Message Passing Query Embedding

Recent works on representation learning for Knowledge Graphs have moved beyond the prob- lem of link prediction, to answering queries of an arbitrary structure. Existing methods are based on ad-hoc mechanisms that require training with a diverse set of query structures. We propose a more general architecture that employs a graph neural network to encode a graph representation of the query, where nodes correspond to entities and variables. The generality of our method allows it to encode a more diverse set of query types in comparison to previous work. Our method shows competitive performance against previous models for complex queries, and in contrast with these models, it can answer complex queries when trained for link prediction only. We show that the model learns entity embeddings that capture the notion of entity type without explicit supervision.

Few words about the Elsevier’s Discovery Lab:
The Elsevier’s Discovery Lab commenced officially in April 2020 and is a collaboration between the Vrije Universiteit Amsterdam, the University of Amsterdam and Elsevier. It will run for 5 years, and will include, besides its directors and managers, 3 PhD students and 2 post-doctoral researchers. The lab’s philosophy is to drive scientific discovery using machine intelligence. The researchers study and develop technology, infrastructure and methods to support the current transformation of science. They focus on data-driven activity, where scientists increasingly rely on intelligent tooling for searching and reading scientific literature, to formulate hypotheses, and to interpret data. Major areas of research focus include: Knowledge Graphs, Reinforcement Learning, and Question Answering.