Skip Navigation
Saturday, January 20, 2018 RSS Feed
Volgenau School of Engineering
The B.S. in Information Technology degree program is accredited by the Computing Accreditation Commission of ABET,

George Mason University is designated as a National Center of Academic Excellence in Information Assurance/Cyber Defense Research and Education.

Mining Unstructured Social Data for Behavioral Computing and User Modeling

Speaker: Hemant Purohit, PhD Candidate, Wright State University, Ohio, USA

Wednesday, May 6, 2015 – 11am – 12pm – Engineering Building, Room 2901


Social Media or Web 2.0, one source of Big Data, has completely revolutionized information consumption, management, and processing. The opportunity to understand and exploit such data has given rise to the interdisciplinary field of computational social science, which studies an unprecedented level of human interaction data—accessible due to an increasing shift from face-to-face to online communication. The big data challenges of large scale volume, velocity of content generation, sparsity of data behaviors, variety in language complexity and community demographics in this online interaction data present an exciting venue for computational science. For instance, mining social media data may help disaster response organizations better coordinate with citizens, and assist to reduce mounting cost of response, estimated to be 271 billion dollars annually by 2025.

I present a novel Web information processing framework, Identify-Match-Engage (IME), to interpret, manage, and integrate unstructured social media data generated by users (citizen sensors) for addressing cooperation between the citizens and formal organizations, with a disaster response use-case. My new behavioral computing methods model latent and subjective attributes (intent, influence, engagement) of users and communities in online social networks. They fuse knowledge from the Web resources (e.g., Wikipedia, Linked Open Data) and theories of behavior (e.g., social identity) into statistical methods of text mining and machine learning. Unlike traditional behavioral computing restricted to one of the three fundamental dimensions of social networks—user, content, and network, these techniques combine all three dimensions to improve the representation of subjective context, and compensate for sparsity of features to model latent behaviors (e.g., engagement). In the future, this interdisciplinary research can help incorporate human behavioral aspects for designing intelligent cooperative systems, and also contribute to physical-cyber-social computing. This work will impact problems of social good and large-scale online user/group modeling ranging from personalization (individual behavior) to abstraction (group behavior).


Hemant Purohit is an interdisciplinary, computational social science researcher at the Ohio Center of Excellence in Knowledge-enabled Computing (Kno.e.sis), and a PhD candidate working with Prof. Amit Sheth at Wright State University, USA. He was one of the ITU Young Innovator 2014 for UN’s ICT agency, for winning a global challenge on Open Source Technologies for Disaster Management, as well as one of the eight international fellows of USAID, Google and ICT4Peace foundation at an influential humanitarian technology event CrisisMappers ICCM-2013, UN Nairobi. He has jointly presented tutorials on crisis computing at prestigious conferences, AAAI ICWSM-2013 and SIAM SDM-2014, and also served as reviewer for conferences and journals including HICSS, ICWSM, WWW, JCSCW, ACM TOIT, ACM TIST, etc. More about Hemant:

Point of contact: Massimiliano Albanese (