一句话：提出基于 NLP 的方法，仅利用提交描述自动判断提交能否集成到软件发布，准确率达 92.9%。
持续集成是全球软件项目维护的关键环节，审批流程、冲突解决、测试与验证会累积较长的解决时间，其中验证过程涉及对提交描述的理解。本文提出一种基于自然语言处理（NLP）的智能方法，仅根据提交描述评估该提交是否可以集成到特定软件发布中，从而支持提交集成决策的自动化。实验结果表明，该方法的准确率达到 92.9%。
Continuous Integration is a critical problem for software maintenance in global projects, compromising companies’ performance, which tends to accumulate a high-resolution time due to the approval process, conflict resolution, tests, and validations. The process of the validation involves the commit description interpretation and can be automated by NLP-mechanisms. This paper presents an intelligent NLP-based approach to evaluate whether the commits can be integrated into a certain software release based only on their descriptions. Our experiments showed an accuracy of 92.9%.
DOI 原文 ·
@inproceedings{paperbot6491,
title = {Automatically Deciding on the Integration of Commits Based on Their Descriptions},
author = {Samuel C. Fonseca and Mateus C. Lucena and Tiago M. Reis and Pedro F. Cabral and Walmir A. Silva and Flavia de S. Santos and Felipe T. Giuntini and Juliano Sales},
booktitle = {ASE},
year = {2021},
doi = {10.1109/ase51524.2021.9678906}
}