PhD Candidate
Ohio State University
Areas of Interest
- Artificial Intelligence
- Human-Computer Interaction
- Natural Language Processing
Poster
On Building an Effective and Trustworthy Natural Language Interface: An intelligent agent that can communicate, collaborate, and learn
Abstract
Building Natural Language Interfaces (NLIs) that allow humans to acquire knowledge and complete tasks using natural language has been a long-term pursuit. Yet, existing NLIs are largely suffering from degraded performance when they are deployed to real applications, and meanwhile cannot inform users of their degradation as well as the likely hazards. Trying to address these problems, I argue that a practically effective and trustworthy NLI should be formulated as "an intelligent agent". The agent needs to be introspective of its uncertainties (i.e., sources of degradation) and can collaboratively interact with humans to resolve the uncertainties. It should also be able to learn from humans during the interaction. In my research, I focus on building NLIs to computer programs, aiming to eventually enable humans to command and communicate with computers via natural language. My work has demonstrated the promise of the envisioned intelligent NLIs in programming-related problems, and the findings potentially can also be applied to other tasks.
Bio
Ziyu Yao is a Ph.D. candidate at The Ohio State University (OSU), advised by Prof. Huan Sun. Her research interests lie in Artificial Intelligence (AI) and Natural Language Processing (NLP), particularly building effective and trustworthy natural language interfaces to assist humans with knowledge acquisition and problem solving. She has worked as a research intern at Carnegie Mellon University, Microsoft Research, Fujitsu Laboratories of America, and Tsinghua University. Before OSU, she obtained her B.E. degree in Communication Engineering from Beijing University of Posts and Telecommunications.
