In our paper, we describe three systems and the code for those is here. These draw on code from the repo for the task at: https://github.com/AmericasNLP/americasnlp2024/.
- Transformer.
doit
: batch script to run code and run evaluationanlp.py
: python code for network architecture and training
- Edit trees as transductions
dotest
: batch script to generate and evaluate resultsdolang.py
: python code to run experimentedittree.py
: edit tree implementation, invoked bydolang.py
- Edit trees with morphosyntactic similarity
newmytest
: batch script to generate and evaluate resultsnewmhtest.py
: python code for baseline, augmented by applying similar rules