Learning Semantic Textual Similarity via Topic-informed Discrete Latent Variables

TL;DR

This paper introduces a topic-informed discrete latent variable model for semantic textual similarity, leveraging vector quantization and topic modeling to enhance semantic understanding and outperform existing neural baselines.

Contribution

The paper presents a novel discrete latent variable model that incorporates topic information and a semantic-driven attention mechanism to improve sentence similarity measurement.

Findings

Outperforms several strong neural baselines on English datasets

Utilizes topic modeling to enrich semantic representations

Enhances transformer models with semantic-driven attention

Abstract

Recently, discrete latent variable models have received a surge of interest in both Natural Language Processing (NLP) and Computer Vision (CV), attributed to their comparable performance to the continuous counterparts in representation learning, while being more interpretable in their predictions. In this paper, we develop a topic-informed discrete latent variable model for semantic textual similarity, which learns a shared latent space for sentence-pair representation via vector quantization. Compared with previous models limited to local semantic contexts, our model can explore richer semantic information via topic modeling. We further boost the performance of semantic similarity by injecting the quantized representation into a transformer-based language model with a well-designed semantic-driven attention mechanism. We demonstrate, through extensive experiments across various English language datasets, that our model is able to surpass several strong neural baselines in semantic textual similarity tasks.