SEEDS: Similarity-based Expert Embedding Decision System

📌 Overview

SEEDS (Similarity-based Expert Embedding Decision System) is a Retrieval-Augmented Generation (RAG) based agricultural question-answering (QA) system. It is built upon a domain-specific knowledge graph (KG), representing Cedar Apple Rust disease.

🔍 Core Idea:

• Uses Knowledge Graphs (KGs) with plant defense, pesticide, and fungal infection data.

• Generates embeddings for queries and KG data using OpenAI's text-embedding-ada-002 model.

• Computes cosine similarity to rank the most relevant KG entries for a given query.

• Provides detailed responses using GPT-4 Turbo for expert-level responses.

• Incorporates Explainable AI (XAI) via t-SNE, PCA, and clustering visualizations.

🧪 Goal: Assist plant protection experts and agriculturists in making data-driven decisions for pest control.

---

🚀 Features

• ✅ Knowledge Graph-based Question Answering
• ✅ Embedding-based Similarity Matching
• ✅ Explainable AI (XAI) Visualizations
• ✅ Support for Pesticide & Plant Pathway Data
• ✅ Interactive Dash Web Application
• ✅ Retrieval-Augmented Generation (RAG)

---

📂 Project Structure

SEEDS/
│── dummy data/  # Knowledge graph (KG) and question datasets
│   ├── dummy_mesh.csv
│   ├── dummy_questions.txt
│   ├── synthetic_properties.csv
│   ├── synthetic_triples.csv
│── src/                   # Source code
│   ├── QA_ada.py           # QA system using embedding similarity & detailed response generation by GPT       
│   ├── XAI.py             # Dash web app for Explainable AI visualizations
│── .gitignore             # Ignore venv, embeddings, results, sensitive files
│── LICENSE # Open-source license information
│── README.md              # Project documentation
│── requirements.txt       # Required dependencies

---

⚙️ Installation & Setup

1️⃣ Clone the Repository

git clone https://github.com/YOUR_USERNAME/SEEDS.git
cd SEEDS

2️⃣ Create Virtual Environment (Recommended)

python -m venv .seeds
source .seeds/bin/activate  # On Mac/Linux
.seeds\Scripts\activate     # On Windows

3️⃣ Install Dependencies

pip install -r requirements.txt

4️⃣ Set up `` file

Create a .env file with:

OPENAI_API_KEY=your-api-key-here

---

🎯 Usage

🔹 1. Run the QA System (CLI-Based)

python src/QA_ada.py --prompts "dummy data/dummy_questions.txt" --triples "dummy data/synthetic_triples.csv" --properties "dummy data/synthetic_properties.csv" --output "results/output.csv" --prompt_emb_path "embeddings/prompt_embeddings.pkl" --kg_text_emb_path "embeddings/kg_text_embeddings.pkl" --kg_node_emb_path "embeddings/kg_node_info_embeddings.pkl"

🔹 2. Run the XAI Dashboard (Web App)

python src/XAI.py --prompts "dummy data/dummy_questions.txt" --triples "dummy data/synthetic_triples.csv" --properties "dummy data/synthetic_properties.csv" --mesh_terms "dummy data/dummy_mesh.csv" --embeddings "embeddings/"

• Open in browser: http://127.0.0.1:8050

---

📊 Example Input & Output

📝 Sample Query

What is the effect of Fungicide X on Venturia Fungi?

📌 Top KG Results

1. Fungicide X inhibits Venturia by disrupting cell walls (Similarity: 0.92)
2. Fungicide X affects fungal spore germination (Similarity: 0.89)
3. Venturia causes apple and pear scab (Similarity: 0.78)

🤖 GPT-4 Generated Answer

Fungicide X effectively inhibits the growth of Venturia Fungi by disrupting its cell wall integrity. The mode of action primarily targets fungal spores, preventing germination and further infection. In field trials, it has shown strong antifungal properties against various Venturia strains. Thus, fungicide X can be used to manage apple and pear scab diseases.

---

📜 Citation

If you use this work, please cite:

@misc{SEEDS2024,
author = {Astha Anand and Marc Jacobs},
title = {SEEDS: Similarity-based Expert Embedding Decision System},
year = {2024},
institution = {Fraunhofer Institute for Algorithms and Scientific Computing (SCAI)},
howpublished = {\url{https://github.com/SCAI-BIO/SEEDS}},
note = {GitHub repository} }

---

🔥 Future Work

• Expand KG with more crops, pests, weather data, etc.
• Integrate multimodal features (e.g., plant disease images)
• User feedback loop for iterative model improvement.

---

📬 Contact

• Author: Astha Anand
• Supervisor: Dr. Marc Jacobs
• Affiliation: Fraunhofer Institute for Algorithms and Scientific Computing (SCAI)
• Email: astha.anand@scai.fraunhofer.de

---

🚀 Happy Coding! Let’s improve agricultural pest control with AI! 🎯🌿