Deep Disfluency Server

This project exposes the deep_disfluency tagger - a deep learning tagger for speech disfluencies - as a TCP server.

Connect to the server, wait for a "ready" message, and start sending raw audio. The server transcribes the audio with the help of IBM Watson, and sends tagged words in response. For more information about the tagging check the deep_disfluency package.

Installation

• Obtain a credentials.json file from IBM Watson as described here, and put it in the project directory.
• Create a python 2 virtual environment (sorry for misbehaving) with virtualenv --python python2 env and activate it with source env/bin/activate.
• Install dependencies with pip install -r requirements.txt. Suggestion: use pip-tools instead.
• python server.py

client.py

The client.py script is a full demo client for the server, and a good reference for implementing your own clients. Check it out!

Configuration

The entire configuration is hard coded in server.py ATM. Change the values there for your needs, or send a pull request with a better configuration mechanism 😉