Research software

This is code I've written to obtain results in my publications.

• Dependency-Based Compositional Semantics (DCS): Java/Scala implementation of the Learning Dependency-Based Compositional Semantics paper.
  [Download release 1.0]

• Unsupervised NLP modeling toolkit: Java/Scala code for unsupervised modeling. Papers that use this code: Online EM for Unsupervised Models and Learning Semantic Correspondences with Less Supervision.
  [Download release 1.1]

• Cross-EM word aligner: Java implementation of the Alignment by Agreement paper.
  [Download release 1.3]

• Perceptron learning for sequences: C++ implementation used in my MEng thesis.
  Supports Markov (tagging) and semi-Markov (segmenting) models.
  [Download release 1.1]

• Word clustering: C++ implementation of the Brown algorithm used in my MEng thesis.
  [GitHub]

• Hypertree and windmill farms: C++ implementation of an algorithm for learning bounded tree-width Markov networks, a generalized Union-Find data structure for maintaining hyperforests, and a linear program for finding the worst dynamic program settings.
  [Download here]

Productivity software

This is code I've written to make research more productive by automating the tedious things.

• fig: a general-purpose collection of Java libraries and tools to faciliate writing research code and managing experiments. Features include easy parsing of command-line options and hierarchical logging.
  [GitHub]

• rfig: a package for creating figures and slide presentations in Ruby (via Metapost).
  All my talk slides from 2006-2012 are created using this package (an example of the final product). Note: I am switching from rfig to sfig.
  [GitHub]

• sfig: a package for creating figures and slide presentations in Javascript/SVG (with option to output Metapost like rfig). As the successor to rfig, sfig is cleaner and allows for interactivity. See the sfig tutorial.
  [GitHub]