S2WTM: Spherical Sliced-Wasserstein Autoencoder for Topic Modeling
This addresses a key bottleneck in topic modeling for text analysis, offering incremental improvements over existing methods.
The paper tackles posterior collapse in variational autoencoder-based neural topic models by proposing S2WTM, which uses a spherical sliced-Wasserstein distance to align distributions, resulting in improved topic coherence and diversity over state-of-the-art models.
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.