Experiments in Multivariant Natural Language Processing for Nahuatl

Code accompanying our Vardial paper, "Experiments in Multivariant Natural Language Processing for Nahuatl."

• Virtually all of the code is for preparing the datasets, since the experiments are run with Machamp via config jsons (TODO: add all experiment configs).
• If you have the Western Sierra and Highland Puebla Nahuatl UD treebanks downloaded, just update the paths in data_creation_script.sh to point to them, and run that.