Multi-Modal Bayesian Embeddings for Learning Social Knowledge Graphs

TL;DR

This paper introduces GenVector, a multi-modal Bayesian embedding model that links social network users and knowledge concepts in a shared latent space, improving social knowledge graph learning.

Contribution

The paper presents a novel multi-modal Bayesian embedding approach, GenVector, that effectively combines social network and knowledge base data for social knowledge graph construction.

Findings

GenVector outperforms state-of-the-art methods on three datasets.

Significant reduction in error rate in large-scale online deployment.

Successful integration of social and knowledge data in a shared latent space.

Abstract

We study the extent to which online social networks can be connected to open knowledge bases. The problem is referred to as learning social knowledge graphs. We propose a multi-modal Bayesian embedding model, GenVector, to learn latent topics that generate word and network embeddings. GenVector leverages large-scale unlabeled data with embeddings and represents data of two modalities---i.e., social network users and knowledge concepts---in a shared latent topic space. Experiments on three datasets show that the proposed method clearly outperforms state-of-the-art methods. We then deploy the method on AMiner, a large-scale online academic search system with a network of 38,049,189 researchers with a knowledge base with 35,415,011 concepts. Our method significantly decreases the error rate in an online A/B test with live users.