Program

Lecture 2 – Reasoning in Knowledge Graphs

Ricardo Guimarães, Ana Ozaki

Knowledge Graphs (KGs) are becoming increasingly popular in the industry and academia. They can be represented as labelled graphs conveying structured knowledge in a domain of interest, where nodes and edges are enriched with metaknowledge such as time validity, provenance, language, among others. Once the data is structured as a labelled graph one can apply reasoning techniques to extract relevant and insightful information. We provide an overview of deductive, inductive and abductive reasoning approaches for reasoning in KGs.

Ricardo Guimarães is a postdoctoral research fellow at the University of Bergen. He works in Artificial Intelligence (AI), specifically with Knowledge Representation and Reasoning (KR), focusing on Description Logics. Currently, he is working towards the combination of Knowledge Representation approaches with Machine Learning methods, focussing on Ontologies and Knowledge Graphs.

Ana Ozaki is an associate professor at the University of Bergen, Norway. She is an AI researcher in the field of knowledge representation and reasoning and in learning theory. Ozaki is interested in the formalisation of the learning phenomenon so that questions involving learnability, complexity, and reducibility can be systematically investigated and understood. Her research focuses on learning logical theories formulated in description logic and related formalisms for knowledge representation. She is the principal investigator of the project Learning Description Logic Ontologies funded by RCN.

Slides Video

Lecture 3 – Reasoning about Concepts with Ontologies and Vector Space Embeddings

Steven Schockeart, Víctor Gutiérrez-Basulto, Zied Bouraoui

Ontologies and vector space embeddings are among the most popular frameworks for encoding conceptual knowledge. Ontologies excel at capturing the logical dependencies between concepts in a precise and clearly defined way. Vector space embeddings excel at modelling similarity and analogy. Given these complementary strengths, various research lines have focused on developing frameworks that can combine the best of both worlds. In this chapter, we present an overview of the work in this area. We first discuss the theory of conceptual spaces, which was proposed in the 1990s by Gärdenfors as an intermediate representation layer, in between embeddings and symbolic knowledge bases. Second, we discuss approaches where symbolic knowledge is modelled in terms of geometric constraints, which are used to constrain or regularise vector space embeddings. Finally, we discuss methods in which similarity, and other forms of conceptual relatedness, are derived from vector space embeddings and subsequently used to support flexible forms of reasoning with ontologies.

Lecture 4 – Neuro-Symbolic Methods for Fact Prediction

 

Knowledge bases are typically incomplete. Recent years have seen two approaches to guess missing facts: Rule Mining and Knowledge Graph Embeddings. The first approach is symbolic, and finds rules such as “If two people are married, they most likely live in the same city”. These rules can then be used to predict missing statements. Knowledge Graph Embeddings, on the other hand, are trained to predict missing facts for a knowledge base by mapping entities to a vector space. Each of these approaches has their strengths and weaknesses, and this article provides a survey of works that have taken to combine embeddings and symbolic approaches.

Lecture 5 — Learning and Reasoning with Graph Data: Neural and Statistical-Relational Approaches

Manfred Jaeger

Graph neural networks (GNNs) have emerged in recent years as a very powerful and popular modeling tool for graph and network data. Though much of the work on GNNs has focused on graphs with a single edge relation, they have also been adapted to multi-relational graphs, including knowledge graphs. In such multi-relational domains, the objectives and possible applications of GNNs become quite similar to what for many years has been investigated and developed in the field of statistical relational learning (SRL). In this lecture I will give a brief overview of the main features of GNN and SRL approaches to learning and reasoning with graph data. I will then in more detail analyze their commonalities and differences  with respect to semantics, representation, parameterization, interpretability, and flexibility. A particular focus will be on relational Bayesian networks (RBNs) as the SRL framework that is most closely related to GNNs. I will show how most common GNN architectures can be directly encoded as RBNs, thus enabling the direct integration of “low level” neural model components with the “high level” symbolic representation and flexible inference capabilities of SRL. This lecture does not require any specific pre-requisites, though a little familiarity with first-order logic for knowledge representation will be an advantage.

Manfred Jaeger studied mathematics in Freiburg, Germany, where he obtained his diploma in mathematics in 1991. He subsequently went to the Max-Planck-Institute for Computer Science in Saarbrücken, where he obtained a PhD in Computer Science from Saarland University in 1995. From 1996-2003 he continued as research associate at the Max-Planck-Institute for Computer Science, and in that period also spent time as postdoctoral researcher at Stanford University, the University of Helsinki, and Freiburg University. In 2002 he obtained the Habilitation in computer science from Saarland University. Since 2003 he is associate professor at the Computer Science department at Aalborg University, Denmark.
Manfred Jaeger has served as associate editor for the Journal of Artificial Intelligence Research and the Artificial Intelligence Journal. He is currently a member of the editorial board of Machine Learning.

Slides 1  Slides 2  Slides 3 Video

Lecture 6 – Automating Moral Reasoning

Marija Slavkovik

Artificial Intelligence ethics is concerned with ensuring a nonnegative ethical impact of ethical impact of researching, developing, deploying and using AI systems. One way to accomplish that is to enable those AI systems to make moral decisions in ethically sensitive situations, i.e., automate moral reasoning. Machine ethics is an interdisciplinary research area that is concerned with the problem of automating moral reasoning. This tutorial presents the problem of making moral decisions and gives a general overview of how a computational (artificial) agent can be constructed to make moral decisions. The tutorial is aimed for students in artificial intelligence who are interested in acquiring a starting understanding of the basic concepts and a gateway to the literature in machine ethics.

Slides Video

Workshop Program

Track 1: Knowledge Graphs & Embeddings

Oral Presentations:   

• Aleksandar Pavlovic: Injecting Knowledge Graphs into Machine Learning Models
• Daniel Daza: Inductive Entity Representations from Text via  Link Prediction
• Vincenzo Suriani: Semantic Mapping Generalization in Robotics using Knowledge  Embedding Techniques

Posters: 

• Moritz Blum
• Taha Halal
• Victor Lacerda
• Taraneh Younesian

Track 2: Ontologies & Logic-based Reasoning 

Oral Presentations:

• Chuangtao Ma: Knowledge Graph Enhanced Schema Matching Network for  Heterogeneous Data Integration
• Elena Romanenko: Pattern-based Ontology Summarization
• Filippo De Bortoli: Description Logics with Expressive Cardinality Constraints

Posters:   

• Anton Gnatenko
• Arka Ghosh 
• Nikolaos Kondylidis
• Cosimo Persia: On the Learnability of Possibilistic Theories –  winner of best poster award by public vote! 🙂

Track 3: Applications 

Oral Presentations:

• Lauren DeLong: Neurosymbolic Graph Reasoning: the Future of the Biomedical Domain
• Marina Boudin: Computational approach for drug repositioning: towards an holistic  perspective with knowledge graphs (OREGANO)

Posters:   

• Adriaan Ludl
• Aida Ashrafi
• Aidan ​​Marnane

Track 4: Miscellaneous 

Oral Presentations:

• Emile van Krieken: Bridging the Discrete-Continuous gap in  Neuro-Symbolic AI – winner of the best oral presentation award by public vote! 🙂
• Kun Zhang: Question Answering over Heterogeneous Graphs  
• Nitisha Jain: Knowledge Graph Representation with Embeddings

Posters: 

• Peter Fratric
• Sarang Shaik