A Stratal OT Model (Kiparsky 2000), that finds pairwise ranking probabilities using Expectation Driven Learning (Jarosz 2015). The software is in the form of a .ipynb file (meant to be used with Google Colab) and example input files are also included (which are meant to be uploaded to Google Drive, prior to running the script in Colab).
Hyperparameters for the model are all in the notebook's second block of code. Currently, the software is designed to test the model on a very specific set of test data, but you can easily edit the words it's tested on by changing the values of TEST_DATUM and TEST_DATUM_OUTPUTS, which are also initialized in the second block of code.
Another way that the model is tested is by asking it to make forced choices, given a particular UR, over the course of learning. These forced choices are also particular to the simulations I've been using it for, so if you need to change them, just edit the dictionaries in the forcedChoiceByDatum function.
- Jarosz, G. (2015). Expectation driven learning of phonology. Ms., University of Massachusetts Amherst.
- Kiparsky, P. (2000). Opacity and cyclicity. The linguistic review, 17(2-4), 351-366.