Zhifeng Wang, PhD

About Me

I am currently pursuing a PhD in South China University of Technology, and I am also serving as a visiting scholar in Language and Information Technologies at Carnegie Mellon University. My research interests include digital speech forensic, robust speaker recognition, spoken dialogue systems, human-robot interaction, and natural language generation.

My advisors are Prof. Alexander I. Rudnicky and Prof. Qianhua He.

I’m interested in improving situated, language-based communication with robots by investigating how people convey spatial information, both in reference to the environment and to members of a human-robot team. This work is in the treasure hunting domain as part of the TeamTalk project.

My work has led to better understanding of how people express route instructions to robots, highlighting the importance of landmarks in such instructions. The TeamTalk corpus contains the data we used for this research. In my current work I am developing spoken language interfaces to robots that combine route instructions from humans with knowledge about a robot’s surroundings. The goal of this work is to enable robots to recover from communication problems more effectively. For example, robots could ask questions about ambiguous referents in the environment if they are aware that multiple landmarks of the same type exist in the world (like when a person asks a robot to move to a door, but multiple doors are nearby). I am also interested in how crowdsourcing marketplaces can impact natural language research. I’m also the chief organizer for the Dialogs on Dialogs Reading Group and a staff reporter for the IEEE Speech and Language Processing Technical Committee’s Newsletter. My research work on occasion has contributed to Olympus, an open-source spoken dialogue system framework that originated at CMU.

In 2009, I received an MSc in Artificial Intelligence from the University of Edinburgh, specializing in natural language and language engineering. My MSc Thesis focused on evaluating a natural language generation system that could generate comparisons between music pieces. I worked with Prof. Johanna Moore and Amy Isard under an Edinburgh-Stanford Link studentship. Prior to this, I graduated with a BSc in Computer Science and Applied Mathematics & Statistics from Stony Brook University. There I worked on research in spoken dialogue systems for surveys.