Enhancing Personalized Dialogue Generation with Contrastive Latent Variables: Combining Sparse and Dense Persona

TL;DR

This paper introduces a Contrastive Latent Variable model that combines sparse, dense, and dialogue history data to improve personalized dialogue generation, achieving better personalization in experiments.

Contribution

It proposes a novel CLV model that clusters dense persona descriptions into sparse categories and integrates dialogue history for enhanced personalization.

Findings

Outperforms baselines in Chinese and English datasets

Effectively clusters dense descriptions into meaningful categories

Improves personalization accuracy in dialogue generation

Abstract

The personalized dialogue explores the consistent relationship between dialogue generation and personality. Existing personalized dialogue agents model persona profiles from three resources: sparse or dense persona descriptions and dialogue histories. However, sparse structured persona attributes are explicit but uninformative, dense persona texts contain rich persona descriptions with much noise, and dialogue history query is both noisy and uninformative for persona modeling. In this work, we combine the advantages of the three resources to obtain a richer and more accurate persona. We design a Contrastive Latent Variable-based model (CLV) that clusters the dense persona descriptions into sparse categories, which are combined with the history query to generate personalized responses. Experimental results on Chinese and English datasets demonstrate our model's superiority in personalization.