MeetNotes

MeetNotes is an AI-powered meeting assistant that captures live speech during meetings, generates a live transcript, extracts important key points using Machine Learning (TF-IDF + TextRank), produces an AI-generated meeting summary using the Groq LLM, and exports professional meeting notes as a PDF.

---

Features

• Live speech-to-text transcription using the Web Speech API
• Real-time transcript display
• Key point extraction using TF-IDF and TextRank
• AI-generated meeting summary using Groq (Llama 3.3 70B)
• PDF export of meeting notes
• Chrome Extension frontend
• FastAPI backend

---

System Architecture

                  User Speaks
                       │
                       ▼
        Browser Web Speech API
       (Live Speech Recognition)
                       │
                       ▼
            Live Meeting Transcript
                       │
             User Ends Meeting
                       │
                       ▼
        FastAPI Backend Receives Transcript
                       │
                       ▼
         Transcript Preprocessing
      (Noise Removal & Text Cleaning)
                       │
                       ▼
      TF-IDF + TextRank (Machine Learning)
        Extract Important Key Points
                       │
                       ▼
          Groq LLM (Llama 3.3 70B)
      Generate AI Meeting Summary
                       │
                       ▼
          Generate Professional PDF
                       │
                       ▼
           Download Meeting Notes

---

How It Works

Step 1 — Live Transcription

The user starts the Chrome extension and turns on the microphone.

The browser continuously converts speech into text using the Web Speech API.

The transcript is displayed live throughout the meeting.

---

Step 2 — End Meeting

When the meeting is completed, the user clicks End Meeting.

The complete transcript is sent to the FastAPI backend.

---

Step 3 — Machine Learning Key Point Extraction

The backend preprocesses the transcript by removing greetings, filler words, and conversational noise.

The cleaned transcript is then processed using:

• TF-IDF (Term Frequency–Inverse Document Frequency) to identify important terms.
• TextRank (PageRank) to rank the most informative parts of the meeting.

This produces the meeting's most important discussion points.

---

Step 4 — AI Summary Generation

The extracted key points are sent to the Groq API using the Llama 3.3 70B Large Language Model.

The AI generates a concise meeting summary while using only the extracted key points as context.

This combines traditional Machine Learning with Generative AI to produce accurate and readable meeting notes.

---

Step 5 — PDF Generation

The system combines:

• AI-generated Summary
• Machine Learning extracted Key Points

and generates a professional PDF that can be downloaded by the user.

---

Tech Stack

Frontend

• HTML
• CSS
• JavaScript
• Chrome Extension (Manifest V3)
• Web Speech API

Backend

• FastAPI
• Python

Machine Learning

• TF-IDF
• TextRank
• Scikit-learn
• NetworkX

AI

• Groq API
• Llama-3.3-70B-Versatile

PDF Generation

• FPDF

---

Project Structure

MeetNotes/
│
├── backend/
│   │
│   ├── app/
│   │   │
│   │   ├── routes/
│   │   │      api.py
│   │   │
│   │   ├── services/
│   │   │      extractor.py
│   │   │      summarizer.py
│   │   │      llm.py
│   │   │      pdf_service.py
│   │   │
│   │   └── main.py
│   │
│   ├── run.py
│   └── requirements.txt
│
├── frontend/
│   ├── index.html
│   ├── scriptjs
│   ├── styles.css
│   ├── manifest.json
│   
│
└── README.md

---

Processing Pipeline

Speech
   │
   ▼
Live Transcript
   │
   ▼
Preprocessing
(Remove Greetings & Noise)
   │
   ▼
TF-IDF Vectorization
   │
   ▼
Sentence Similarity Matrix
   │
   ▼
TextRank (PageRank)
   │
   ▼
Top Ranked Key Points
   │
   ▼
Groq LLM
(Llama-3.3-70B)
   │
   ▼
AI Meeting Summary
   │
   ▼
PDF Generation

---

API Endpoints

Process Meeting Transcript

Endpoint

POST /process_text

Request

{
    "text": "Complete meeting transcript..."
}

Response

{
    "summary": "AI-generated meeting summary...",
    "bullets": [
        "Project deployment timeline discussed",
        "Budget approved",
        "Testing begins next Monday",
        "Customer feedback after deployment"
    ]
}

---

Generate PDF

Endpoint

POST /generate-pdf

Request

{
    "summary": "Meeting summary...",
    "bullets": [
        "Point 1",
        "Point 2",
        "Point 3"
    ]
}

Returns

MeetingNotes.pdf

---

Installation

Clone Repository

git clone https://github.com/diyaaghosh/MeetNotes.git
cd MeetNotes

---

Install Backend Dependencies

cd backend
pip install -r requirements.txt

---

Create Environment File

Create a .env file.

GROQ_API_KEY=your_groq_api_key

---

Run Backend

python run.py

Backend runs at

http://localhost:10000

---

Load Chrome Extension

1. Open Chrome.
2. Navigate to:

chrome://extensions

3. Enable Developer Mode.
4. Click Load unpacked.
5. Select the frontend folder.

---

Sample Output

Key Points

• Project deployment timeline discussed.
• Budget approved after cost review.
• Testing begins next Monday.
• Customer feedback collected after deployment.
• Review meeting scheduled for Friday.

---

AI Summary

The meeting focused on the project deployment timeline and overall project progress. The team approved the project budget after reviewing the estimated costs. Testing is scheduled to begin next Monday, followed by customer feedback collection after deployment. The next review meeting has been scheduled for Friday to assess progress and discuss any remaining issues.

---

Future Enhancements

• Speaker Identification
• Automatic Action Item Detection
• Meeting Title Generation
• Multi-language Support
• Calendar Integration
• Email Meeting Notes
• Semantic Keyword Extraction using Transformer Embeddings
• Sentiment Analysis
• Meeting History Dashboard
• Cloud Storage Integration

---

Technologies Used

Category	Technology
Frontend	HTML, CSS, JavaScript
Speech Recognition	Web Speech API
Backend	FastAPI
Machine Learning	TF-IDF, TextRank
AI	Groq (Llama-3.3-70B-Versatile)
PDF	FPDF
Language	Python

---

Project Summary

MeetNotes AI combines Machine Learning and Generative AI to automate meeting documentation.

• TF-IDF + TextRank identify the most important discussion points from the transcript.
• Groq's Llama 3.3 70B transforms those key points into a concise, professional meeting summary.
• The generated summary and key points are exported into a structured PDF, allowing users to focus on discussions instead of taking manual notes.