Hasan T Karaoglu Epidemics in Blogspace Introduction Blogs are different! - PowerPoint Presentation

Hasan T Karaoglu Epidemics in Blogspace

IntroductionBlogs are different! Methods are different! Contents are different!Some methods on Some Content of Some BlogsDiscussion Outline

Blogs are a popular way to share personal journals, discuss matters of public opinion, have collaborative conversations,aggregate content on similar topics. Blogs also disseminate new content novel ideas How does content spread across, what kinds of content spreads, and at what rate? Introduction

Epidemics : one way of modeling these aspects Physics of Information Diffusion Disease Propagation ModelSusceptibleInfectedRecoveredMutation?Threshold Model for Social Networks Introduction - Epidemics

Youtube, Flickr (Content Sharing )AmazonCNN, MSNBC (Web)Linkedln (Professional Networking)Orkut , Facebook, Yonja (Social Networking) Twitter (?) Blogger, Blogspot , LiveJournal, Slashdot (Blogspace) Blogs are different

Blogs are different High level of reciprocity Symmetric indegree – outdegree In contrast to Web (high authority sites)

Blogs are different

Blogs are different Average Path Length is very short in compared to Web. (Directionality ?)

Blogs are different Joint Degree Distribution (High Degree Nodes Connect to Other High Degree Nodes) Epidemics on Network Core? Youtube Celebrities?

Blogs are different Strongly Connected Core Analysis Slowly Increasing Shortest Path High Clustering

Blogs are different Strong Local Clustering (people tend to be introduced to other people via mutual friends)

EpidemicsGossip Influence Map (Word of Mouth) Recommendation Based Web (Data) MiningMathematical Modeling (Markov Chains, Information Theory, …)… Methods are different

Contents are different Recommendation News (Political, Fun, Paparazzi)GossipMedia (Music, News, Excerpts)

Infection Inference technique introduced by Adamic et al.Link inferenceLink classificationClassifier training Problems and Challenges Some methods on Some Content of Some Blogs

Some methods on Some Content of Some Blogs Pattern Used for Classifier Training The number of common blogs explicitly linked to by both blogs (indicating whether two blogs are in the same community)The number of non-blog links (i.e. URLs) shared by the two Text similarity Order and frequency of repeated infections. Specifically, the number of times one blog mentions a URL before the other and the number of times They both mention the URL on the same day. In-link and out-link counts for the two blogs

Some methods on Some Content of Some Blogs Text Similarity s(A,B) = nAB / √n A / √nB

Some methods on Some Content of Some Blogs Timing of Infection

Some methods on Some Content of Some Blogs Link Inference Blog URL and Text Similarity Patterns Three-way Classifier (57%) reciprocated links, one way links, unlinked pairs Two-way Classifier (SVM 91.2% Logistic Regression 91.9%) linked unlinked pairs Infection Inference n A-before-B /n A, n A-after-B /n A, n A-same-day-B /n A Timing Patterns (75%) with all 6 timing patterns and text/blog similarity patterns (61 – 75%) link-in / link-out counts

Some methods on Some Content of Some Blogs Visualization Heuristics using classifiersTwo types of graph Directed Acyclic Graph Most likely tree

Some methods on Some Content of Some Blogs Epidemic Propagation Model by Gruhl et al. Topics IndividualsTopicsTopic = Chatter + Spike + (Resonance)

Some methods on Some Content of Some Blogs Epidemic Propagation Model by Gruhl et al. Topics IndividualsTopicsTopic = Chatter + Spike + (Resonance)

Some methods on Some Content of Some Blogs

Some methods on Some Content of Some Blogs aoccdrnig to rscheearch at an elingsh uinervtisy it deosn’t mttaer in waht oredr the ltteers in a wrod are, the olny iprmoetnt tihng is taht the frist and lsat ltteer is at the rghit pclae

Some methods on Some Content of Some Blogs Power-law Characteristic for Individuals Different Posting Behaviors for Individuals

Some methods on Some Content of Some Blogs Propagation Model Cascading ModelCopy Probability κ (v,w) Noticing Probability r( v,w )For 7K topics, r mean 0.28 and std 0.22,κ quite low, mean 0.04 and std 0.07, Even bloggers who commonly read from another source are selective in the topics they choose to write about.

Could we use these models to extract further pattern or characteristics ? Classification of Hoax, Fake News ? Prediction of Popular songs, videos at their inception…..Discussion

Thanks! Q & A

D. W. Drezner , and H. Farrell, “Web of Influence,” Foreign Policy, vol. 145, pp. 32-40, Dec. 2004E. Adar and L. A. Adamic, “Tracking Information Epidemics in Blogspace,” Proceedings of the 2005 IEEE/WIC/ACM International Conference on Web Intelligence, pp. 207–214, 2005. D. Gruhl, R. Guha, D. Liben-Nowell, and A. Tomkins, “Information diffusion through blogspace,” Proceedings of the 13th international conference on World Wide Web, pp. 491-501,2004. A. Mislove, M. Marcon , K. P. Gummadi, P. Druschel , and B. Bhattacharjee , “Measurement and Analysis of Online Social Networks,” Proceedings of the 7th ACM SIGCOMM conference on Internet measurement, pp. 29-42, 2007 M. Cha, J. A. N. Perez, and H. Haddadi , "Flash Floods and Ripples: The Spread of Media Content through the Blogosphere", 3rd Int'l AAAI Conference on Weblogs and Social Media (ICWSM) Data Challenge Workshop, May 17 - 20, 2009, San Jose, CaliforniaM . Young, The Technical Writer's Handbook. Mill Valley, CA: University Science, 1989. Z. Fanzi , Q. Zhengding , L. Dongsheng , and Y. Jianhai , “Shape-based time series similarity measure and pattern discovery algorithm”, Journal of Electronics, vol. 22, pp. 142-148, Aug. 2007 References