A study analyzed 1,684 tweets from 705 users during the 2009 Black Saturday bushfires in Victoria, Australia. The tweets were categorized and 12% contained useful information like maps, updates from emergency services, and on-the-ground reports. 22% provided location data and 25% had valuable first-hand accounts. While Twitter was used to share information during the crisis, future research could explore better aggregating and validating information on the platform to support emergency communication.
More than Just Lines on a Map: Best Practices for U.S Bike Routes
Twitter Provides Valuable Insights During Crisis
1. Priceless Tweets!
A study on twitter messages
A study on twitter messages
posted during crisis: Black
Saturday.
d
Suku Sinnappan
Cathy Farrel
y
Elizabeth Stewart
Swinburne University of Technology
# Locusts
# Fire
# Fire
# Flood
2. Outline
• Background
• Objectives
• Motivation
• Significance
• Literature and Methodology
Literature and Methodology
• Analysis and Discussion
• Future research
3. Background
• Web 2.0 as a channel for communication
• 2009 3 billion tweets worldwide
2009 ‐ 3 billion tweets worldwide
• 9000 tweets a minute (O’Brien 2009)
• Lightweight (140 char), crowdsourcing
• Twitter has been used for numerous purposes
Twitter has been used for numerous purposes
– Politics, PR, marketing, education, Personal,
customer relationship, official, crisis etc
t l ti hi ffi i l i i t
4. Objectives
• Web 2.0 has not been used to effectively
g p p
communicate during crisis despite its potential
in Australia
• Can Twitter help and How?
Can Twitter help and How?
• What kind of information is tweeted during
crisis in Australia?
• Can Twitter be used an alternative channel for
Can Twitter be used an alternative channel for
communication during crisis?
5. Motivation
• There are few motivations leading up to the
y
study
– Web 2.0 used for official info dissemination but
not communication despite huge public uptake
not communication despite huge public uptake
– The failure of traditional communication channels
during crisis
during crisis
– Inadequate ground level information during crisis
– Valuable information untapped from Web 2.0
6. Significance
• An alternative communicate channel (both
y) g
ways) for large scale disaster
• Reporting valuable information effectively
(public and official communication)
(public and official communication)
• Potential reduction in casualties or damage
7. Literature
• T itt b t t diti
Twitter beats traditional media to deliver up to date
l di t d li t d t
information
• Iran election, Mumbai Terrorist, Bangkok Red Shirt protest,
, , g p ,
Gaza Flotilla. Including yesterday’s student hostage crisis at
Marinette, Wisconsin
• Sichuan earthquake (Gabarain 2008) Californian fires
Sichuan earthquake (Gabarain, 2008), Californian fires
(Wagner, 2008, Sutton Et al, 2008), New England Ice Storm
(Ragan, 2009), Gulf of Mexico Hurricane (Janega, 2008),
Cyclone Nagris i M
C l N i in Myanmar (Palatino, 2008) and Mississippi
(P l ti 2008) d Mi i i i
Hurricane (Brown, 2009), Red River Flood (Starbird et al.
2010), Mass emergency (Hughes et al. 2009), Haiti (Palen et
al. 2010), Seattle Violent Crisis (Herevin and Zach, 2010).
8. Methodology
• Gather tweets during crisis to analyse
• 2009 Victoria bush fires using ‘fires’ as used by
2009 Victoria bush fires using fires as used by
Herevin and Zach (2010)
• N d d f
Needed to form categories to make sense of
i k f
the tweets – used Namaan et al (2010) as
base
• 1684 tweets between 6nd – 14th Feb 2009
1684 tweets between 6 Feb, 2009
• 705 unique users
9. Analysis and Discussion (1/7)
Analysis and Discussion (1/7)
No. of
• Top 10 (as shown) No. User Name
No. of Followers (as
Tweets of 19 July
• Only 2 official
Only 2 official 2010))
1 774melbourne 63 13,845
channels 2 geehall1 49 655
3 retrogrrl 29 1,223
,
• E
Example of official 4
l f ffi i l SassyCupcakes 20 148
5 tellyworth 20 91
dissemination 6 duncanriley 19 8,525
,
7 problogger 19 100,859
• Not reported in 8 stephenedgar 18 1,331
royal commission 9
royal commission 10 wolfcat 18 2,585
3,996
abcnewsMelbrne 16
10. Analysis and Discussion (2/7)
Analysis and Discussion (2/7)
• Validating data collected against another crisis
( g
(Uganda pub bombing after world cup 2010)
p g p )
• Tweets were categorised by batches of 100s
based on Namaan et al while referring to
based on Namaan et al while referring to
Sutton et al. 2008, Hughes et al. 2009,
Heverin and Zach, 2010.
