Exploring the shapes of stories using indico sentiment analysis APIs. Implements a couple hacks to enable sentiment models to work reasonably well across long context windows Sentiment analysis is difficult to implement on long stories. This repo implements a couple hacks, to demonstrate how the indico sentiment API can be used to pick up long-range emotional landscape across long stories and movie scripts. We think Kurt Vonnegut would be impressed!
Navigate to a good place on your filesystem, then:
git clone https://github.com/IndicoDataSolutions/plotlines
If you don't have ipython/Jupyter installed yet, then install it:
pip install -U ipython
Launch the notebook server. It will bring up a browser window.
ipython notebook
Click on plotlines.ipynb and walk through the code. Either use the menu Cell > Run all to run everything at once, or use Shift + Enter to execute one cell at a time.
Follow the links in the notebook to get an indico API key, install any necessary modules, and find input data (e.g., movie scripts).