U.S. Semantic Technologies Symposium Series
Session - Current and Future Trends of Knowledge Graph Representation and Reasoning
Session Organizers
- Monireh Ebrahimi, Kansas State Univerity
- Bassem Makni, IBM T.J. Watson Research Center
- Aaron Eberhart, Kansas State Univerity
- Amir Hossein Yazdavar, Kansas State Univerity
Abstract
How can we represent knowledge graphs such that the resulting semantic structures representation becomes useful in a downstream task at hand particularly logical reasoning and question answering? The Semantic Technology community, as shown by annual initiatives like the Semantic Deep Learn-ing workshop (the 4th edition was co-located with the recent ISWC 2018), and by the increasing numbers of journals and conference papers focused on the integration be-tween semantic web and machine learning research, is more and more interested in the problem of applying deep models to the semantic web problems. In this regard, deep models can be viewed as a tool or source of insight for overcoming the key challenges of knowledge representation learning and reasoning. While representation learning has got tremendous attention in terms of learning senses and their trans-formation to embeddings, which are highly useful for word sense disambiguation, entity linking, and semantic search; the logical reasoning has received relatively less attention. This presentations-session aims at exposing existing solutions towards representation learning for knowledge graph reasoning, their limitations and the potential benefits of state-of-the-art deep learning approaches to address them. The format is intended to promote a targeted discussion around such problems, by engaging experts from the us2tscommunity.In the first part of the presentations session, the presenters will introduce the survey of current works on knowledge graph representation. The second part of the session will focus on the most relevant techniques in the state of the art, highlighting best practices and limitations in RDF graph reasoning. Finally, the presenters will illustrate how machine learning approaches can be used to complement knowledge representation and reasoning systems, opening the discussion to the attendants.