Lunch at 12:30pm, talk at 1pm, in 148 Fitzpatrick

Title: Towards human-AI collaboration on natural language based user tasks

Abstract: In this talk, I will introduce my recent work in designing human-AI collaborative systems to facilitate users’ daily natural language based tasks. First, I will talk about StoryBuddy, a human-AI interactive storytelling system allowing flexible parent involvement and chat-based question answering. Also I will briefly talk about our ACL work about the FairytaleQA dataset and storybook-based QA generation models behind StoryBuddy. Besides, I will introduce our latest work PaTAT that features human-AI collaborative qualitative coding with explainable rule synthesis, and Label Sleuth, an open-sourced human-in-the-loop platform for annotating text classification data. Lastly, I will discuss an ongoing GPT-3 based human-AI co-writing project and wish to seek feedback and ideas from the audience.

Bio: Zheng Zhang is a first-year PhD student in Computer Science and Engineering at University of Notre Dame, advised by Dr. Toby Jia-jun Li. Zheng’s work falls in the intersection of human-computer interaction, natural language processing, program synthesis, and speech processing, where he uses human-centered methods to design, build, and evaluate human-AI collaborative systems to strengthen the efficiency, trustworthiness, explainability, and accuracy of different phases in the AI model development and deployment, ranging from data annotation to model’s decision making. Besides, he is also passionate about leveraging the state-of-the-art AI techniques to benefit a variety of human needs and optimizing user experience accordingly.