Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Persona Extraction Through Semantic Similarity for Emotional Support Conversation Generation

Seunghee Han, Se Jin Park, Chae Won Kim, Yong Man Ro

TL;DR

The experimental results demonstrate that high-quality persona information inferred by PESS (Persona Extraction through Semantic Similarity) is effective in generating emotionally supportive responses.

Abstract

Providing emotional support through dialogue systems is becoming increasingly important in today's world, as it can support both mental health and social interactions in many conversation scenarios. Previous works have shown that using persona is effective for generating empathetic and supportive responses. They have often relied on pre-provided persona rather than inferring them during conversations. However, it is not always possible to obtain a user persona before the conversation begins. To address this challenge, we propose PESS (Persona Extraction through Semantic Similarity), a novel framework that can automatically infer informative and consistent persona from dialogues. We devise completeness loss and consistency loss based on semantic similarity scores. The completeness loss encourages the model to generate missing persona information, and the consistency loss guides the model to distinguish between consistent and inconsistent persona. Our experimental results demonstrate that high-quality persona information inferred by PESS is effective in generating emotionally supportive responses.

Persona Extraction Through Semantic Similarity for Emotional Support Conversation Generation

TL;DR

The experimental results demonstrate that high-quality persona information inferred by PESS (Persona Extraction through Semantic Similarity) is effective in generating emotionally supportive responses.

Abstract

Providing emotional support through dialogue systems is becoming increasingly important in today's world, as it can support both mental health and social interactions in many conversation scenarios. Previous works have shown that using persona is effective for generating empathetic and supportive responses. They have often relied on pre-provided persona rather than inferring them during conversations. However, it is not always possible to obtain a user persona before the conversation begins. To address this challenge, we propose PESS (Persona Extraction through Semantic Similarity), a novel framework that can automatically infer informative and consistent persona from dialogues. We devise completeness loss and consistency loss based on semantic similarity scores. The completeness loss encourages the model to generate missing persona information, and the consistency loss guides the model to distinguish between consistent and inconsistent persona. Our experimental results demonstrate that high-quality persona information inferred by PESS is effective in generating emotionally supportive responses.
Paper Structure (13 sections, 6 equations, 2 figures, 5 tables)

This paper contains 13 sections, 6 equations, 2 figures, 5 tables.

Table of Contents

  1. INTRODUCTION
  2. METHOD
  3. Problem Definition
  4. Semantic Similarity for Consistency Measurement
  5. Completeness Loss
  6. Consistency Loss
  7. Training Objectives
  8. Response Generation
  9. EXPERIMENTS
  10. Setup
  11. Persona Extractor Evaluation
  12. Response Generation Evaluation
  13. CONCLUSION

Figures (2)

  • Figure 1: The overall process of obtaining the consistent persona $\mathcal{P}^\text{con}$ and the missing persona $\mathcal{P}^\text{miss}$ based on semantic similarity score matrix $\mathcal{S }$. New target persona $\mathcal{P}^\text{new}$ is constructed by combining $\mathcal{P}^\text{con}$ and $\mathcal{P}^\text{miss}$.
  • Figure 2: Overview of PESS-GEN. PESS generates the persona of speaker A, given the speaker A's utterances $\mathcal{U}_A$. The predicted persona $\mathcal{P}^g$ is passed to the response generation model with the dialogue history $\mathcal{U}$ to generate the speaker B's following response $u^B_n$.