Not an image classifier - an NLP project for Reddit's Am I The Asshole subreddit.

• create data set from reddit posts
  • pull posts from API
  • label the posts by using the comments as votes (rules-based keyword approach)
  • evaluate approach on small subset of data
• create an AssholeClassifier
  • document classification task: given the post, classify as asshole or not
• create a reddit bot that comments with the prediction
  • create a feedback loop for data annotation
• I don't know, maybe get published?

preprocess.py takes the raw data and performs any necessary preprocessing. Outputs are found in data/processed.

train.py is the script used to train the model on the data

inference.py is used to load the model and classify new instances.

models holds all the model architectures we've tried.

1. create a virtual environment.

   python3 -m venv venv_classhole
   source venv_classhole/bin/activate
   

2. activate your environment. On mac, use

   source venv_classhole/bin/activate
   

   On Windows, use

   venv_classhole\Scripts\activate
   

3. Install dependencies

   pip install -r requirements.txt