“What Our Favorites Can Tell Us:  A Computer-Directed Reading, Topic Modeling and Trend Analysis Investigation into 2,000 New York Times Bestsellers and Genre Fiction Winners from 2005 to 2016”

1. Abstract

This study investigates what content and themes appear “at scale” in recently published literature, specifically New York Times bestsellers and genre fiction winners between 2005 and 2016. Topic modelling, a form of natural language processing and statistical analysis that gleans “topics” or related word groups by looking at the probability that words occur in the same documents within a corpus, is applied in this study to investigate a dataset of over 2,000 novels at the sentence level. This investigation embodies a DH sense of "play" by co-creating with computer (in this case, a statistical model) to identify both familiar and "latent" topics. The aim of this content analysis is to identify significant themes that resonate with recent social and political issues like the depiction of gun violence across genres, representation of minority characters, health-care related content and in tracking how the content of popular texts follows trends across a decade.