neulab / incremental_tree_edit Goto Github PK

Hi authors, your work is interesting. May I ask steps of applying this model to a different language (i.e., JavaScript)?

1. Create ASDL grammar on tranX?
2. Convert raw data (source code) to the ginhubedit dataset format?
3. Make edit sequence generation (could you please suggest how to do?)

Data Preprocessing and Sanity Check
Todo: when applying the model to a different programming language, how to check/debug the ASDL grammar implementation, the abstract syntactic tree preprocessing, and the ground-truth edit sequence generation?

Hi. I am trying to transfer your model to my Java dataset. It is not a easy work because of the difference between the grammar of C# and Java. I have already produced the ASDLTree with Field class modified.

Now I have difficulty in understanding the CSharpHypothesis class.

-What is its function?
-Why does the SubstitutionSystem use Hypothesis intead of CSharpHypothesis?

It will be a great help if you could answer these questions.

After I run "bash scripts/githubedits/train.sh source_data/githubedits/configs/graph2iteredit.seq_edit_encoder.20p.json"

and the err.log says

Could you please guide me how to fix this err?

Hi,
Your work is great. I want to reproduce your model on Java datasets. But the Syntax.xml and grammar.json file in asdl/lang/csharp is for Csharp. Are there any public resources for Java?
Thank you.