S2WTM: Spherical Sliced-Wasserstein Autoencoder for Topic Modeling

TL;DR

This paper introduces S2WTM, a novel spherical autoencoder for topic modeling that effectively aligns latent distributions on a hypersphere, overcoming posterior collapse and producing more coherent topics.

Contribution

S2WTM is the first to use Spherical Sliced-Wasserstein distance in a hyperspherical autoencoder for improved topic modeling performance.

Findings

Outperforms state-of-the-art models in topic coherence

Generates more diverse and meaningful topics

Enhances downstream task performance

Abstract

Modeling latent representations in a hyperspherical space has proven effective for capturing directional similarities in high-dimensional text data, benefiting topic modeling. Variational autoencoder-based neural topic models (VAE-NTMs) commonly adopt the von Mises-Fisher prior to encode hyperspherical structure. However, VAE-NTMs often suffer from posterior collapse, where the KL divergence term in the objective function highly diminishes, leading to ineffective latent representations. To mitigate this issue while modeling hyperspherical structure in the latent space, we propose the Spherical Sliced Wasserstein Autoencoder for Topic Modeling (S2WTM). S2WTM employs a prior distribution supported on the unit hypersphere and leverages the Spherical Sliced-Wasserstein distance to align the aggregated posterior distribution with the prior. Experimental results demonstrate that S2WTM outperforms state-of-the-art topic models, generating more coherent and diverse topics while improving performance on downstream tasks.