Conceptual Blending Project

Overview

The Conceptual Blending Project develops an AI-driven system that integrates MeTTa symbolic reasoning with GPT-based conceptual blending. It implements four distinct conceptual blending networks:

• Simplex Network: Blends concepts using frame-role relationships.
• Mirror Network: Maps structural similarities between concepts.
• Single-Scope Network: Expands concepts within unified cognitive spaces.
• Double-Scope Network: Creates bidirectional concept mappings.

The system processes natural language queries and explicit concept pairs, leveraging structured reasoning to generate novel conceptual combinations stored in a MeTTa space for further analysis.

Directory Structure

.
|-- conceptual_blending
|   |-- agents
|   |   |-- gpt_agent.py          # GPT-based agent for generating blended concepts
|   |   |-- __init__.py           # Initialize the agent module
|   |   |-- llmagent.py           # Handles API authentication and requests
|   |-- __init__.py               # Initialize the conceptual_blending module
|   |-- main.py                   # Main logic for blending concepts and running agents
|   |-- prompts                   # Prompt templates for different blending networks
|-- requirements.txt             # Python dependencies for the project
|-- run-conceptual-blending.metta # MeTTa script for reasoning and running the project
|-- .env                         # Environment variables for API authentication

Setup and Run

1. Clone the repository and navigate to the project folder:

git clone https://github.com/iCog-Labs-Dev/conceptBlending.git
cd conceptBlending/

2. Create a virtual environment and activate it:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install the dependencies:

pip install -r requirements.txt

4. Configure API Keys

Create a .env file and add one of the following API credentials:

# ===================
# DEVELOPMENT (GitHub Marketplace OpenAI API)
# ===================
GITHUB_TOKEN=your_github_marketplace_api_key_here

# ===================
# PRODUCTION (OpenAI API)
# ===================
OPENAI_API_KEY=your_openai_api_key_here

5. Run the Project

Execute the project using the MeTTa script to test the conceptual blending functionality:

metta run-conceptual-blending.metta

The output will be:

(simplexBlend (blend music painting) auditoryCanvas)
(mirroredConcept (mirror light wisdom) illumination)
(singleScope (expand nature) interconnectedVitality)
(singleScope (expand technology) syntheticIngenuity)
(doubleScope (expand emotion mathematics) emotionalEquations)

---

Conceptual Blending Workflow

1. Agents

GPT Agent

• Purpose: Handles natural language input, identifies concepts, and generates blended concepts using AI.
• Functionality: Combines the concepts and produces a structured blended concept.

Example: For the input ! (gpt_simplex "Music" "Painting"), the GPT agent would return:

(simplexBlend (blend music painting) auditoryCanvas)

Blending Networks

• Simplex Network: Basic concept mapping using two explicit concepts.
• Mirror Network: Concept reflection using two explicit concepts.
• Single-Scope Network: Expands a single concept within a unified cognitive space.
• Double-Scope Network: Merges two distinct conceptual structures.

MeTTa Reasoning Agent

• Purpose: Acts as the symbolic reasoner, storing, reasoning over, and utilizing knowledge generated from agents.

---

Sample Workflow

;; Import the conceptual blending module for knowledge representation
! (import! &self conceptual_blending)

;; Ask the GPT agent with two explicit concepts for different networks
! (gpt_simplex "Music" "Painting")
! (gpt_mirror "Light" "Wisdom")
! (gpt_single "Nature" "Technology")
! (gpt_double "Emotion" "Mathematics")

;; Combine GPT with other agents and workflows
;; Store the blended knowledge (GPT response) from each network into the knowledge space
! (add-reduct &self (gpt_simplex "Music" "Painting"))
! (add-reduct &self (gpt_mirror "Light" "Wisdom"))
! (add-reduct &self (gpt_single "Nature" "Technology"))
! (add-reduct &self (gpt_double "Emotion" "Mathematics"))

;; Retrieve and list all knowledge stored in the current space
! (match &self ($x) $x)