Moving Other Way: Exploring Word Mover Distance Extensions

TL;DR

This paper explores extensions to the Word Mover's Distance metric, incorporating word frequency and vector space geometry, and evaluates their effectiveness on document classification tasks.

Contribution

It introduces and empirically tests novel extensions of WMD, improving classification accuracy over the original metric.

Findings

Some extensions outperform WMD in classification error

Incorporating word frequency improves similarity measurement

Geometry-based modifications enhance WMD performance

Abstract

The word mover's distance (WMD) is a popular semantic similarity metric for two texts. This position paper studies several possible extensions of WMD. We experiment with the frequency of words in the corpus as a weighting factor and the geometry of the word vector space. We validate possible extensions of WMD on six document classification datasets. Some proposed extensions show better results in terms of the k-nearest neighbor classification error than WMD.