PyAnnote-Api-Toolkit

Goal

The PyAnnote-Api-Toolkit application is designed to help users identify audio recordings or segments where the pyannote diarization system struggles to achieve high confidence. The application allows users to upload an audio recording, which is then diarized. Users can visualize the diarization results through a diagram, listen to the audio, and provide an overall appreciation score. The tool also enables analysis through configurable thresholds, including:

• Mean score across all samples (combined with human scores).
• Turn-level confidence score thresholds.
• Sample-level confidence score thresholds.

Author

This project is developed by Alex CHOUX, who holds a Master's Degree in Computer Science, with a specialization in Data Science and AI for NLP.

---

Application Structure

The application consists of 6 distinct services, orchestrated using Docker Compose:

Service	Description
HTML Page	Dockerized using Nginx, serving as the client interface.
Python Server	Manages requests between the HTML page and other services.
Webhook + Ngrok	Handles diarization jobs and sends results to the server.
PyAnnote Wrapper	Interface with the PyAnnote API.
Mongo Gateway	Interface for interacting with the MongoDB database.
MongoDB	Database for storing application data.

Note: Only the webhook is exposed to the internet via Ngrok. The HTML page is the only other service with an accessible port for user interaction.

---

Prerequisites

To run the application, you need:

• Docker Compose (to build and launch the services).
• A .env file containing the required environment variables.

.env File Structure

NGROK_SUBDOMAIN=your_ngrok_subdomain
NGROK_AUTH_TOKEN=your_ngrok_auth_token
TOKEN_PYANNOTE=your_pyannote_api_token
PYANNOTEAI_WEBHOOK_SIGNING_SECRET=your_pyannote_webhook_signing_secret

---

Forwards

Right now, the application is not designed to use speaker identification. It would be the further step in the development of the application. Moreover, some minor considerations must be addressed, such as:

• The expiration management of the uploaded audio files.
• The job_id expiration.
• Allowing the user to use recordings stored on a web storage like S3.