Personalization in Dialogue Systems

The field of dialogue systems is moving towards more personalized and user-centric approaches. Researchers are focusing on developing methods that can effectively model user traits, preferences, and goals to generate tailored responses. This shift is driven by the need to improve user experience and engagement in various applications, such as conversational search, recommendation, and task-oriented dialogue. Innovative approaches, including the use of diffusion models, large language models, and customer personas, are being explored to achieve this goal. Noteworthy papers in this area include: Simulating Before Planning, which introduces the User-Tailored Dialogue Policy Planning framework to model user traits and feedback. Know Me, Respond to Me, which presents the PERSONAMEM benchmark to evaluate LLMs' ability to recognize dynamic user profiles and generate personalized responses. You Are What You Bought, which proposes the concept of customer personas to provide readable and informative explicit user representations for e-commerce applications. PicPersona-TOD, which introduces a novel dataset for personalizing utterance style in task-oriented dialogue with image persona.

Sources

Simulating Before Planning: Constructing Intrinsic User World Model for User-Tailored Dialogue Policy Planning

Know Me, Respond to Me: Benchmarking LLMs for Dynamic User Profiling and Personalized Responses at Scale

Target Concrete Score Matching: A Holistic Framework for Discrete Diffusion

Planning with Diffusion Models for Target-Oriented Dialogue Systems

DIVE: Inverting Conditional Diffusion Models for Discriminative Tasks

You Are What You Bought: Generating Customer Personas for E-commerce Applications

PicPersona-TOD : A Dataset for Personalizing Utterance Style in Task-Oriented Dialogue with Image Persona

Built with on top of