Knowledge-Grounded Dialogue Flow Management for Social Robots and Conversational Agents

Area of Science:

• Artificial Intelligence
• Human-Computer Interaction
• Computational Linguistics

Background:

• Conversational agents and social robots require sophisticated dialogue management systems.
• Existing systems often struggle with maintaining coherent conversations and understanding user intent.
• The need for knowledge-based approaches to improve conversational quality is significant.

Purpose of the Study:

• To propose and evaluate a knowledge-based conversation system for social robots.
• To develop a dialogue management algorithm that selects appropriate topics based on user input.
• To enhance conversational coherence and avoid purely reactive responses.

Main Methods:

• Development of an ontology for conversation topics and their relationships.
• Implementation of a dialogue management algorithm focusing on user intent and topic selection.
• Comparative evaluation of five conversational agents with 100 participants using SASSI and a custom coherence survey.

Main Results:

• The proposed system demonstrated a more coherent conversational flow compared to keyword-based and random topic selection agents.
• Performance was assessed against a human baseline and a leading commercial chatbot (Replika).
• Subjective user perception of coherence and interaction quality was measured.

Conclusions:

• The knowledge-based approach significantly improves conversational coherence in social robots.
• Effective dialogue management is crucial for creating engaging and intuitive human-robot interactions.
• The proposed system offers a promising direction for developing more advanced conversational agents.