Skip to content

A (soon-to-be) nlp tool for seeing how obnoxious a writer you are.

License

Notifications You must be signed in to change notification settings

dvfeinblum/lexicount

Repository files navigation

Lexicount

This is a little tool I wrote to see how obnoxious a writer I am.

Getting Started

Setup

The backend for this tool is a simple postgres database. For ease of use, you can set it up using docker:

cd sqitch/
docker-compose up

Also, there's a redis that's used for caching stuff. That's also nice and easy to set up:

docker pull redis
docker run -d -p 6379:6379 redis

Now that you have the db, you're actually ready to go. Set up a venv and install requirements with:

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

NOTE: NLTK requires various downloads to work properly. Once you've finished installing the requirements, you should open a python REPL and run the following:

>>> import nltk
>>> nltk.download('averaged_perceptron_tagger')

Running the thing

Once you get that working, you're basically all set. Simply run:

python src/run.py

and it should get going on its own.

Oh! Also. If you would like to crawl along your own blog, there are a couple things you have to do:

  1. Make sure you have a blogpost.com blog. Currently, there's some hardcoded stuff that assumes you'll be using a feed from one of those.
  2. Set this envvar:
BLOG_FEED_URL="http://[YOUR USER NAME].blogspot.com/feeds/posts/default"

and that's it! The script will do the rest.

Pre-Commit Hooks

In order to keep my code squeaky clean, this repo features a pre-commit hook that runs autopep8. The formatter needs to be initialized after creating the venv. This can be done by running

pre-commit install