Semiparametric Latent Topic Modeling on Consumer-Generated Corpora

TL;DR

This paper introduces a semiparametric topic model combining nonnegative matrix factorization and regression, improving sparse topic reconstruction and prediction in small, limited-vocabulary consumer corpora.

Contribution

It presents a novel two-step semiparametric approach that enhances topic structure recovery and interpretability in consumer-generated texts.

Findings

Better reconstruction of sparse topics in small corpora

Improved prediction of topics in new documents

Produces interpretable and useful topic definitions

Abstract

Legacy procedures for topic modelling have generally suffered problems of overfitting and a weakness towards reconstructing sparse topic structures. With motivation from a consumer-generated corpora, this paper proposes semiparametric topic model, a two-step approach utilizing nonnegative matrix factorization and semiparametric regression in topic modeling. The model enables the reconstruction of sparse topic structures in the corpus and provides a generative model for predicting topics in new documents entering the corpus. Assuming the presence of auxiliary information related to the topics, this approach exhibits better performance in discovering underlying topic structures in cases where the corpora are small and limited in vocabulary. In an actual consumer feedback corpus, the model also demonstrably provides interpretable and useful topic definitions comparable with those produced by other methods.