Bio
Xiaodan Zhu is the director of Text Analytics and Machine Learning (TAML) lab at Queen’s University, Canada. He is a Mitchell Professor and the AI Lead of the Ingenuity Labs Research Institute at Queen’s, and an Assistant Professor at the Department of Electrical and Computer Engineering. He is a Faculty Affiliate at the Vector Institute for Artificial Intelligence. Xiaodan received his Ph.D. from the Department of Computer Science at the University of Toronto. His recent research interests include natural language inference, sentiment analysis, dialogue modelling, summarization, NLP for legal and financial text analytics, neural symbolic models for NLP, image captioning and grounding, and in general, artificial intelligence. He recently served on the COLING ‘20 and ACL ‘19 Best Paper Selection Committee. He served as a chair for the 33rd Canadian Conference on Artificial Intelligence, senior area chair for NAACL ’22, ACL ‘21, co-chair for SemEval ‘20 and ‘19, workshop chair for COLING ‘20, and various other service roles for his research community.
Keynote: Towards Robust Fact Verification Based on Text and Tables
Verifying if a statement is true based on the provided evidence is important for many legal and financial applications. Fact verification intrinsically needs to perform reasoning over the provided evidence, which is a central and challenging problem in NLP and AI that needs better models and further understanding. This talk will discuss two recent trends of research along the line. In the first, the evidence is not only collected from unstructured text but also from semi-structured and structured data such as tables. I will present verification models based on program execution and statement decomposition. In the second trend, efforts have been made to leverage the complementary strengths of neural networks and symbolic models, aiming to build logic into neural-network-based inference models and hence provide explainability towards rendering more trustable models.