Message Passing Query Embedding

TL;DR

This paper introduces a graph neural network-based approach for encoding complex knowledge graph queries, enabling more diverse query types and learning entity types without explicit supervision, with competitive performance on complex query answering.

Contribution

It presents a general architecture using GNNs for query embedding that surpasses previous ad-hoc methods in diversity and training efficiency.

Findings

Competitive performance on complex query answering

Can learn entity types without explicit supervision

Handles a broader set of query structures

Abstract

Recent works on representation learning for Knowledge Graphs have moved beyond the problem 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.