Hierarchical Latent Word Clustering

TL;DR

This paper introduces a Bayesian non-parametric model that extends Hierarchical Dirichlet Allocation to discover tree-structured word clusters, revealing meaningful hierarchical relationships in text data.

Contribution

It proposes a novel hierarchical clustering method for words using an extended Hierarchical Dirichlet Allocation model, enabling the extraction of structured word clusters.

Findings

Discovered meaningful hierarchical word structures in NIPS corpus.

Identified hierarchical clusters in radiology reports.

Demonstrated the model's effectiveness in capturing word relationships.

Abstract

This paper presents a new Bayesian non-parametric model by extending the usage of Hierarchical Dirichlet Allocation to extract tree structured word clusters from text data. The inference algorithm of the model collects words in a cluster if they share similar distribution over documents. In our experiments, we observed meaningful hierarchical structures on NIPS corpus and radiology reports collected from public repositories.