Audio Response Evaluator

This project evaluates spoken responses from users based on several linguistic parameters including:

• Fluency
• Vocabulary
• Grammar
• Topic Relevance

It uses OpenAI's Whisper for transcription, spaCy for NLP processing, and FastAPI to expose the functionality via a simple API.

---

🔧 Features

• 🎙 Upload audio responses (e.g., .mp3, .mp4,.wav)
• 🧠 Automatic transcription using Whisper
• ✍️ Evaluate linguistic features (fluency, vocabulary, etc.)
• 🚀 FastAPI-based backend for scalable deployment
• 📊 Returns structured JSON scores per parameter

---

📁 Project Structure

audio_abex/
├── audio_evaluator.py       # Core logic to analyze transcription
├── main.py                  # FastAPI server
├── utils/                   # Helper functions and tools
├── venv/                    # Virtual environment
└── requirements.txt         # Python dependencies

---

🚀 Getting Started

1. Clone the repository

git clone https://github.com/vishalgoyal316/analyze_audio.git
cd analyze_audio

2. Set up a virtual environment (Python 3.11 recommended)

python3 -m venv venv
source venv/bin/activate

3. Install dependencies

pip install -r requirements.txt
python -m spacy download en_core_web_sm

4. Run the server

uvicorn main:app --reload

Visit http://localhost:8000/docs for Swagger API documentation.

---

📤 API Usage

POST /analyze-audio

Request:

• audio_file: form-data file upload (audio format)

Response:

{
  "fluency": 8.5,
  "grammar": 9.0,
  "vocabulary": 7.8,
  "relevance": 8.2,
  "transcript": "This is the spoken content."
}

---

🧠 Technologies Used

• FastAPI
• OpenAI Whisper
• spaCy
• PyTorch

---

📜 License

MIT License. See LICENSE file for details.

---

🙋‍♂️ Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you'd like to change.