Transcript Online Social Networks (OSNs)

Role of Online Social Networks during
disasters & political movements
Saptarshi Ghosh
Department of Computer Science and Technology
Bengal Engineering and Science University Shibpur
Online Social Networks (OSNs)

Among the most popular sites in today’s Web




More than few billion users world-wide
Celebrities, media houses, politicians all using OSNs
Quick ways of disseminating information, real-time news
Huge data readily available


Plethora of user-generated content: text, images, videos, …
Automated means of collecting data rather than surveys
(on which traditional social media research had to depend)
Variety in online social media
Multi-disciplinary research on OSNs

Tools from a wide variety of disciplines used to
study OSNs

Sociology – how human beings behave in society

Computer networks & distributed systems

Network science, complex network theory

Data mining, machine learning, information retrieval,
natural language processing, …
Mining information on recent events

Facebook, Twitter are valuable sources of news on
events happening ‘now’ [Yardi, ICWSM 2010]

Natural calamities, e.g., hurricanes, floods, earthquakes
[Sakaki, WWW 2010][Qu, CSCW 2011]

Man-made calamities, e.g., bomb blasts, riots

Spread of epidemics, e.g., dengue [Gomide, WebSci 2011]

Elections, political unrests
OSNs after calamities
OSNs after calamities
Activity in Twitter after earthquake
No longer only a comic strip, but close to reality
Sakaki et. al., “Earthquake shakes Twitter users: real-time event
detection by social sensors”, WWW 2010
Profile of a Twitter user
Example tweets
Use of OSNs during & after disasters

Qu et al. Microblogging after a Major Disaster in China: A
Case Study of the 2010 Yushu Earthquake. CSCW 2011




Analyed citizens’ activity in such scenario
How information spreads
How microblogging facilitated disaster response
Muralidharan, Rasmussen. Hope for Haiti: An analysis of
Facebook and Twitter usage during the earthquake relief
efforts. Public Relations Review (Science Direct)

Analyzed tweets posted by media & non-profit organizations
Types of posts / tweets

Different types of tweets posted during & after
disasters




Situational Updates
Opinion and sentimental tweets
Help Tweets
Event Analysis
Types of tweets
Situational update
Types of tweets
Help Tweet
Types of tweets
Opinion and
Sentiment
Types of tweets
Event analysis
Utilizing information in OSNs


During an important event, posts generated in OSNs
at the rate of hundreds to thousands per second
Several well-known challenges / research issues







Extracting important information
Summarization of data
Authority / expert identification
Public sentiment / opinion mining
Spam detection
Dealing with misinformation, rumours
… and many others
Extracting important information

Important information during a calamity



Situational updates (SU)
How to identify SU posts from among all posts?
Use of NLP and ML techniques



[Vieweg, ICWSM 2011]
NLP to identify objectivity, formal / informal register,
personal / impersonal tone of tweets
Trained ML classifier based on these features
85% - 90% accuracy in SU / non-SU classification
Summarization of data

Tweets posted too fast for human comprehension


Ways to organize data: extract important posts, automatic
summarization, …
Summarization of sets of tweets on a common topic
[Sharifi, HLT-NAACL 2010][Inouye, SocialCom, 2011]

Continuous summarization of tweet streams
2013]
[Shou, SIGIR
Identify influential users / experts

Several metrics of influence




#followers, PageRank, #times retweeted in Twitter, …
Topical experts

[Cha, ICWSM 2010]
[Weng, WSDM 2010] [Pal, WSDM 2011] [Ghosh, SIGIR 2012]
Authoritative sources of information on specific topics
How to measure topic-specific expertise of users?
Experts during specific events


Community leaders during emergencies [Tyshchuk, ASONAM 2013]
Geographically ‘local’ sources [Yardi, ICWSM 2007]
Emotion / opinion mining

Identify user’s emotion / opinion from posts in OSN
[Bollen, WWW 2010]

Identify opinion on movies / political issues [Fang, WSDM 2012]

Summarization of opinions [Ganesan, WWW 2012]

Twitter used to predict success of movies, election results
[Tumasjan, ICWSM 2010]

Extension: estimate sentiment of a country / whole world
on issues of national / international importance
Spam detection

Identify spam / users with malicious intentions



Identify spam in Facebook [Gao, IMC 2010] ,Twitter [Lee, SIGIR 2010],
Youtube [Benevenuto, SIGIR 2009], blogs [Shin, Infocom 2011], …
Identify spam in tweets related to trending topics / events
happening now [Benevenuto, CEAS 2010][Martinez-Romo, Expert Systems 2013]
Sybil detection [Yu, SIGCOMM 2006][Viswanath, SIGCOMM 2010]
Dealing with rumor / misinformation

Rumors frequently posted, often unintentionally

Detecting rumors in tweets



[Gupta, PSOSM 2012][Castillo, WWW 2011]
Classify credible / non-credible, rank tweets wrt credibility
Features: text-based (swear words, emoticons), userbased (#followers, retweeting behavior), how propagated
Spread of rumors in social networks


Why rumors spread quickly [Doerr, ACM Communications, 2012]
How to control rumors [Tripathy, CIKM 2010]
Rumors in Twitter after London riots
http://www.guardian.co.uk/uk/interactive/2011/dec/07/london-riots-twitter
Thank You
Contact: [email protected]