Automatic generation of natural language summaries for Java classes

By Laura Moreno, Jairo Aponte, G. Sridhara, Andrian Marcus, Lori L. Pollock, K. Vijay-Shanker. IEEE International Conference on Program Comprehension, 2013

369 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Information Systems

View Paper: https://www.semanticscholar.org/paper/47bdc0c4eedcb226c0c5f88471b69c5a1b67dfba

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Ge Li
2. Charles Sutton
3. Zhi Jin
4. Miltiadis Allamanis
5. Huan Sun
6. Ziyu Yao
7. Alexander LeClair
8. Collin McMillan

In Information Systems

1. Ge Li
2. Charles Sutton
3. Miltiadis Allamanis
4. Zhi Jin
5. Denys Poshyvanyk
6. Huan Sun
7. Ziyu Yao
8. Collin McMillan

Citations in the corpus

• Learning to Generate Pseudo-Code from Source Code Using Statistical Machine Translation (T) (2015, neutral): “1 is one example of the type of parallel corpus targeted in this study, in which have one-by-one correspondences between each line in the source code and pseudo-code.”
• Autofolding for Source Code Summarization (2014, methodology): “Work in this area includes synthesis of API usage examples [37], extraction of API usage patterns [38], [39], and generation of natural language summaries for source code [12], [13].”
• RepoAgent: An LLM-Powered Open-Source Framework for Repository-level Code Documentation Generation (2024, neutral): “Early methods were rule-based or template-driven (Haiduc et al., 2010; Sridhara et al., 2010; Moreno et al., 2013; Rodeghero et al., 2014).”
• Deep Code Comment Generation (2018, methodology): “[25] defined heuristics and stereotypes to synthesize comments for Java classes.”
• Improved Code Summarization via a Graph Neural Network (2020, methodology): “ation” was coined around ten years ago [19] and since that time the •eld has proliferated. At •rst, the dominant strategy was based on sentence templates and heuristics derived from empirical studies [15, 36, 40, 44, 50, 51]. Starting around 2016, data-driven strategies based on neural networks came”
• Measuring Program Comprehension: A Large-Scale Field Study with Professionals (2018, background): “the automated generation of comments [38], [54], [55], [61], and documents [25], [35].”
• CoaCor: Code Annotation for Code Retrieval with Reinforcement Learning (2019, methodology): “Earlier works tackled the problem using template-based approaches [37, 49] and topic n-grams models [38] while the recent techniques [2, 19–22, 31] are mostly built upon deep neural networks.”
• A Neural Model for Generating Natural Language Summaries of Program Subroutines (2019, background): “…code summa-rization” [9], with significant effort focused on generation of summaries of subroutines: For several years, significant progress was made based on content selection and sentence templates [1], [10]–[14] or even somewhat-idiosyncratic solutions such as mimicking human eye movements [15].”
• RCLinker: Automated Linking of Issue Reports and Commits Leveraging Rich Contextual Information (2015, positive): “ChangeScribe [15], [36] is an approach for automatic generation of commit messages, which combines code summarization techniques [25], [45], stereotypes detection [16], [17], [44], and impact analysis.”