🧠 Ontology Population with LLMs

This project demonstrates a framework for automatically populating OWL ontologies from unstructured natural language text using Large Language Models (LLMs). It builds on the ontology population methodology presented in the paper "Robot Situation and Task Awareness using Large Language Models and Ontologies" and is written entirely in Python.

---

📌 Project Overview

The system enables intuitive and efficient human-robot interaction by transforming natural language scene descriptions into structured ontological knowledge. This knowledge is represented in OWL format which enables further reasoning and inference, which then can be used to enhance robotic planning.

The pipeline performs:

• Entity extraction
• Property (attribute) extraction
• Relation extraction
• Ontology instantiation via Owlready2

---

🧾 Folder Structure

.
├── ontologies/
│   ├── modified_ontologies/
│   └── test_ontologies/
├── src/
│   ├── ontology_manager.py
│   └── ontopopulator.py
├── tests/
│   └── onto_instanciation/
├── Example_texts/
│   └── natural_text.txt
├── config.template.json
├── requirements.txt
├── setup.py
└── README.md

---

⚙️ Setup the Environment

Make sure you have Python 3.8+ installed.

python -m venv venv
source venv/bin/activate       # On Windows: venv\Scripts\Activate.ps1
pip install -r requirements.txt

---

🔑 Configuration

All parameters are stored in a config.json file. Here's a template:

{
  "API_KEY": "your-openai-api-key",
  "ontology": "your-ontology-name.owl",
  "ontologies_directory_path": "path/to/ontologies/",
  "text": "path/to/text.txt",
  "SAVE_PATH": "path/to/save/instantiated-ontology/"
}

---

🚀 Usage

from src.ontopopulator import OntoPopulator
import json

# Load config
with open("config.json") as f:
    config = json.load(f)

# Instantiate and run
api_key = config.get("API_KEY")
op = OntoPopulator(config.get("ontologies_directory_path"), config.get("ontology"), api_key=api_key)
results = op.populate_from_text(text_path=config.get("text"))

# Access outputs
print(results)
onto = results.get("ontology")

---

🧪 Testing

An example code to execute the program can be found at tests/onto_instanciation/. And an example text for the population can be found at tests/onto_instanciation/Example_texts

---

📝 License

Specify your license here (e.g., MIT, Apache 2.0). If you don’t have one, consider using Choose a License.

---

🤝 Acknowledgements

This project uses:

• OpenAI GPT-4o-mini
• Owlready2
• Langchain

---