A Comparative Study of Distributional and Symbolic Paradigms for Relational Learning

Many real-world domains can be expressed as graphs and, more generally, as multi-relational knowledge graphs. Though reasoning and learning with knowledge graphs has traditionally been addressed by symbolic approaches such as Statistical relational learning, recent methods in (deep) representation learning have shown promising results for specialised tasks such as knowledge base completion. These approaches, also known as distributional, abandon the traditional symbolic paradigm by replacing symbols with vectors in Euclidean space. With few exceptions, symbolic and distributional approaches are explored in different communities and little is known about their respective strengths and weaknesses. In this work, we compare distributional and symbolic relational learning approaches on various standard relational classification and knowledge base completion tasks. Furthermore, we analyse the properties of the datasets and relate them to the performance of the methods in the comparison. The results reveal possible indicators that could help in choosing one approach over the other for particular knowledge graphs.

Improving Classification of Interlinked Entities Using Only the Network Structure

This paper presents a classifier architecture that is able to deal with classification of interlinked entities when the only information available is the existing relationships between these entities, i.e. no semantic content is known for either the entities or their relationships. After proposing a classifier to deal with this problem, we provide extensive experimental evaluation showing that our proposed method is sound and that it is able to achieve high accuracy, in most cases much higher than other already existing algorithms configured to tackle this very same problem. The contributions of this paper are twofold: first, it presents a classifier for interlinked entities that outperforms most of the existing algorithms when the only information available is the relationships between these entities; second, it reveals the power of using label independent (LI) features extracted from network structural properties in the bootstrapping phases of relational classification.