Case Studies Web Science (VU) (707.000) Elisabeth Lex KTI, TU Graz June 6, 2016 Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 1 / 24
Case StudiesWeb Science (VU) (707.000)
Elisabeth Lex
KTI, TU Graz
June 6, 2016
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 1 / 24
Study Structural Features of Online Social Networks
Study of linking patterns and discussion topics of political bloggers1
Goal: measure degree of interaction between liberal and conservativeblogs, uncover differences in structure of the two communities
2004 first presidential elections where blogs played a role, precedingelections, increasing readership
Data set: large set of political blogs ( 1500, balanced set) over theperiod of two months preceding the U.S. Presidential Election of 2004
Community detection (Girvan-Newman method)
1The political blogosphere and the 2004 US election: divided they blog, Adamic,Lada A., and Glance N. , Proceedings of the 3rd international workshop on Linkdiscovery, p.36-43, (2005)
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 2 / 24
Liberal and conservative blogs
Figure: Community structure of the political blogs. Red for conservative, blue forliberal. Orange links go from liberal to conservative, purple from conservative toliberal
Divided blogosphere: liberals and conservatives link primarily within theircommunities (91%), far fewer cross-links
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 3 / 24
Findings
Stronger interaction patterns within conservative community(in-between, but also links outside of the community) but fewer posts
Differences in links to mainstream media (e.g., more conservativeblogs linked to Fox News)
Both communities focus on different news, topics, political figures
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 4 / 24
Online Conversational Practices of Political Parties onTwitter
Study of online conversational practices of German politicians onTwitter during the German federal election 20132
Investigated homophily in follow, retweet and mention Twitternetworks of politicians
Homophily: the tendency of a political party to communicate within partyboundaries
2Lietz, Haiko, Claudia Wagner, Arnim Bleier, and Markus Strohmaier. 2014. ”Whenpoliticians talk: assessing online conversational practices of political parties on Twitter(Best Paper Award).” In Proceedings of the Eighth International AAAI Conference onWeblogs and Social Media (ICWSM2014)
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 5 / 24
Online Conversational Practices of Political Parties onTwitter
Study of online conversational practices of German politicians onTwitter during the German federal election 20132
Investigated homophily in follow, retweet and mention Twitternetworks of politicians
Homophily: the tendency of a political party to communicate within partyboundaries
2Lietz, Haiko, Claudia Wagner, Arnim Bleier, and Markus Strohmaier. 2014. ”Whenpoliticians talk: assessing online conversational practices of political parties on Twitter(Best Paper Award).” In Proceedings of the Eighth International AAAI Conference onWeblogs and Social Media (ICWSM2014)
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 5 / 24
Online Conversational Practices of Political Parties onTwitter
Figure: Following, retweeting, and mentioning networks, colors = party
Structural differences: e.g., Homophily effects stronger in follow andretweet networks, weaker in mention network
Two parties (CDU/CSU and FDP) tightly knit in the follow andretweet networks, which formed previous coalition
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 6 / 24
Online Conversational Practices of Political Parties onTwitter
Figure: Using sociological constructs to assess political online conversations
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 7 / 24
Analysing Researchers on Social media
Identification and classification of scientists on Twitter3
Goal: generate Twitter user directory for e.g. computer scientists
Why?
Twitter, due to the speed of retweets, is a good medium forinformation diffusion from which scientists can benefit4
3Asmelash Teka Hadgu and Robert Jaeschke. Identifying and Analyzing Researcherson Twitter. In Proceedings of the 6th Annual ACM Web Science Conferencem 23-30,2014.
4H. Kwak, C. Lee, H. Park, and S. Moon. What is Twitter, a social network or anews media? In Proc. Int. Conf. on World Wide Web, pages 591-600, 2010
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 8 / 24
Analysing Researchers on Social media
Identification and classification of scientists on Twitter3
Goal: generate Twitter user directory for e.g. computer scientists
Why?
Twitter, due to the speed of retweets, is a good medium forinformation diffusion from which scientists can benefit4
3Asmelash Teka Hadgu and Robert Jaeschke. Identifying and Analyzing Researcherson Twitter. In Proceedings of the 6th Annual ACM Web Science Conferencem 23-30,2014.
4H. Kwak, C. Lee, H. Park, and S. Moon. What is Twitter, a social network or anews media? In Proc. Int. Conf. on World Wide Web, pages 591-600, 2010
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 8 / 24
Analysing Researchers on Social media
Goal of the study by Hadgu and Jaeschke:
Insights into how researchers use TwitterSupport users to find researchers on Twitter who work in specific fieldsInvestigate visibility of scientific articles in the social web: altmetrics
“Scientists”: users being interested in the topics of the targetdiscipline and having similar, Twitter- based features like users thathave published scientific papers
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 9 / 24
Analysing Researchers on Social media
What kind of computer scientists use Twitter - Empirical analysis ofidentified users: age, popularity, influence, and social network
E.g. Who are the most influential researchers on Twitter?
