Info

Segment Summary:

About the Project

Hi, my name is Thomas, welcome to my books project! This work attempts to visualize something many of us have felt or experienced - the ways in which a book might make us feel. How do great authors use cadence of emotions to make you laugh and then cry, and are there universal patterns for what "great" writing looks like?

To study this, I borrowed some recent advancements in the computer science world of Natural Language Processing. Facebook's large language model BART lies at the core of this project for two primary tasks.

By plotting these points out and examining a moving average of their behavior, we can get a sense of how concepts are growing and evolving through the course of the book. Details on the implementation and the full code is publicly available for your convenience at this GitHub repository.

Contact Information

For more information, connect with me on social media:

Mobile

Thanks so much for visiting my site! This project is not meant to be viewed on mobile devices. For the full experience please view on a desktop or laptop computer.

About the Project

This work attempts to visualize something many of us have felt or experienced - the ways in which a book might make us feel. To study this, I borrowed some recent advancements in the computer science world of Natural Language Processing. Facebook's large language model BART lies at the core of this project for two primary tasks. By plotting these points out and examining a moving average of their behavior, we can get a sense of how concepts are growing and evolving through the course of the book. As an example, the chart below shows the progression of the concept of "oppression" through the course of George Orwell's 1984.

Oppression
The image above is chaotic, with a number of sharp hills and valleys in the plot. By applying a moving average to the data, we can smooth out the plot and get a better sense of the overall trend.
Oppression Smooth
This image shows the same data as the first but with a moving average applied. The resulting plot is smoother, highlighting the overall trends more clearly and reducing the impact of short-term fluctuations.

Details on the implementation and the full code is publicly available for your convenience at this GitHub repository.

Contact Information

For more information, connect with me on social media:

LinkedIn LinkedIn Substack Substack