Heverin and Zach 2010
11. Analysis and Discussion (3/7)
• * Naaman et al. (# 6‐13)
• Opinions/Complaints and No. Categories of tweets No. of tweets
Statements/Random
1 Web Li k
W b Links to maps, websites, other resources,
bi h 400
Thoughts were merged as
2 Geographical data i.e. location, name of a place 730
they were inapplicable
• Each tweet was allowed to Other factual description i.e. Time, Figure, Cardinal directions, Other organisations
3 (formal names), types of crisis (flood, fire, tsunami etc), communication outlets 875
be coded into more than 1 (radio, TV, telephone), climate related information
category due to information
Suggestions, request followers to act, actionable information, and current status
• Ave 1.98 categories, 4
g , 4
update
264
iterations 5 Seeking information, help or answers 119
• 2 dropped (#12 ,#13), new 6* Statements or random thoughts – a comment by the user on a situation 331
(
(#1‐#5)) 7* Me
M now – current status about the user i.e. ‘tired and upset’
t t t b t th i ‘ti d d t’ 218
• Tweets that were less 8* Presence maintenance – instant display of user existence i.e. ‘I’m backkkk’ 3
informative (# 7,8,11) were 9* Other’s status – informing followers about other’s situation 2
miniscule as compared to
miniscule as compared to Communication with Followers – reply tweets, forwarding tweets, linking others in
10* 491
80% reported by Namaan et a message
al ‐ while 65% were 11* Self promotion i.e. ‘Check out my blog’ 1
informative 12
12* Anecdote me – an exciting-funny incident about the user 0
• Tweets need to be decoded 13* Anecdote others - an exciting-funny incident about others 0
in context
12. Analysis and Discussion (4/7)
Analysis and Discussion (4/7)
• Objective 2 – could Twitter be used during times of crisis as an alternative
Objective 2 could Twitter be used during times of crisis as an alternative
communication channel?
• 12% carried links ‐ Google maps, Country Fire Authority updates,
Weatherzone.com, Bureau of Meteorology, ABC.net.au,
Weatherzone com Bureau of Meteorology ABC net au
Incidentalert.com.au, TheAge.com.au and social sites such as Flickr and
Photobucket.
• 22% had Geo‐location data ‐ “Ah, too late. Yackandanda is under alert,
, ,
and the Beechworth fires are heading south‐east. 100+ homes gone, 14+
dead. #bushfires”.
• 25% had rich ground level information–
– “fires in narre north, two streets away from my house. shit.”
– “ADSL has been down all afternoon, Sky was surreal bronze colour, due to
fires”
– “t
“two routes out of Yea are closed and the other two lead towards other fires”
t t fY l d d th th t l d t d th fi ”
– “Fighting fires !”
13. Analysis and Discussion (5/7)
Analysis and Discussion (5/7)
• O
Operationalising the tweets using ‘#’ to increase search
ti li i th t t i ‘#’ t i h
ability. Eg Palen et al (2010) Haiti project (EPIC)‐
– #name (name),
#name (name),
– #loc (location) ,
– #num (amount or capacity),
– #contact (email, phone, link, other) or #con,
– #photo (link to photo),
– #source (source of info) or #src, #status (e.g., open, closed,
#source (source of info) or #src #status (e g open closed
injured, etc.),
– #date (date or time),
– #info (other information).
• Contextualise the # terms to suit disaster and region
14. Analysis and Discussion (6/7)
Analysis and Discussion (6/7)
• F
For eg –
– “Anyone have news on whether the fires are near Healesville?”
– “#help Anyone have news on whether the #crisis fires are near #loc
Healesville?”.
H l ill ?”
• 5% had direct actionable data ‐ “@royrussel ABC Gippsland
Transmitter burned by fires tune into 531AM , 828AM, 104.7FM,
105.5FM or 90.7FM for local update ‐ @774Melbourne”,
105 5FM 90 7FM f l l d t @774M lb ”
• 119 tweets were seeking for information/ assistance (similar to
phone calls to SES/CFA/Police)
– “I'm looking after an evacu‐cat from Warburton to help out
@firstdogonmoon 's friend. Bush fires affect pets too!”
– “Anyone have news on whether the fires are near Healesville?”
– ” Does anyone know if the fires touched Smith's Gully? Worried about
an ex‐colleague.”
15. Analysis and Discussion (7/7)
Analysis and Discussion (7/7)
– Tweets also prove that Twitter was used for crisis
communication during the bushfires by – public
and state authorities :
• “CFA urges those not directly affected by the Vic fires to
get info from ABC Melbourne http://bit.ly/MsGj
(expand) rather than the CFA site”
– A way forward will be to deploy bots to detect
f d ill b d l b d
searchable # tags, aggregate information with geo‐
location tag and channel info by using web 2.0
l i d h li f b i b20
itself . This offers an organic solution to users.
16. Future Research
Future Research
• Th
There are many ways to extend the current work
d h k
• Exploring ways to aggregate information (both direction)
• Ed
Educating both public and local authorities about Web 2.0
i b h bli d l l h ii b W b20
communication (device and software)
• Involving other web 2 0 tools – Facebook Myspace Orkut
Involving other web 2.0 tools Facebook, Myspace, Orkut
etc
• Applying the same study to flood, hurricane, terrorist,
earth quakes, or other natural disasters such locust
attacks.
• Validating the quality of information i e spam and info
Validating the quality of information i.e. spam and info
accuracy (eg. Flood info ABC invalid link)