mfuery/markov

Markov Chain Text Generator: Dad Joke Edition

Prompt / Motivation

Markov Chain implementation at learning to generate dad jokes using icanhazdadjoke.com

Dev Setup

Python Setup

1. Install Python 3.7:

brew install python

2. Install virtualenv:

pip install virtualenv

3. Install virtualenvwrapper:

pip install virtualenvwrapper

4. Add these to your ~/.bash_profile. This will cause virtualenv to create all your virtual envs in your home directory.

export VIRTUALENVWRAPPER_PYTHON=/usr/local/bin/python
export VIRTUALENVWRAPPER_VIRTUALENV_ARGS='--no-site-packages'
export WORKON_HOME=$HOME/.venv
source /usr/local/bin/virtualenvwrapper.sh

5. Setup virtual environment.

mkvirtualenv dadjokes

6. Install requirements.txt for applicable environment.

pip install -r requirements/dev.txt

Database

Dev environment is currently set up to use SQLite by default.

Migrate the database.

./manage.py migrate

Data sources are data driven (hear an echo?). So load the initial data into the database:

./manage.py loaddata database.json

Local Server

This serves the django app locally.

./manage.py runserver

Command-line Test

Run this to test out the API download, Markov Chain, and joke generation, all at once.

./manage.py fetch_jokes 1000

Webpack and React.js

Install dependencies and start the webpack/React dev server.

cd src/frontend/
npm start

Next Steps

• Train the Markov model using fitness functions.
• Convert Markov Chain build step to use matrices.
• Parallelize the Markov Chain build process to use multiple threads.
• Better caching.
• More data sources.