012). Spatial computing and social
user description, status count, [and] length of a media in the context of disaster
tweet” (Canini, Suh, & Pirolli, 2011; Morris et al., management. IEEE Intelligent
2012) will be used to assess the credibility of Systems, 27(6), 90-96.
tweets in the Himalayan region. • Bilham, R., Gaur, V., & Molnar, P. (2001).
A natural disaster event will be selected as a case Pollen Tubes Labeled With Green
study to analyse the granularity of geographic Fluorescent Protein Show That the Pollen
AGILE 2016 – Helsinki, June 14-17, 2016
Tube Also Pen- Etrates an, (August), ken2=exp=1450710999~acl=/static/pdf/9
1442–1444. 38/art%253A10.1023%252FA%253A102
• Canini, K. R., Suh, B., & Pirolli, P. L. 1
(2011). Finding Credible Information • Kavanaugh, A., & Yang, S. (2011).
Sources in Social Networks Based on Microblogging in crisis situations :, 1–6.
Content and Social Structure. 2011 IEEE • Kellenberg, D. K., & Mobarak, A. M.
Third Int’l Conference on Privacy, (2008). Does rising income increase or
Security, Risk and Trust and 2011 IEEE decrease damage risk from natural
Third Int’l Conference on Social disasters? Journal of Urban Economics,
Computing, 1–8. 63(3), 788–802.
http://doi.org/10.1109/PASSAT/SocialCo http://doi.org/10.1016/j.jue.2007.05.003
m.2011.91 • Kwak, H., Lee, C., Park, H., & Moon, S.
• Clay Whybark, D. (2007). Issues in (2010). What is Twitter , a Social Network
managing disaster relief inventories. or a News Media? The International World
International Journal of Production Wide Web Conference Committee
Economics, 108(1–2), 228–235. (IW3C2), 1–10.
http://doi.org/10.1016/j.ijpe.2006.12.012 http://doi.org/10.1145/1772690.1772751
• Cutter, Sussan L., (2012). Hazards, • Lavé, J., & Avouac, J. P. (2000). Active
vulnerability and environmental justice. folding of fluvial terraces across the
New York, NY: Taylor and Francis Siwaliks Hills, Himalayas of central
• Earle, P. S., Bowden, D. C., & Guy, M. Nepal. Journal of Geophysical Research,
(2011). Twitter earthquake detection: 105(B3), 5735.
Earthquake monitoring in a social world. http://doi.org/10.1029/1999JB900292
Annals of Geophysics, 54(6), 708–715. • Lingad, J., Karimi, S., & Yin, J. (2013).
http://doi.org/10.4401/ag-5364 Location extraction from disaster-related
• Gelernter, J., & Balaji, S. (2013). An microblogs. WWW 2013 Companion -
algorithm for local geoparsing of Proceedings of the 22nd International
microtext. GeoInformatica, 17(4), 635– Conference on World Wide Web, 1017–
667. http://doi.org/10.1007/s10707-012- 1020. Retrieved from
0173-8 http://www.scopus.com/inward/record.url
• Hossmann, T., Carta, P., Schatzmann, D., ?eid=2-s2.0-
Legendre, F., Gunningberg, P., & Rohner, 84892685783&partnerID=tZOtx3y1
C. (2011). Twitter in Disaster Mode: • Mendoza, M., Poblete, B., & Castillo, C.
http://doi.org/10.1145/2079360.2079367 (2010). Twitter Under Crisis : Can we trust
• Jongman, B., Wagemaker, J., Romero, B., what we RT ?, 71–79.
& de Perez, E. (2015). Early Flood • Middleton, S. E., Middleton, L., &
Detection for Rapid Humanitarian Modafferi, S. (2014). Real-time crisis
Response: Harnessing Near Real-Time mapping of natural disasters using social
Satellite and Twitter Signals. ISPRS media. IEEE Intelligent Systems, 29(2), 9–
International Journal of Geo-Information, 17. http://doi.org/10.1109/MIS.2013.126
4(4), 2246–2266. • Morris, M. R., Counts, S., Roseway, A.,
http://doi.org/10.3390/ijgi4042246 Hoff, A., & Schwarz, J. (2012). Tweeting
• Kattelmann, R. (2003). Glacial Lake is believing? Proceedings of the ACM
Outburst Floods in the Nepal Himalaya: A 2012 Conference on Computer Supported
Manageable Hazard? Retrieved December Cooperative Work - CSCW ’12, 441–450.
21, 2015, from http://doi.org/10.1145/2145204.2145274
http://download.springer.com/static/pdf/9 • Murthy, D. (2013). NEW MEDIA AND
38/art%3A10.1023%2FA%3A102113010 NATURAL DISASTERS: Blogs and the
1283.pdf?originUrl=http://link.springer.c 2004 Indian Ocean tsunami. Information
om/article/10.1023/A:1021130101283&to Communication and Society, 16(7), 1176–
AGILE 2016 – Helsinki, June 14-17, 2016
1192. development. IEEE Transactions on
http://doi.org/10.1080/1369118X.2011.61 Knowledge and Data Engineering, 25(4),
1815 919–931.
