Emotion Detection for Social Robots Based on NLP Transformers and an Emotion Ontology

Area of Science:

• Robotics
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Social robots require understanding human emotions for adaptive behavior.
• Emotion detection is crucial for robots interacting with humans across various media.
• Storing emotional data in semantic repositories enables advanced AI applications.

Purpose of the Study:

• To propose a framework for social robots to detect and store human emotions.
• To develop a proof-of-concept focusing on text-based emotion detection.
• To integrate emotion data into semantic repositories for enhanced robot functionality.

Main Methods:

• Developed a framework utilizing EMOTION ONTOLOGY (EMONTO) for emotion representation.
• Implemented text-based emotion detection using NLP transformers and speech-to-text conversion.
• Integrated the framework with a museum ontology for a tour-guide robot case study.

Main Results:

• Achieved emotion classification model performance comparable to state-of-the-art transformer models.
• Successfully demonstrated emotion detection from text and speech in a museum tour-guide robot scenario.
• Established a clear roadmap for future improvements and framework expansion.

Conclusions:

• The proposed framework enables social robots to effectively detect and semantically store human emotions.
• This integration enhances the potential for intelligent applications, such as personalized museum experiences.
• The system provides a scalable solution for emotion-aware human-robot interaction.