A Fabula of Syuzhet II

Continuing the tale of digital humanities and sentiment analysis


  1. See Fabula of Syuzhet, the first of this Storify series, to read the earlier discussion.
  2. Timeline of further Syuzhet discussion
    March 24: Matthew Jockers offers a further response to Annie Swafford, A Ringing Endorsement of Smoothing.
    March 30: Annie Swafford publishes Why Syuzhet Doesn’t Work and How We Know, underlining her thoughts on Syuzhet and responding to Andrew Pipers’ points about validation.
    April 1: Jockers publishes My Sentiments (Exactly?) which describes "gut-checking" the sentiment values returned by machine methods and comparing types of tagging.
    April 6: Jockers publishes Requiem for a Low Pass Filter, conceding errors in Syuzhet caused by the low pass filter.
  3. Distant reading,the use of computers to discern patterns in enormous numbers of texts, has been lampooned inside and outside of academia. But digital humanists continue to forge experiments to limn huge bodies of texts as data and to publish their results.
  4. In February 2015, Matthew Jockers, Associate Professor of English at the University of Nebraska, Lincoln and Director of the Nebraska Literary Lab, announced Syuzhet, his new software package for the sentiment analysis of literary texts. Jockers incorporated a mathematical tool used by physicists (the Fourier transform) into Syuzhet to compare the "plot shapes" of 40,000 novels and found six common, perhaps archetypal, plot shapes. The appearance of Syuzhet and Jockers' claim to have discovered that novels have a small, finite number of plots was noted in the academic community, among data scientists and commercial practitioners of sentiment analysis, and in the news media. In March, Annie Swafford, Assistant Professor for Interdisciplinary and Digital Teaching and Scholarship at the State University of New York, New Paltz, published a critique of the Syuzhet package. Swafford's critique touched off a far-ranging discussion across academic disciplines of the value of text mining tools for literary research. This Storify, and the earlier one in the series, attempts to gather representative elements from the discussion.
  5. To better follow the discussion, reading the comments on the blog posts, some of which contain new voices and perspectives, is recommended.
  6. (Selected tweets in roughly chronological order; some grouped for clarity.)

  7. Jockers publishes a blog post that goes toe to toe with Swafford's critique, explaining that Syuzhet is not designed to provide a perfect mapping of emotional valence. He promises another post to identify an “ideal” number of components for the low pass filter.
  8. A new Storify appears on the scene in which Heather Froehlich, a corpus linguistics Ph.D. student at the University of Strathclyde, voices her concerns about the lack of understanding of lexical ambiguity within computational approaches to literary scholarship.
  9. Programmer and mathematician Daniel Lepage takes Jockers to task about his method of “smoothing.”