In the past years, knowledge graphs have grown from niche academic endeavours to becoming critical assets that many IT companies have invested in. We now have extensive collections of facts enabling impressive new knowledge-driven applications. But what is the future of knowledge graphs? In this talk, I will outline several noteworthy future directions for knowledge graphs. The first is that of a more unified knowledge integration from heterogeneous sources. The second is that of a much tighter coupling with the massive amounts of natural language available online. The third is the integration of multimodal data, especially images and video. Finally, the fourth direction is a move towards cognitive and neural approaches to knowledge. Examples from my work on resources such as UWN/MENTA, Lexvo.org, FrameBase, WebChild, and Knowlywood will be given to show possible implementations of some of these ideas.


Gerard de Melo is a Tenure-Track Assistant Professor at Tsinghua University, Beijing, where he is heading the Web Mining and Language Technology group. He has published over 50 research papers in these areas, being awarded Best Paper awards at CIKM 2010, ICGL 2008, and the NAACL 2015 Workshop on Vector Space Modeling, as well as an ACL 2014 Best Paper Honorable Mention, a Best Student Paper Award nomination at ESWC 2015, and the WWW 2011 Best Demonstration Award, among others. Prior to joining Tsinghua, de Melo had spent two years as a Visiting Scholar at UC Berkeley, working in ICSI's AI/FrameNet group. He received his doctoral degree at the Max Planck Institute for Informatics in Germany. For more information, please refer to his home page at http://gerard.demelo.org.