基于深度学习的代码审查意见有效性评估

doi:10.16180/j.cnki.issn1007-7820.2020.01.008

摘要/Abstract

摘要：

针对代码审查过程中的代码审查意见对于开发者可能无价值的问题,文中提出了一种基于深度学习长短期记忆网络的代码审查意见有效性评估方法。该方法通过提取代码审查意见中与审查意见有效性相关的特征,并根据这些特征构建评估模型,从而评估审查意见对于开发人员是否有价值。为了验证方法的有效性,文中选取了GitHub上开源Eclipse项目中的审查信息作为实验数据,并将所提方法与其它机器学习方法对比。实验结果表明,该方法可以有效评估审查意见的价值。

关键词: 软件工程, 代码审查, 深度学习, 长短期记忆模型, 词向量, 自然语言处理。

Abstract:

Aiming at the code review comments during the code review process may be of no value to developers, a code review comments evaluation method based on deep learning Long Short-Term Memory networks was proposed, which can effectively extract the features related to validity of the code review comments, then a code review comments evaluation model based on these features was built to judge whether the review comments are useful to the developer. In order to verify the method, the extensive experiments were conducted based on the open source Eclipse project on GitHub as experimental data, and this paper compared the method with other machine learning methods. The experimental results demonstrated that the method effectively evaluate whether the review comment is meaningful and significantly better than the comparison methods.

Key words: software engineering, code review, deep learning, long short-term memory, word embedding, natural language processing

中图分类号:

TP391.1

段雨佳,鞠婷. 基于深度学习的代码审查意见有效性评估[J]. 电子科技, 2020, 33(1): 39-45.

DUAN Yujia,JU Ting. Evaluation of Code Review Comments Based on Deep Learning[J]. Electronic Science and Technology, 2020, 33(1): 39-45.

图/表 8

图1

图2

图3

图4

图5

表1

图6

表2

参考文献 20

[1]	Host M, Johansson C . Evaluation of code review methods through interviews and experimentation[J]. Journal of Systems and Software, 2000,52(2-3):113-120.
[2]	Bacchelli A, Bird C. Expectations, outcomes, and challenges of modern code review[C]. San Francisco: International Conference on Software Engineering, 2013.
[3]	Rong G, Li J, Xie M, et al. The effect of checklist in code review for inexperienced students: an empirical study[C] .Nanjing:IEEE Conference on Software Engineering Education & Training,IEEE, 2012.
[4]	Mcgraw G . Automated code review tools for security[J]. Computer, 2008,41(12):108-111.
[5]	Remillard J . Source code review systems[J]. IEEE Software, 2005,22(1):74-77.
[6]	Bosu A, Greiler M, Bird C. Characteristics of useful code reviews:an empirical study at Microsoft[C]. Florence: Mining Software Repositories, 2015.
[7]	Greff K, Srivastava R K, Koutník J , et al. LSTM: a search space odyssey[J]. IEEE Transactions on Neural Networks and Learning Systems, 2017,28(10):2222-2232.
[8]	Mcintosh S, Kamei Y, Adams B , et al. An empirical study of the impact of modern code review practices on software quality[J]. Empirical Software Engineering, 2016,21(5):2146-2189.
[9]	Thongtanunam P, Kula R G, Cruz A E C, et al. Improving code review effectiveness through reviewer recommendations[C]. Hyderabad: International Workshop on Cooperative and Human Aspects of Software Engineering,ACM, 2014.
[10]	Rahman M M, Roy C K, Kula R G. Predicting usefulness of code review comments using textual features and developer experience[C]. Buenos Aires:International Conference on Mining Software Repositories,IEEE, 2017.
[11]	胡星, 王千祥 . 深度学习在缺陷修复者推荐中的应用[J]. 计算机科学与探索, 2017,11(5):700-707.
	Hu Xing, Wang Qianxiang . Application of deep learning in recommendation of bug reports assignment[J]. Journal of Frontiers of Computer Science and Technology, 2017,11(5):700-707.
[12]	Cherry J M, Adler C, Ball C , et al. SGD:saccharomyces genome database[J]. Nucleic Acids Research, 1998,26(1):73-79.
[13]	Perozzi B, Al-Rfou’ R, Kulkarni V , et al. Inducing language networks from continuous space word representations[J]. Computer Science, 2014,549(1):261-273.
[14]	Lecun Y, Bengio Y, Hinton G . Deep learning[J]. Nature, 2015,521(7553):436-439.
[15]	Saul L K, Jaakkola T, Jordan M I . Mean field theory for sigmoid belief networks[J]. Journal of Artificial Intelligence Research, 1996,4(1):61-76.
[16]	Levy O, Goldberg Y . Neural word embedding as implicit matrix factorization[J]. Advances in Neural Information Processing Systems, 2014(3):2177-2185
[17]	Reimer, Paula J Baillie, , et al. INTCAL04 terrestrial radiocarbon age calibration,26-0 ka BP[J]. Radiocarbon, 2004,46(5):1029-1058.
[18]	Memisevic R, Zach C, Hinton G, et al. Gated softmax classification[C]. Vancouver:Advances in Neural Information Processing Systems, 2010.
[19]	Srivastava N, Hinton G, Krizhevsky A , et al. Dropout: a simple way to prevent neural networks from overfitting[J]. The Journal of Machine Learning Research, 2014,15(1):1929-1958.
[20]	Cassel M, Lima F. Evaluating one-hot encoding finite state machines for SEU reliability in SRAM-based FPGAs[C]. Lake of Como:IEEE International Symposium on On-Line Testing,IEEE, 2006.

	Bayes	SVM	RNN	LSTM
准确率	0.597	0.715	0.802	0.884
召回率	0.625	0.613	0.633	0.627
F₁ score	0.611	0.661	0.708	0.734