Traversing Knowledge Graphs in Vector Space

TL;DR

This paper introduces a compositional training method for knowledge graph embedding models that significantly improves their ability to answer path queries and enhances overall knowledge base completion performance.

Contribution

It proposes a novel compositional training objective that reduces cascading errors and acts as structural regularization, leading to state-of-the-art results.

Findings

Dramatic improvement in path query answering accuracy.

Enhanced knowledge base completion performance with up to 43% error reduction.

Achieved new state-of-the-art results on standard benchmarks.

Abstract

Path queries on a knowledge graph can be used to answer compositional questions such as "What languages are spoken by people living in Lisbon?". However, knowledge graphs often have missing facts (edges) which disrupts path queries. Recent models for knowledge base completion impute missing facts by embedding knowledge graphs in vector spaces. We show that these models can be recursively applied to answer path queries, but that they suffer from cascading errors. This motivates a new "compositional" training objective, which dramatically improves all models' ability to answer path queries, in some cases more than doubling accuracy. On a standard knowledge base completion task, we also demonstrate that compositional training acts as a novel form of structural regularization, reliably improving performance across all base models (reducing errors by up to 43%) and achieving new state-of-the-art results.