Commonsense reasoning has long been a core artificial intelligence (AI) problem. However, there has long been a lack of scalable commonsense acquisition methods and principle ways of applying the commonsense in the past. An important reason is that we do not have a good enough commonsense representation methodology. In this talk, I will first introduce why representing commonsense with higher-order selectional preference over eventualities (i.e., events and states) is a good option and how we can construct a large-scale eventuality-centric commonsense knowledge graph (KG) ASER at low cost. After that, as it is infeasible to build up a large enough knowledge graph to cover all events in the world, I will then introduce how we can generalize knowledge about observed events to unseen ones. In the end, I will introduce the proposed commonsense inference learning framework CKBQA, which enables us to explore a generalizable commonsense inference model.
Dr. Hongming Zhang is currently a research scholar at University of Pennsylvania, working with Prof. Dan Roth. He got his Ph.D. degree from HKUST in 2021. His research interest is primarily on commonsense reasoning and multi-modal event understanding. Specifically, he aims at building a large-scale eventuality-centric knowledge graph to connect different modalities and help machines understand the commonsense. More information is available at http://www.cse.ust.hk/~hzhangal/.
Host: Muhao Chen, POC: Pete Zamar
YOU ONLY NEED TO REGISTER ONCE TO ATTEND THE ENTIRE SERIES - We will send you email announcements with details of the upcoming speakers.
Register in advance for this webinar: https://usc.zoom.us/webinar/register/WN__0VhakI6Q6i3JsasdmNWcA
After registering, you will receive an email confirmation containing information about joining the Zoom webinar.
The recording for this AI Seminar talk will be posted on our USC/ISI YouTube page within 1-2 business days: https://www.youtube.com/user/USCISI.