• Murthy, D., & Longwell, S. A. (2013). http://doi.org/10.1109/TKDE.2012.29
Twitter and Disasters. Information, • Singh, A. (2010). Climate Change and
Communication & Society, 16(6), 837– Disasters in the Hindu Kush Himalayan
855. Region. Article, New Delhi, India:
http://doi.org/10.1080/1369118X.2012.69 Solution Exchange. Retrieved from
6123 http://amritapeoplesvoice.blogspot.com
• Nasikhin, & Adriani, M. (2007). Location • Terpstra, T., Stronkman, R., Vries, a De,
Identification for the Geographic & Paradies, G. L. (2012). Towards a
information Retrieval. realtime Twitter analysis during crises for
• Petrović, S., Osborne, M., & Lavrenko, V. operational crisis management.
(2010). Streaming first story detection Proceedings of ISCRAM 2012, (April), 1–
with application to twitter. NAACL HLT 9.
2010 - Human Language Technologies: • Thayyen, R. J., & Gergan, J. T. (2010).
The 2010 Annual Conference of the North Role of glaciers in watershed hydrology: a
American Chapter of the Association for preliminary study of a “Himalayan
Computational Linguistics, Proceedings catchment.” The Cryosphere, 4(1), 115–
of the Main Conference, (June), 181–189. 128. http://doi.org/10.5194/tc-4-115-2010
Retrieved from • Verma, S., Vieweg, S., Corvey, W. J.,
http://www.scopus.com/inward/record.url Palen, L., Martin, J. H., Palmer, M., …
?eid=2-s2.0- Anderson, K. M. (2011). Natural
80053272732&partnerID=tZOtx3y1 Language Processing to the Rescue ?:
• Poblete, B., Castillo, C., & Mendoza, M. Extracting “ Situational Awareness ”
(2011). Information Credibility on Tweets During Mass Emergency.
Twitter, 45–59. • Vieweg, S., Hughes, A. L., Starbird, K., &
• Ratcliffe, J. H. (2004). Geocoding crime Palen, L. (2010). Microblogging during
and a first estimate of a minimum two natural hazards events. Proceedings of
acceptable hit rate. International Journal the 28th International Conference on
of Geographical Information Science, Human Factors in Computing Systems -
18(January), 61–72. CHI ’10, 1079.
http://doi.org/10.1080/136588103100015 http://doi.org/10.1145/1753326.1753486
96076 • Viviroli, D., & Weingartner, R. (2004).
• Richardson, S. D., & Reynolds, J. M. The hydrological significance of
(2000). An overview of glacial hazards in mountains: from regional to global scale.
the Himalayas. Quaternary International, Hydrology and Earth System Sciences,
65–66, 31–47. 8(6), 1017–1030.
http://doi.org/10.1016/S1040- http://doi.org/10.5194/hess-8-1017-2004
6182(99)00035-X • Witham, C. S. (2005). Volcanic disasters
• Sadler, J. C., & Ramage, C. S. (1976). The and incidents: A new database. Journal of
Role of Mountains in the South Asian Volcanology and Geothermal Research,
Monsoon Circulation. Journal of the 148(3–4), 191–233.
Atmospheric Sciences. http://doi.org/10.1016/j.jvolgeores.2005.0
http://doi.org/10.1175/1520- 4.017
0469(1976)033<2255:COROMI>2.0.CO; • Yamada, T., & Sharma, C. (1993). Glacier
2 lakes and outburst floods in the Nepal
• Sakaki, T., Okazaki, M., & Matsuo, Y. Himalaya. IAHS Publications-
(2013). Tweet analysis for real-time event Publications of …, (218), 319–330.
detection and earthquake reporting system Retrieved from
AGILE 2016 – Helsinki, June 14-17, 2016
http://ks360352.kimsufi.com/redbooks/a2
18/iahs_218_0319.pdf
• Imran, M., Castillo, C., Lucas, J., Meier,
P., et al. (2014) AIDR: Artificial
intelligence for disaster response. In:
Proceedings of the companion publication
of the 23rd international conference on
World wide web companion. [Online].
2014 pp. 159–162. Available from:
doi:10.1145/2567948.2577034.
• Kumar, S., Barbier, G., Ali Abbasi, M.A.
& Liu, H. (2011) TweetTracker: An
Analysis Tool for Humanitarian and
Disaster Relief. In: Fifth International
AAAI Conference on Weblogs and Social
Media. 2011 pp. 661–662.
• Thom, D., Kruger, R., Ertl, T., Bechstedt,
U., et al. (2015) Can twitter really save
your life? A case study of visual social
media analytics for situation awareness.
In: 2015 IEEE Pacific Visualization
Symposium (PacificVis). [Online]. 2015
pp. 183–190. Available from:
doi:10.1109/PACIFICVIS.2015.7156376.
• Abel, F., Hauff, C., Houben, G.-J.,
Stronkman, R., et al. (2012) Twitcident:
Fighting fire with information from social
web streams. In: Proceedings of the 21st
international conference companion on
World Wide Web - WWW ’12 Companion.
[Online]. 2012 p. 305. Available from:
doi:10.1145/2187980.2188035.