Nothing but net? content and network factors in information diffusion

31/12/2015 - 12:00
Speaker: 
Seminar: 
מיקום: 
Abstract: 

Information diffusion is the process in which nuggets of information spread in a network (typically a social network). This is a complex process that depends on the network topology, the social structures and the information itself (content/language). In this talk I will discuss information diffusion from these different yet complementary perspectives. In the first part  of the talk I will focus on the features of the diffusing information.  I will  present a gradient boosted trees algorithm modified for learning user preferences of Twitter hashtags. In the second part of my talk I will focus on the network structure. I will use exponential random graph models (ERGM) in order to learn what latent factors contribute to network formation and I will show how network structure and social roles contribute to the information spread. Specifically, I will present promising results obtained on political networks in the American political systems and analysis of the partisan use of political hashtags.

In both parts I put emphasis on interpretable models that go beyond accurate prediction and  facilitate a better understanding of complex social processes.  

Bio sketch:

Oren is a postdoctoral fellow at the Harvard's School for Engineering and Applied Science (SEAS) jointly with the Network Science Institute at Northeastern University and is also affiliated with Harvard's Institute for Quantitative Social Science (IQSS). He received  his Ph.D. in Computer Science from the Hebrew University at 2013. In his work he combines natural language processing and network analysis in order to model and predict complex social dynamics. His work was published in both the NLP and the web/data science communities. Oren co-organized the workshop on NLP and Social Dynamics at ACL 2014, is co-organizing the EMNLP 2016 workshop on NLP and Computational Social Science  and will be giving the tutorial on Understanding Offline Political Systems by Mining Online Political Data at WSDM 2016. His research on sarcasm detection (with Dmitry Davidov and Ari Rappoport) was listed in Time Magazine special issue as one of the 50 best inventions of the year (2010).

    Homepage: http://people.seas.harvard.edu/~orentsur, Oren http://people.seas.harvard.edu/~orentsur/