Augmenting word2vec with latent Dirichlet allocation within a clinical application

TL;DR

This paper introduces three hybrid models combining LDA and word2vec to improve Alzheimer's disease detection from speech transcripts, achieving state-of-the-art results on the DementiaBank dataset.

Contribution

The paper proposes novel hybrid models integrating LDA and word2vec specifically for clinical speech analysis, demonstrating improved diagnostic accuracy.

Findings

Two models outperform current state-of-the-art F-scores

Models effectively distinguish Alzheimer's from non-Alzheimer's speech

Hybrid approach enhances clinical language analysis

Abstract

This paper presents three hybrid models that directly combine latent Dirichlet allocation and word embedding for distinguishing between speakers with and without Alzheimer's disease from transcripts of picture descriptions. Two of our models get F-scores over the current state-of-the-art using automatic methods on the DementiaBank dataset.