Jockers will discuss his ongoing development and validation of the “Syuzhet” R software package that implements sentiment analysis as a way to track emotional valence in a narrative fiction. Jockers will show how the highs and lows of emotional language serve as a reliable proxy for plot movement, and he’ll describe how he validated Syuzhet’s output against a collection of 167,034 human coded sentences. He’ll end with a discussion of his attempt to use clustering and several different distance / similarity metrics to identify a finite set of archetypal plot shapes within a collection of 20,000+ novels.