Page 141 - 《软件学报》2025年第9期

4052 软件学报 2025 年第 36 卷第 9 期

同时, 设计了一种 GCF 方法, 通过融合 bug 报告文本语义和 bug-开发者二部图两个模态数据重构节点特征, 并基
于二部图结构捕获 bug-开发者之间的相关性. 最终, 将 bug 分派任务建模为 GCF 模块上的链接预测, 并根据预测
结果得到了 bug 分派的推荐方案. 在 GC、MC 及 MF 这 3 个公开数据集上进行的大量实验分析表明, CBT-MF
对 bug 分派的性能较好, 整体表现出了明显的优越性. 该方法有望为软件 bug 分派开辟新的研究思路, 提高 bug 分
派的准确性和效率.
CBT-MF 还有许多值得扩展的工作. 首先, 需要继续深入探究 bug-开发者相关性的有效表征, 并研究应用其他
聚合器, 如最大池化、LSTM 等, 以提升 CBT-MF 方法的性能. 其次, 进一步研究自适应的数据增强方案, 通过基于
自适应的数据增强方案对不均衡 bug 数据进行优化和调整, 以更好地适应不同分布和特征的数据. 第三, 不断探
索 CBT-MF 在其他软件 bug 分派场景应用的可行性, 以提高 CBT-MF 方法的表达能力和泛化性. 最后, 计划探索
处理重复报告和动态属性变化等问题的有效方法, 以提升 CBT-MF 的推广性和实用性. 此外, 在 CBT-MF 中, 主要
强调 bug 和开发者文本表示向量和二部图数据的融合及特征提取, 因此在 bug 分派时选取了简单的内积交互函数
进行相关性预测. 事实上, 还有其他更复杂的选择, 例如基于神经网络的交互函数, 其不仅可以通过向量传播层来
丰富初始向量, 而且还允许通过参数调整来控制向量传播的范围. 因此, 探索更有效的分派预测方法也是未来研究
工作的一部分.

References:
[1] Zhang TL, Chen R, Yang X, Zhu HY. Approach of bug reports classification based on cost extreme learning machine. Ruan Jian Xue
Bao/Journal of Software, 2019, 30(5): 1386–1406 (in Chinese with English abstract). http://www.jos.org.cn/1000-9825/5725.htm [doi: 10.
13328/j.cnki.jos.005725]
[2] Sun JM, Xing ZC, Xia X, Lu QH, Xu XW, Zhu LM. Aspect-level information discrepancies across heterogeneous vulnerability reports:
Severity, types and detection methods. ACM Trans. on Software Engineering and Methodology, 2023, 33(2): 49. [doi: 10.1145/3624734]
[3] Banerjee S, Syed Z, Helmick J, Culp M, Ryan K, Cukic B. Automated triaging of very large bug repositories. Information and Software
Technology, 2017, 89: 1–13. [doi: 10.1016/j.infsof.2016.09.006]
[4] Binh NT, Tinh PD, Chien NH. BKiosk system—A solution for context-specific services supporting contextualization stage of E-
government systems. In: Proc. of the 6th Int’l Conf. on E-commerce, E-business and E-government. Plymouth: ACM, 2022. 403–411.
[doi: 10.1145/3537693.3537716]
[5] Anvik J, Hiew L, Murphy GC. Who should fix this bug? In: Proc. of the 28th Int’l Conf. on Software Engineering. Shanghai: ACM, 2006.
361–370. [doi: 10.1145/1134285.1134336]
[6] Sharma M, Tandon A, Kumari M, Singh VB. Reduction of redundant rules in association rule mining-based bug assignment. Int’l Journal
of Reliability, Quality and Safety Engineering, 2017, 24(6): 1740005. [doi: 10.1142/S0218539317400058]
[7] Anvik J, Murphy GC. Reducing the effort of bug report triage: Recommenders for development-oriented decisions. ACM Trans. on
Software Engineering and Methodology (TOSEM), 2011, 20(3): 10. [doi: 10.1145/2000791.2000794]
[8] Xuan JF, Jiang H, Ren ZL, Zou WQ. Developer prioritization in bug repositories. In: Proc. of the 34th Int’l Conf. on Software
Engineering (ICSE). Zurich: IEEE, 2012. 25–35. [doi: 10.1109/ICSE.2012.6227209]
[9] Jonsson L, Borg M, Broman D, Sandahl K, Eldh S, Runeson P. Automated bug assignment: Ensemble-based machine learning in large
scale industrial contexts. Empirical Software Engineering, 2016, 21(4): 1533–1578. [doi: 10.1007/s10664-015-9401-9]
[10] He JY, Meng ZP, Chen X, Wang Z, Fan XY. Semi-supervised ensemble learning approach for cross-project defect prediction. Ruan Jian
Xue Bao/Journal of Software, 2017, 28(6): 1455–1473 (in Chinese with English abstract). http://www.jos.org.cn/1000-9825/5228.htm
[doi: 10.13328/j.cnki.jos.005228]
[11] Baltrušaitis T, Ahuja C, Morency LP. Multimodal machine learning: A survey and taxonomy. IEEE Trans. on Pattern Analysis and
Machine Intelligence, 2019, 41(2): 423–443. [doi: 10.1109/TPAMI.2018.2798607]
[12] Akbari H, Yuan LZ, Qian R, Chuang WH, Chang SF, Cui Y, Gong BQ. VATT: Transformers for multimodal self-supervised learning
from raw video, audio and text. In: Proc. of the 35th Int’l Conf. on Neural Information Processing Systems. Curran Associates Inc., 2021.
24206–24221.
[13] Anand V, Ramesh R, Jin BS, Wang ZY, Lei XX, Lin CY. Multimodal language modelling on knowledge graphs for deep video
understanding. In: Proc. of the 29th ACM Int’l Conf. on Multimedia. ACM, 2021. 4868–4872. [doi: 10.1145/3474085.3479220]
[14] Dai J, Li QS, Xue H, Luo Z, Wang YL, Zhan SY. Graph collaborative filtering-based bug triaging. Journal of Systems and Software,