Finding Interpretable Concept Spaces in Node Embeddings using Knowledge Bases

TL;DR

This paper introduces a method to identify human-interpretable concept subspaces within pre-trained node embeddings by leveraging external knowledge bases, enabling better understanding of the embeddings' structure.

Contribution

It presents a novel approach to explain node embeddings by discovering interpretable concept subspaces guided by knowledge bases, which was not previously explored.

Findings

Low error in identifying fine-grained concepts

Effective linear transformations for concept subspaces

Utilizes external knowledge bases for interpretability

Abstract

In this paper we propose and study the novel problem of explaining node embeddings by finding embedded human interpretable subspaces in already trained unsupervised node representation embeddings. We use an external knowledge base that is organized as a taxonomy of human-understandable concepts over entities as a guide to identify subspaces in node embeddings learned from an entity graph derived from Wikipedia. We propose a method that given a concept finds a linear transformation to a subspace where the structure of the concept is retained. Our initial experiments show that we obtain low error in finding fine-grained concepts.