Learning Lenient Parsing & Typing via Indirect Supervision

Validation Set: We have shared 3 CSV files containing human-annotated validation sets of our paper.

AST and Code Correction: For generating AST and Code correction, please check the two files, AST.py and Top1.py.
In AST.py, we present the output of different parts of the program with an example. Please read that one before Top1.py.
We follow the implementation of https://github.com/Lsdefine/attention-is-all-you-need-keras. Please check the remaining code in the above link. We made a minor correction in the dataloader.py to use two separate vocabulary cutoff for input and output. dataloader1.py, transformer1.py, etc. are an exact replication of dataloader.py and transformer.py.
Since we are using two models, we did it to avoid any conflict.

TypeFix: Check the code in TypeFix.zip