Conceptual Blending Module

This project implements a set of AI-driven conceptual blending systems that integrate symbolic reasoning (MeTTa) and optimization techniques to generate creative conceptual combinations. The system supports GPT-enhanced conceptualization, constraint-based scoring, and information-theoretic evaluation of conceptual blends.

Submodules

🔹 Info-Theoretic Approach with CMA-ES and GA

This is the main executable pipeline that combines:

• Symbolic concept representation in MeTTa
• Information-theoretic analysis (entropy, emergence, mutual information)
• Optimization using CMA-ES and Genetic Algorithms
• Integration with GPT and ConceptNet agents
• Evaluation through Optimality Constraints (e.g., topology, unpacking, good reason)

🔹 Naive Blending Approach (Graph-Matching-Based)

A simpler conceptual blending prototype based on graph matching and GPT.

What is Conceptual Blending

Conceptual blending is a cognitive process of combining elements from multiple conceptual spaces to form novel and meaningful ideas. This project automates blending using:

• Symbolic AI (MeTTa) to represent and manipulate concepts.
• Information-theoretic methods to quantify blend quality.
• LLMs (like GPT) to extract and augment conceptual features.
• Optimization strategies to search for high-fitness blends.

---