ResearchPortal — AI-Powered Research Paper Analyzer

---

📖 Table of Contents

• Overview
• Key Features
• Architecture & Workflow
• Project Structure
• Supported LLM Models
• Technology Stack
• Getting Started
• Challenges & Solutions
• Roadmap
• Contributing
• Acknowledgements
• Author

---

Overview

ResearchPortal is an end-to-end Retrieval-Augmented Generation (RAG) system designed to extract, analyze, and answer questions from research paper PDFs with minimal hallucination.

By combining semantic search powered by FAISS with LLM reasoning via Groq and LLaMA 3, the platform delivers accurate, context-aware responses for academic research workflows.

How It Works

Upload PDF  →  Select Section  →  View Summary  →  Ask Questions

---

Key Features

Feature	Description
📤 PDF Upload	Upload research papers in PDF format with a simple web interface
🔍 Section Detection	Automatically identifies and splits structured academic sections using regex-based parsing
🤖 LLM Refinement	Cleans and normalizes noisy text extracted from PDFs using Groq-powered LLM processing
📝 Section-wise Summaries	Generates detailed, AI-driven summaries for each detected section of the paper
💬 RAG Chat System	Interactive Q&A interface to query the contents of the uploaded paper
🧠 Semantic Search	FAISS vector store combined with HuggingFace sentence embeddings for precise retrieval
🌐 Flask Web UI	Lightweight, interactive web interface built with Flask

---

Architecture & Workflow

The system follows a multi-stage pipeline from PDF ingestion to answer generation:

┌─────────────────┐
│   PDF Upload    │
└────────┬────────┘
         ▼
┌─────────────────┐
│  Text Extraction │   ← PyPDF2
└────────┬────────┘
         ▼
┌─────────────────┐
│ Section Detection│  ← Regex-based parsing
└────────┬────────┘
         ▼
┌─────────────────┐
│  LLM Refinement  │  ← Groq (text cleanup)
└────────┬────────┘
         ▼
┌─────────────────┐
│  Text Chunking   │
└────────┬────────┘
         ▼
┌─────────────────┐
│   Embedding      │  ← HuggingFace Embeddings
└────────┬────────┘
         ▼
┌─────────────────┐
│ FAISS Vector     │  ← Semantic search store
│ Search           │
└────────┬────────┘
         ▼
┌─────────────────┐
│  LLM Answer       │  ← Groq + LLaMA 3
│  Generation       │
└─────────────────┘

---

Project Structure

ResearchPortal/
├── app.py                          # Flask application entry point
├── src/
│   ├── load_and_extract_text.py    # PDF text extraction (PyPDF2)
│   ├── detect_and_split_sections.py # Regex-based section detection
│   ├── get_summary.py              # LLM-powered section summarization
│   ├── create_vector_db.py         # FAISS vector store creation
│   └── RAG_retrival_chain.py       # RAG retrieval & answer chain
├── templates/
│   └── index.html                  # Web UI template
├── uploads/                        # Uploaded PDFs (git-ignored)
├── .env                            # Environment variables (git-ignored)
├── requirements.txt               # Python dependencies
└── README.md                       # Project documentation

---

Supported LLM Models

All models are served through the Groq API platform:

Model	Context Window	Best For
llama3-8b-8192	8,192 tokens	⚡ Recommended — Fast inference, balanced performance
llama3-70b-8192	8,192 tokens	High-quality reasoning responses
mixtral-8x7b-32768	32,768 tokens	Long-context processing
gemma-7b-it	8,192 tokens	Lightweight alternative

---

Technology Stack

Backend

Technology	Role
Flask	Web framework & API server
LangChain	RAG orchestration & chaining
FAISS	Vector similarity search
PyPDF2	PDF text extraction
Groq API	LLM inference backend
HuggingFace Embeddings	Sentence-level text embedding

Frontend

Technology	Role
HTML / CSS / JavaScript	Web interface
Font Awesome	Iconography
Google Fonts	Typography

---

Getting Started

Prerequisites

• Python 3.9 or higher
• A Groq API key (available at groq.com)

1. Clone the Repository

git clone https://github.com/Intelli2Byte/Automated-Research-Synthesis-ARS-.git
cd Automated-Research-Synthesis-ARS-

2. Create a Virtual Environment

python -m venv venv

# Activate on Windows
venv\Scripts\activate

# Activate on macOS / Linux
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

4. Configure Environment Variables

Create a .env file in the project root:

GROQ_API_KEY=your_api_key_here

5. Launch the Application

python app.py

The web interface will be available at http://127.0.0.1:5000 by default.

---

Challenges & Solutions

Challenge	Approach
Noisy PDF text extraction	LLM-based refinement pipeline to clean and normalize extracted text
Accurate section identification	Regex-based detection supplemented with LLM processing
Reducing LLM hallucination	RAG architecture grounds responses in retrieved source context
Optimizing retrieval performance	FAISS vector index with HuggingFace embeddings for efficient semantic search

---

Roadmap

• Multi-paper comparison — Analyze and compare multiple research papers side by side
• Citation extraction — Automatically extract and format citations from papers
• UI enhancements — Improved user experience and responsive design
• Cloud deployment — Production-ready cloud hosting and scaling

---

Contributing

Contributions are welcome! Please follow these steps:

1. Fork the repository

2. Create a feature branch

   git checkout -b feature/your-feature-name

3. Commit your changes

   git commit -m "Add: feature description"

4. Push to your branch

   git push origin feature/your-feature-name

5. Open a Pull Request — Describe your changes and their purpose

---

Acknowledgements

This project builds upon the following open-source tools and platforms:

• Groq — Fast LLM inference
• LangChain — LLM application framework
• HuggingFace — Embedding models
• FAISS — Vector similarity search
• PyPDF2 — PDF processing

---

Author

Neha Maurya

Platform	Link
📧 Email	mauryaneha2006@gmail.com
🔗 LinkedIn	linkedin.com/in/neha-maurya-644a1a290

---

💼 Internship Opportunities

I'm actively seeking internship opportunities in AI, Machine Learning, and Full-Stack Development. If you feel my work on this project demonstrates a good fit for your team, I'd love to hear from you!

• 📧 Reach out: mauryaneha2006@gmail.com
• 🔗 Connect on LinkedIn: linkedin.com/in/neha-maurya-644a1a290

I'm eager to contribute, learn, and grow — if I'd be a good fit, let me in! 🚀

---

Built with ❤️ for academic research workflows.