Bilingual Dialogue Dataset with Personality and Emotion Annotations for Personality Recognition in Education

Area of Science:

• Computational Linguistics
• Artificial Intelligence
• Affective Computing

Background:

• Existing dialogue datasets lack crucial personality and emotion annotations, hindering the development of personalized and human-like conversational AI.
• This limitation impacts user experience and the ability of natural language processing (NLP) models to capture nuanced human interaction.

Purpose of the Study:

• To construct comprehensive bilingual (Chinese and English) dialogue datasets enriched with Big Five personality traits and emotion annotations.
• To address the scarcity of diverse personality representations in existing dialogue resources.

Main Methods:

• Utilized the AutoGen tool within a multi-agent framework to generate multi-turn question-answering dialogue datasets based on fables.
• Created persona agents with diverse personalities to ensure heterogeneity and overcome limitations in personality diversity.
• Integrated emotion annotations for each utterance.

Main Results:

• Successfully constructed bilingual dialogue datasets with integrated Big Five personality traits and emotion annotations.
• Validated the quality of utterances and investigated the alignment between conversational content and speaker personality traits.
• Demonstrated enhanced personality heterogeneity through persona agent creation.

Conclusions:

• The developed dataset is a valuable resource for advancing emotionally intelligent dialogue systems and research in personality and affective computing.
• The dataset facilitates the development of NLP models capable of understanding and generating dialogues that reflect personality and emotion.
• Enables the creation of emotion-aware systems for automatic personality trait detection.