People/Web Search Calendar Emergency Info A-Z Index UVA Email University of Virginia

Computer Science Colloquia


Hongning Wang, University of Illinois, Champaign-Urbana

Tuesday, March 18, 2014

3:30 PM, Rice Hall, Rm. 130 (Light refreshments after the seminar Rice Hall 4th floor atrium)

HOST: Jane Qi

Beyond experts and engineering: exploiting data for automated control

ABSTRACT

Mining actionable knowledge from big data provides remarkable opportunities for optimizing operations and decision making in many application domains. Since a large portion of data is produced by humans, and, more importantly, they are the ultimate consumer of knowledge, putting humans into the loop of mining big data is essential to maximize the utility of the mined knowledge. However, most existing work on big data mining has taken a data-centric perspective and emphasized efficiency and scalability of the computational infrastructures and algorithms for mining large data sets, but without much consideration of the human factors.

In this talk, I will argue for the importance of "human-centric big data mining," where the dual role of humans as both data producers and knowledge consumers must be explicitly considered; and I will present my work on mining human-generated data for knowledge discovery, as well as exploiting the mined knowledge about people to optimize knowledge services for human consumers. Specifically, I will first introduce a new data mining problem called Latent Aspect Rating Analysis (LARA) for analyzing opinionated text data to discover individual users' latent opinions and preferences at the level of topical aspects. To solve the LARA problem, a unified generative model is proposed, and it proves to be effective and enables a wide range of novel applications that bring benefit to both ordinary consumers and business intelligence analysts. I will then discuss how to optimize a knowledge service system (e.g., a search engine system) via mining users' interactive behaviors recorded in the system's log data. By mining such log data, we can automatically organize the scattered long-term interaction behaviors of users into semantically coherent user-tasks, which can then be further exploited to explicitly model individual users' information need and decision preferences and optimize the service of a big data application system in a personalized manner.

Bio: Hongning Wang is a Ph.D. candidate from the Department of Computer Science at University of Illinois at Champaign-Urbana, supervised by Professor ChengXiang Zhai. His research interests include data mining, information retrieval, and machine learning, with a particular focus on computational user modeling and knowledge discovery. He has published over 20 research papers on these topics in top venues in data mining and information retrieval areas, including KDD, WWW, SIGIR and WSDM. He is the recipient of the 2012 Google PhD Fellowship in Search and Information Retrieval, and 2012 Yahoo! Key Scientific Challenges in Web Information Management. He has served on program committees for several major conferences such as ICML, ECML/PKDD, and ECIR, and reviewed for multiple journals, including IEEE TKDE, ACM TOIS, Neurocomputing and BMC Bioinformatics.