Leveraging Class Hierarchy for Code Comprehension

Abstract

Object-oriented programming languages enable a hierarchical class structure, which provides rich contextual information to guide code comprehension and synthesis. In this work, we propose the novel task of generating comments for overriding methods to facilitate code comprehension. To address this task, we formulate a deep learning framework which (1) exploits context from the comments of overridden methods and class names; (2) learns to generate comments in overriding methods that are more specific than those in the overridden methods; and (3) ensures that the generated comments are compatible with comments of overridden methods.

Cite

Text

Zhang et al. "Leveraging Class Hierarchy for Code Comprehension." NeurIPS 2020 Workshops: CAP, 2020.

Markdown

[Zhang et al. "Leveraging Class Hierarchy for Code Comprehension." NeurIPS 2020 Workshops: CAP, 2020.](https://mlanthology.org/neuripsw/2020/zhang2020neuripsw-leveraging/)

BibTeX

@inproceedings{zhang2020neuripsw-leveraging,
  title     = {{Leveraging Class Hierarchy for Code Comprehension}},
  author    = {Zhang, Jiyang and Panthaplackel, Sheena and Nie, Pengyu and Li, Junyi Jessy and Mooney, Ray and Gligoric, Milos},
  booktitle = {NeurIPS 2020 Workshops: CAP},
  year      = {2020},
  url       = {https://mlanthology.org/neuripsw/2020/zhang2020neuripsw-leveraging/}
}