This repository hosts code for a Twitter bot (@SimpsonsHaiku) that tweets haikus mined from episodes of The Simpsons.
This project was inspired by @nythaikus. The dataset used here is hosted on Kaggle here, and was originally scraped by Todd Schneider from the dearly departed simpsonsworld.com, covering the first 26 seasons. Usage and some exploratory data analysis can be found at haiku_demo.ipynb.
To run the project and generate your own haiku, you'll need to clone the repository in the usual way, and install the relevant packages/activate the conda environment shared in environment.yaml. Direct dependences are also specified in the requirements.txt file.
git clone https://github.com/mwestt/simpsons-haiku.git
conda env create -f environment.yaml
conda activate simpsons_haiku
We start by instantiating the core haiku object, of class SimpsonsHaiku from haiku.py and extracting a DataFrame of haikus. We can then call the generate_haiku method to sample a haiku along with associated metadata.
from haiku import SimpsonsHaiku
simpsons_haiku = SimpsonsHaiku()
simpsons_haiku.generate_haiku_df() # Saves haiku DataFrame to `haiku_df.csv` by default
haiku, metadata = simpsons_haiku.generate_haiku()Or you can pass in the path to a haiku_df or the DataFrame itself, and avoid waiting for the data wrangling, for example:
simpsons_haiku = SimpsonsHaiku('haiku_df.csv') # Faster, no data wrangling
haiku, metadata = simpsons_haiku.generate_haiku()>>> print(haiku)It doesn't take a
whiz to see that you're looking
out for number one