Now that I'm in the end stages of the Ph.D., I'm finally getting organized and making some effort to post some of my research code. I've attempted to remove deadline-induced sloppiness and add documentation, but it has not gone through processes of code review or other niceties for production-level code.

If you do end up using any of this software, please add the appropriate citations (listed in each section). I welcome feedback, though may not be able to respond to all support and feature requests.

Cascades, which are described in our SDM 2007 paper, are subgraphs (weakly connected components) of conversations, built from links. Since 2008 I have been building a package for processing them. I have cleaned up part of it and included that part here, in case someone wants to find isomorphic cascades and generate some cool-looking graphviz's from social media link data.

• Package of everything, including Java code, script, IO files, Readme, and resulting pdfs. [tar.gz]
• Automatically-generated images of a "Cascade Markov Model" [pdf][png] - Shows the typical evolution of cascades. For each cascade structure shown, p_stay indicates how often the cascade stopped at that size. Numbers along the edges between structures show relatively how often a cascade transitions into a larger cascade structure.
• Automatically-generated images of an authored CMM [pdf] [png] - Same as above, only where author isomorphism is also required. Each author in a single cascade is a different color.
• README
• Java files:
  • Cascade.java - Cascade data structure and isomorphisms.
  • BuildCascadesEfficient.java - Input link file, output cascade file.
  • CascadeIterator.java - Input cascde file, output Cascades one by one.
  • CascadeMarkovModel.java and LearnCascadeMarkovModel.java - Code that enumerates the cascades as they are built, and assigns transition probabilities to go from one cascade structure to the next larger one (see the PDFs).
  • CascadeUtils.java and CascadeViz.java - A few misc methods.
  • CascadeTest.java - A little JUnit test code-- in progress.

(Coming soon)

(Coming soon)