Lecture Notes Generator

This project creates comprehensive lecture notes from various source materials using RAG (Retrieval-Augmented Generation) with Ollama.

Overview

The pipeline consists of five main stages:

1. Converting PPTX to PDF: Converts PowerPoint presentations to PDF format.
2. Audio Extraction: Extracts audio from video lectures.
3. Audio Transcription: Transcribes audio to text using OpenAI's Whisper model.
4. PDF Processing: Extracts text and images from PDFs with automatic caption generation.
5. RAG Notes Generation: Creates comprehensive, well-structured lecture notes using RAG with Ollama.

Requirements

• Python 3.8+ with required packages (listed in requirements.txt)
• LibreOffice (for PPTX to PDF conversion)
• FFmpeg (for audio extraction)
• Ollama with the mistral:latest model

NOTE: Works best on a pure Linux system.

Installation

1. Install the required Python packages:

pip install -r requirements.txt

2. Install external dependencies:

# On Debian/Ubuntu
sudo apt-get update
sudo apt-get install ffmpeg libreoffice

# On macOS
brew install ffmpeg libreoffice

3. Install Ollama from https://ollama.ai/ and pull the Mistral model:

ollama pull mistral:latest

Usage

End-to-End Pipeline

Run the entire pipeline with:

python scripts/generate_all_notes.py

This will:

• Convert all PPTX files to PDF
• Extract audio from all video files
• Transcribe all audio files
• Process all PDF files
• Generate lecture notes for all topics

Process Specific Topics

To process only specific topics:

python scripts/generate_all_notes.py --topics BST binary_tree_traversal

Skip Specific Steps

You can skip specific pipeline steps:

# Skip PPTX conversion
python scripts/generate_all_notes.py --skip-pptx

# Skip audio extraction and transcription
python scripts/generate_all_notes.py --skip-audio

# Skip PDF processing
python scripts/generate_all_notes.py --skip-pdf

# Only run the notes generator
python scripts/generate_all_notes.py --only-notes

Pipeline Components

The project consists of the following main components:

• pptx_to_pdf.py: Converts PPTX files to PDF
• audio.py: Extracts audio from video files
• transcribe.py: Transcribes audio files to text
• extraction.py: Extracts content from PDF files
• simple_rag/: Contains RAG-related components
  • vector_store.py: TF-IDF based vector store
  • rag_application.py: RAG query system using Ollama
  • lecture_notes_generator.py: Generates lecture notes

Output

The generated notes will be saved in:

• Markdown: simple_rag/lecture_notes/markdown/
• PDF: simple_rag/lecture_notes/

Data Structure

Place your lecture materials in LLM_DATASET/ with one folder per topic:

LLM_DATASET/
└── topic_name/
    ├── presentation.pptx
    ├── presentation.pdf
    ├── lecture.mp4
    ├── mp3/
    │   └── lecture.mp3
    ├── lecture.mp3.txt
    ├── presentation_gen.txt
    ├── presentation_captions.csv
    └── images/
        └── img_1_1.png

Troubleshooting

• Ollama Connection: Ensure the Ollama service is running on http://localhost:11434
• PDF Extraction: If PDF extraction fails, check that the PDF is not password-protected
• LibreOffice: If PPTX conversion fails, ensure LibreOffice is correctly installed