Title:

Data Mining for General Task Resolution: Methodology and Analysis

Abstract:

As the advent and prosperity of various problem solving and question answering systems such as crowdsoucing systems,Google now, Apple Siri and so on, automated problem solving and question answering seem to receive more and more attention, and embrace a bright future. In this talk, I will talk about our work along this line. The presentation is organized in two parts. In the first part, I will talk about task resolution in human-human collaborative environments, where human experts are organized into an expert network to resolve problems from customers. We aim at analyzing the expert behaviors in such collaborative environments and ultimately optimizing the system in terms of task resolution efficiency. In the second part, I will talk about question answering in human-computer interactive environments, where we develop automated algorithms to answer general questions such as "Where ShanghaiTech Univ. is located?" and "who is the dean of SIST in ShanghaiTech?". I will present two question answering frameworks, namely schemaless graph querying and natural language question answering. Though from different perspectives, both frameworks utilize huge knowledge bases such as freebase to answer questions and satisfy people's information need. Finally, we conclude this talk with a brief summary and potential future works. 

Bio:

Huan Sun is currently starting her  fifth year as a Ph.D. student in the Department of Computer Science, University of California, Santa Barbara. She has been working with Prof. Xifeng Yan since Fall 2010. Before coming to UCSB, Huan received her B.S. in EE from the University of Science and Technology of China (USTC) in 2010.  She won the Regents' Special Fellowship to support her graduate study in UCSB.  In the past, she worked as an intern at Microsoft Research Asia, IBM T.J. Watson Research Center, and Microsoft Research, Redmond respectively.