MailOrg: Intelligent Email Clustering & Summarization System

Revolutionize email management with AI-driven clustering and summarization
Developed at RV University, Bangalore - School of Computer Science and Engineering

---

🚀 Key Features

• Semantic Clustering: Group similar emails using BERT embeddings + K-Means
• Smart Summarization: Generate 2-line summaries via BART model
• Decision Support: Auto-suggest archive/delete/priority actions
• Storage Optimization: Reduce redundancy by 38% in testing
• Real-World Ready: Tested on 12,000+ genuine emails

---

📂 Dataset Overview

Property	Details
Total Emails	12,486 (3 personal accounts)
Training Set	9,988 emails
Test Set	2,498 emails
Categories	10 semantic groups
Avg. Email Length	127 words

---

🔍 System Architecture

graph TD
    A[Raw Emails] --> B[Text Preprocessing]
    B --> C[BERT Vectorization]
    C --> D[K-Means Clustering]
    D --> E[Cluster Analysis]
    B --> F[BART Summarization]
    F --> G[Summary Evaluation]
    E --> H[User Interface]
    G --> H

Loading

---

🛠️ Tech Stack

Component	Implementation Details
NLP Framework	Hugging Face Transformers
Embeddings	BERT-base-uncased (768-dim)
Clustering	K-Means++ with Euclidean distance
Summarization	DistilBART-CNN-12-6
Evaluation	ROUGE, Silhouette Score, DBI
Infrastructure	Apple M3, Google Colab (T4 GPU)

---

📊 Performance Metrics

Clustering Results

Metric	Train Score	Test Score
Silhouette Score	0.14	0.11
Davies-Bouldin Index	2.34	2.67
Calinski-Harabasz Score	418	213

Summarization Results

Metric	ROUGE-1	ROUGE-2	ROUGE-L	F1-Score
Daily Emails	0.42	0.31	0.38	0.68
Monthly Set	0.12	0.08	0.10	0.31

---

🚀 Getting Started

Prerequisites

• Python 3.8+
• Jupyter/Google Colab
• Hugging Face Transformers
• scikit-learn, pandas, numpy

Installation

git clone https://github.com/yourusername/MailOrg.git
cd MailOrg
pip install -r requirements.txt

Usage Workflow

1. Data Preparation
   Place raw emails in /data/raw directory as CSV files

2. Run Processing Pipeline

# Step 1: Preprocess and summarize
!jupyter nbconvert --execute Data_Preprocessing_and_Summarization.ipynb

# Step 2: Generate embeddings
!jupyter nbconvert --execute BERT_Vectorization_and_Clustering.ipynb

# Step 3: Evaluate results
!jupyter nbconvert --execute Evaluation_and_Result_Analysis.ipynb

---

📝 Sample Output

Original Email
"Dear Valued Customer, Your recent order #4512 has shipped via FedEx (tracking: 9274-2834-5532). Expected delivery: March 15. Contact support@example.com for queries."

Processed Result

📁 Cluster: Shipping Notifications (ID: 5)
📌 Summary: Order #4512 shipped via FedEx. Tracking: 9274***5532
💡 Action Suggested: Archive after review

---

📄 License

Distributed under MIT License. See LICENSE for details.

---

📧 Contact

Team Member	Email Address
C.J. Sakshi	cjsakshi.btech23@rvu.edu.in
Mohammed Ismail	mohammedi.btech23@rvu.edu.in
Pema Tshering Sherpa	pemats.btech23@rvu.edu.in
Rakshitha K	rakshithak.btech23@rvu.edu.in

Academic Advisor: Dr. Shabber Basha S H, RV University

---

---