Nitfol: Converse with beasts in their own tongue
A CoreML-based Natural Language Parser trained to understand interactive fiction commands.
Gloth: Fold dough 83 times
Nitfol includes the gloth tool, which generates training data based on interactive fiction commands. It uses this to create a word tagger model for use with Nitfol.
Running swift run gloth from the Nitfol project root creates an Artifacts directory, containing the generated testing and training data, and the resulting word tagging model. As gloth generates the model, it logs training, validation and evaluation accuracy, and displays a confusion matrix.
When you're satisfied with the Gloth.mlmodel results, you can drag the compiled Gloth.mlmodelc into the Sources/Nitfol/Resources directory for use in Nitfol.