Skip to content

README.md

Latest commit

 

History

History
110 lines (73 loc) · 3.3 KB

README.md

File metadata and controls

110 lines (73 loc) · 3.3 KB
title 📚 Notebook Directory & Workflow
nav_order 3

📚 Notebook Directory & Workflow

The notebooks in this directory are categorized into two groups: the Preparation Pipeline (mandatory for database enrichment) and Layer Demonstrations (standalone execution of specific system components).

⚙️ 1. Preparation Pipeline (Data Enrichment)

These notebooks must be executed to "bake" the logic and weights into the Neo4j graph. Run these if you are not using a pre-configured database dump.

Notebook Purpose Impact on Database
calculate-weights.ipynb Calculates Information Content (IC) for each symptom based on disease frequency. Adds weight property to :Symptom nodes.
generate-embeddings.ipynb Generates high-dimensional vectors using the multilingual-e5-large model. Adds embedding property to :Symptom nodes.

🧪 2. Layer Demonstrations (Individual Components)

These notebooks are used for research, validation, and testing each layer of the NeSy framework in isolation. They do not modify the database.

  • nlp-llm.ipynb (Neural Layer)

    • Focus: Symptom extraction.

    • Action: Demonstrates how an LLM parses unstructured user inputs into structured symptom lists.

  • inference-and-scoring.ipynb (Symbolic Layer)

    • Focus: Graph reasoning and ranking.

    • Action: Executes the hybrid Vector Search + Normalized Weighted Sum scoring directly against Neo4j to rank potential diseases.

  • xai-llm.ipynb (Neural Layer)

    • Focus: Clinical transparency.

    • Action: Takes raw inference results and generates natural language explanations, bridging the gap between numbers and human-readable diagnostics.

⚙️ Setting Up the Local Jupyter Kernel

Since this folder has its own dedicated environment, follow these steps to ensure VS Code uses the correct dependencies.

🛠️ Option 1: Manual Setup (Step-by-Step)

1. Navigate to the notebooks/ directory:

  cd NeSy/notebooks

2. Create a virtual environment:

  python -m venv .venv

3. Activate virtual environment:

  • Windows 
      .venv\Scripts\activate
  • macOS/Linux 
      source .venv/bin/activate

3. Install dependencies:

  pip install -r requirements.txt

4. Install Jupyter kernel:

  python -m ipykernel install --user --name=nesy-notebooks --display-name="NeSy Notebooks (venv)"

5. Select the Kernel in VS Code:

  1. Open any .ipynb file.
  2. Click Select Kernel in the top-right corner.
  3. Navigate to Jupyter Kernel -> NeSy Notebooks (venv).

Note: If the kernel does not appear immediately, restart VS Code and try selecting the kernel again.

🤖 Option 2: Automated Setup (One-Click Script)

We’ve included ready-to-run setup scripts in this folder — just double-click or run them to install everything automatically.

🖥️ For Windows

Option A: Double-click (easiest)

  1. Find setup.bat in the notebooks/ folder
  2. Double-click it — it will automatically:
    • Create .venv
    • Install dependencies
    • Register the Jupyter kernel
    • Show you next steps

Option B: Run from terminal

  1. Open terminal in notebooks/ folder

  2. Type:

    ./setup.bat

💡 No need to edit or create anything — script is pre-configured for this project.