Scene Detection in Literary Fiction

Dennis Tenen

xpmethod.plaintext.in

Since our Last Time

Writing a book

Microscope!

Running a lab

Three research clusters:

  • theory / method
  • public discourse
  • minimal computing

Two projects of interest to this group:

  1. Emerging Consensus in Scientific Literature (w/ Dan Jurafsky, Dan MacFarland of Stanford, and Laura Kurgan of Spatial Information Lab)
  1. Scene Detection in Literary fiction (w/ Kathy Mckewan)

Scene Detection

Original Paper

  1. Raymond Williams
  1. Bakhtin

Suggested improvement: networks over time

Complexity! Lack of models for theoretical primitives.

  • space
  • time
  • actors
  • relations

Chunking. Preliminary observations:

  • Typographical space (paragraph)
  • Diegetic space (scene)

How do the two relate?

What is a scene?

Burke, Kenneth. A Grammar of Motives. University of California Press, 1969.

Rasheed, Z., and M. Shah. “Scene Detection in Hollywood Movies and TV Shows.” In 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings, 2:II – 343–48 vol.2, 2003.

Complications:

  • actors can become scenes
  • scenes can become actors
  • scenes can involve mental states and ideologies

Proposed algorithm:

  1. Separate actors from background through grammatical categories (object, subject, transitive verb, A blanks B)

  2. Cluster scene particulates semantically to detect potential boundaries

  3. Multiple passes to detect quality / movement (by analogy with video).

Ground truth, a stumbling block as always. To overcome: expert tagging, parallel copora. Any ideas?

We need better primitives to build more complex models!