1. Collected examples
These example files are all contained in a single .zip file, which you can download when you click here.
The Perceptron files are just input files. After downloading, you need to specify their location in the script (which you download from the Perceptron page to the left and open in R).
The Solver examples also include driver files, which you’ll open in R after downloading. You’ll need to specify the location of the input files, and of Solver.R (which you download from the HG in R page to the left, along with Tableau.R, which must reside in the same directory as Solver.R).
Some comments on the examples
aave.txt, chicano.txt: A simple case of phonologically conditioned variation – two patterns of t/d-deletion. See Coetzee and Pater (2011) on the data underlying this example, and for an introduction to the on-line approach to learning implemented in Perceptron.
anttila-gj.txt: A much-discussed case of phonological variation in the OT/HG literature. The data for the input files were adapted from a Praat file posted by Paul Boersma here, collapsing the equivalent cases as discussed in Goldwater and Johnson (2003). In Solver.R, this case serves as a replication of Goldwater and Johnson (2003), whose batch learning method is essentially the one it implements. See this handout for a gentle introduction to this method and to L1 and L2 priors, as well as an explanation of the “SGA” vs. “Perceptron” variants of on-line MaxEnt learning. The blog mentioned in the handout is the LSA course blog here, which has other related materials and handouts.
batama.txt: A much-discussed simple hidden structure problem in phonology (originally from Tesar and Smolensky 2000: MIT Press). See this handout on the problem, and the solutions implemented in Perceptron and Solver (NB I’ve only included the input file for Solver).
english-syntax.txt, german-syntax.txt: A much-discussed simple hidden structure problem in syntax (originally from Gibson and Wexler 1994 and Fodor 1998). See these work notes on the problem and the Perceptron-implemented solution (it works in Solver too, but I haven’t included the input file).
head.txt: Another syntax input file that can be used to demonstrate some interesting results of agent-based learning as implemented in Perceptron – see this handout.
2. Files for Kager and Pater (2012)
Click here for a zip file for the Solver driver file and data file needed to run the simulations in Kager and Pater (2012).