Simple character recommender for story writing. Built at Hack@Brown 2018.
- Training data comes from Nate the Snake
- Label is each character in the text
- Features are the previous
num_previouscharacters (defined inutils.py)
- Logistic regression model trained on data using one-hot encodings for the characters
- Flask server allows user to interact with the trained model
- Create a virtualenv (
virtualenv env) - Activate virtualenv (
source env/bin/activate) - Download dependencies (
pip install -r requirements.txt)
- Paste training data into
text.txt - Change paramters in
utils.py - Run
python fileparse.pyto generate the csv intraining_data.csv - Run
python train_model.pyto train and save model tomodel.pkl
- Run
python web.py - Go to
localhost:5000
- Webcrawl wikipedia for more training data