Project Debater is the first AI system that can debate human experts on complex topics. Its live demonstration in February 2019 received massive media coverage. This research effort has resulted in more than 50 scientific papers, and many datasets freely available for research purposes. I will discuss the scientific challenges that arise when building such a system, including argument mining, stance classification, principled argument detection, narrative generation, and rebutting a human opponent. Many of the underlying capabilities of Project Debater are now available for academic research as web APIs. I will also present a systematic evaluation of Project Debater’s performance, published last year in the Nature magazine. Finally, I will introduce Key Point Analysis (KPA), a novel summarization technology coming out of Project Debater. KPA identifies the main points and their prevalence in a large collection of comments, and generates a concise summary that is human-readable, quantitative and actionable.
Roy Bar-Haim is a senior research scientist and project lead in the Debating Technologies group, IBM Research - Haifa. Roy joined Project Debater in its early stages, and was responsible for several of its core components. He is currently leading a research team working on quantitative summarization of opinions, as well as on leveraging AI and language technologies for cloud compliance. Roy holds a Ph.D in Computer Science from Bar-Ilan University. Before joining IBM, Roy led NLP teams in several startup companies. Together with his colleagues, he presented tutorials on debating technologies in AACL 2020, ACL 2021 and IJCAI 2021.