Hyperbolic Centroid Calculations for Text Classification

TL;DR

This paper explores hyperbolic centroid methods for text classification, addressing the challenge of averaging hyperbolic word embeddings, and evaluates their effectiveness in NLP tasks.

Contribution

It introduces and analyzes hyperbolic centroid schemes for document representation, a novel approach in hyperbolic NLP embeddings.

Findings

Hyperbolic centroid schemes outperform Euclidean averaging in classification tasks.

Certain hyperbolic schemes significantly improve text classification accuracy.

The study provides insights into the geometric properties of hyperbolic embeddings.

Abstract

A new development in NLP is the construction of hyperbolic word embeddings. As opposed to their Euclidean counterparts, hyperbolic embeddings are represented not by vectors, but by points in hyperbolic space. This makes the most common basic scheme for constructing document representations, namely the averaging of word vectors, meaningless in the hyperbolic setting. We reinterpret the vector mean as the centroid of the points represented by the vectors, and investigate various hyperbolic centroid schemes and their effectiveness at text classification.