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 10 / 24
Selected Results
Figure: Researchers ranked by number of researchers who retweet (r), follow (f)and mention (m) them, and number of publications (p) in DBLP
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 11 / 24
Information Diffusion in Social Media
Information and behaviors propagate over online social networks dueto social exposure (i.e., number of friends of a user that have exposedinformation to her)
Recent study: propagation of hashtags and urls (“meme”) consideringtopical alignment5
A user diffuses or adopts a meme if she writes it in one of her tweets
Consider the topical distributions, i.e., the topicality, of active usersand popular memes
5Przemyslaw Grabowicz, Niloy Ganguly, and Krishna Gummadi. 2015. MicroscopicDescription and Prediction of Information Diffusion in Social Media: Quantifying theImpact of Topical Interests. In Proceedings of the 24th International Conference onWorld Wide Web Companion (WWW ’15 Companion)
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 12 / 24
Information Diffusion in Social Media
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 13 / 24
Results
Measure adoption probability as a function of social exposure k andtopical alignment S(user ,meme)
Findings: probability of adoption increases with social exposure, aswell as with the topical alignment between user and meme
Max of adoption probability slightly larger for social exposures thanfor topical alignments
Adoption probability grows much more steadily with topical similarity,reaches plateau only few social exposures
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 14 / 24
Information Diffusion in Networks of Researchers
Mapping of activities of researchers in academics to activity on socialnetworks6
Link authors of research papers to their twitter activity
Analysis of information flow between different users and/or researchareas
6Pujari, S., Hadgu, A., Lex, E., Jaschke, R. (2015) Social Activity versus AcademicActivity: A Case Study of Computer Scientists on Twitter In Proceedings of the 15thInternational Conference on Knowledge Technologies and Data-Driven Business(i-KNOW 2015)
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 15 / 24
Dynamics in Online Collaboration Networks
How do individual agents produce complex behaviour without centralcontrol?7
Goal: Study impact of social status on opinion diffusion incollaboration graphs to configure the degree of influence of socialstatus on network dynamicsEmpirical networks with heterogeneous community structure (Reddit,Wikipedia editions)Synthetic networks as benchmarking
7Hasani-Mavriqi, I., Geigl, F., Pujari, S., Lex, E., Helic, D. (2015) Influence of Statuson Consensus Building in Collaboration Networks. In Proceedings of the IEEE/ACMInternational Conference on Advances in Social Networks Analysis and Mining(ASONAM 2015)
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 16 / 24
A Web Science Approach Towards Recommender Systems- Example
Recommender systems: approach for Web users to cope withinformation overload
Collaborative Filtering (CF): one of the most successful methods
CF aims to recommend resources to user based on digital traces sheleaves behind on the Web
I.e., her interactions with resources and the interactions of othersimilar users
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 17 / 24
Recommender Systems - Example
Example: personalized resource recommendation - user-basedCollaborative Filtering (CFU)
CFU : resulting resources not ranked
Plus: neglects user-resource dynamics that shape attention andinterpretation
E.g. One single ressource can cause a shift in a user’s attention!
Idea: Combine user-based CF with a model from cognitive psychologythat captures such dynamics8
Goal: better personalize recommendations (rank) and improveaccuracy
8Seitlinger, P., Kowald, D., Kopeinik, S., Hasani-Mavriqi, I., Ley, T., and Lex, E.(2015). Attention Please! A Hybrid Resource Recommender MimickingAttention-Interpretation Dynamics. In Proceedings of 24th International World WideWeb Conference, WWW’15
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 18 / 24
Results
Evaluation: Social tagging systems (Bibsonomy, Delicious, CiteULike)
Compared approach to several baselines (e.g., Most Popular (MP)User-Based Collaborative Filtering (CFU))
We found: our approach improves CFU on all three datasets
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 19 / 24
Learning from Human Cognition
How do activation processes in human memory influence whether users inonline systems re-use information?
Goal: Design a decision support algorithm that mimics how peopleaccess information in their long term memory
Exploit cognitive architecture ACT-R from human memory theory topredict reuse probability
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 20 / 24
Learning from Human Cognition
Results
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 21 / 24
Example Theses
Social Learning Analytics
Social learning analytics make use of data generated by learners’online activity in order to identify behaviors and patterns within thelearning environment.
Goal of this thesis: collect data about learners and their contextsfrom the Social Web (e.g. StackExchange, Reddit) and to analyzethis data using Social Network Analysis (e.g. using graph-tool)
Focus is on studying direct interactions (e.g. messaging, friending,following) and/or indirect interactions, which happen when learnersleave behind activity traces such as e.g. ratings or tags.
Work will be in the context of the H2020 project AFEL (Analytics forEveryday Learning)9
9http://afel.know-center.tugraz.atElisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 22 / 24
Example Theses
Opinion Dynamics in Social Networks
The aim of this thesis is to build on our existing work, in which westudied the impact of social status on Opinion Dynamics andconsensus building in Online collaboration networks. Seehttp://www.know-center.tugraz.at/cms/wp-content/
uploads/2015/08/ASONAM_2015_Paper.pdf for details.
Goal: extend our existing analysis to other types of networks.
Our open source framework for Opinion Dynamics can be used(Python, graph-tool) and extended.
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 23 / 24
Example Theses
Implementing a Recommender System Using a Cognitive Model on theWeb
Goal: implement a computational cognitive model that explainsbehavior of users on the Web and to deploy this model forrecommender systems.
Evaluate model on well-known datasets from the research communityand compare it to other state-of-the art approaches.
Algorithm should be implemented in Java or Python
Elisabeth Lex (KTI, TU Graz) Case Studies June 6, 2016 24 / 24