Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CAPE: A Chinese Dataset for Appraisal-based Emotional Generation using Large Language Models

June M. Liu, He Cao, Renliang Sun, Rui Wang, Yu Li, Jiaxing Zhang

TL;DR

This study introduces a two-stage automatic data generation framework to create CAPE, a Chinese dataset named Cognitive Appraisal theory-based Emotional corpus that facilitates the generation of dialogues with contextually appropriate emotional responses by accounting for diverse personal and situational factors.

Abstract

Generating emotionally appropriate responses in conversations with large language models presents a significant challenge due to the complexities of human emotions and cognitive processes, which remain largely underexplored in their critical role in social interactions. In this study, we introduce a two-stage automatic data generation framework to create CAPE, a Chinese dataset named Cognitive Appraisal theory-based Emotional corpus. This corpus facilitates the generation of dialogues with contextually appropriate emotional responses by accounting for diverse personal and situational factors. We propose two tasks utilizing this dataset: emotion prediction and next utterance prediction. Both automated and human evaluations demonstrate that agents trained on our dataset can deliver responses that are more aligned with human emotional expressions. Our study shows the potential for advancing emotional expression in conversational agents, paving the way for more nuanced and meaningful human-computer interactions.

CAPE: A Chinese Dataset for Appraisal-based Emotional Generation using Large Language Models

TL;DR

This study introduces a two-stage automatic data generation framework to create CAPE, a Chinese dataset named Cognitive Appraisal theory-based Emotional corpus that facilitates the generation of dialogues with contextually appropriate emotional responses by accounting for diverse personal and situational factors.

Abstract

Generating emotionally appropriate responses in conversations with large language models presents a significant challenge due to the complexities of human emotions and cognitive processes, which remain largely underexplored in their critical role in social interactions. In this study, we introduce a two-stage automatic data generation framework to create CAPE, a Chinese dataset named Cognitive Appraisal theory-based Emotional corpus. This corpus facilitates the generation of dialogues with contextually appropriate emotional responses by accounting for diverse personal and situational factors. We propose two tasks utilizing this dataset: emotion prediction and next utterance prediction. Both automated and human evaluations demonstrate that agents trained on our dataset can deliver responses that are more aligned with human emotional expressions. Our study shows the potential for advancing emotional expression in conversational agents, paving the way for more nuanced and meaningful human-computer interactions.

Paper Structure

This paper contains 30 sections, 1 equation, 8 figures, 10 tables.

Table of Contents

  1. Introduction
  2. Related works
  3. Textual Emotion Recognition
  4. Emotional Support Conversation
  5. Emotional Utterance Generation
  6. Method
  7. Definitions
  8. Data-generative Framework
  9. Data Quality Control
  10. Dataset Statistics
  11. Experiments
  12. Tasks
  13. Models
  14. Evaluation Metrics
  15. Results
  16. ...and 15 more sections

Figures (8)

  • Figure 1: An example of an emotion generation conversation with the analysis on the results of LLMs.
  • Figure 2: Overview of CAT-BEAR. It contains two stages: (1) AA and BB are initially assigned specific personalities, goals, and situational construal, which are used to generate their beliefs and knowledge (2) the appraisal process, where individuals evaluate the interaction across six dimensions (unpleasantness, control, responsibility, certainty, effort, and attention) to generate emotions and utterances.
  • Figure 3: Emotions distribution of CAPE
  • Figure 4: Guideline for GPT-4 to generate belief and knowledge
  • Figure 5: Guideline for GPT-4 to find the most appropriate emotion
  • ...and 3 more figures