Implementation of the ISSRE 2021 paper "Improving Code Summarization Through Automated Quality Assurance"
Python >=3.6
numpy
scipy
pickle
nltk==3.2.5
sklearn
six==1.11.0
rouge==1.0.0
git+https://github.com/Maluuba/nlg-eval.git@master
typing==3.6.2
1.Dataset
We use the dataset of Deepcom (https://github.com/xing-hu/EMSE-DeepCom)
2.Train your model and get the summaries generated by Deepcom,
Rencos (https://github.com/zhangj111/rencos),
NMT (https://github.com/OpenNMT/)
3.Use Ensum to the summaries.
python Ensum.py approach-name approach-name
For example:
python Ensum.py deepcom deepcom
4.Evaluate your results
nlg-eval --hypothesis=examples/hyp.txt --references=examples/ref.txt