Constructing Word Embeddings from the Random Hyperbolic Graph

Abstract

Recent studies have shown that the appropriate space for word embeddings is not the Euclidean space, but negatively curved, hyperbolic space. We randomly throw points in the hyperbolic disc and claim that these points are already word representations. However, it is yet to be uncovered which point corresponds to which word of the human language of interest. This correspondence can be approximately established using a pointwise mutual information between words and graph matching techniques. The embeddings were evaluated at WS353 task, and then separately on its similarity and relatedness parts.