Intelligence is a human's ability to acquire and apply knowledge. How can machines explore world knowledge towards artificial intelligence? Knowledge Graph (KG) is one of the most effective ways to organize knowledge in a machine-readable form. Recently, many KGs have been curated and successfully applied to various real-world applications, ranging from Information Retrieval to Question Answering. However, current KGs are far from complete, which limits the application of knowledge. Therefore, we propose to build a Universal KG, including various types of knowledge, e.g., commonsense, facts, languages, and domains.
To achieve this, we propose a self-guided framework to automatically extract knowledge from natural languages, and to integrate existing structured knowledge. The basic idea is to learn more knowledge based on what we have learned. We will show (1) how to introduce existing knowledge to guide information extraction for scalability, (2) an efficient training-free graph matching framework with cross-graph knowledge inference and transfer, to robustly integrate heterogeneous knowledge, and (3) a self-supervised common semantic space to provide better knowledge guidance between texts and KG, and between different languages.
About the Speaker
Dr CAO Yixin is a Research Assistant Professor with Nanyang Technology University. Before that, he was a Research Fellow with NExT++, National University of Singapore (NUS). He received his Ph.D. in Computer Science from Tsinghua University in 2018. His research interests span natural language processing, knowledge graph and recommendation. Various part of his work have been published in top conferences, such as ACL, EMNLP, AAAI and WWW. Moreover, he has served as (Senior) PC member for several conferences including ACL, EMNLP, NAACL, AAAI, IJCAI, NeurIPs, and ICML, and as reviewers for journals including TPAMI, TACL, TKDE, and TOIS.
He is a tenure-track faculty candidate for the Artificial Intelligence & Data Science, Data Management & Analytics cluster.
