知识驱动的软件缺陷分析与理解关键技术研究

开发人员要实现软件缺陷精准修复，应对软件缺陷进行深入分析和准确理解。但是，现有软件缺陷分析技术对隐藏在缺陷数据中的知识、知识间的关联关系以及反映缺陷发生或修复过程的事件等方面的挖掘与分析存在明显不足，无法有效辅助开发人员对软件缺陷进行全面、准确的理解。针对上述问题，本项目旨在研究知识驱动的软件缺陷分析与理解关键技术。具体而言，对缺陷知识及其关联关系进行刻画和定义，建立缺陷知识分类体系，为缺陷知识的识别奠定基础；针对历史修复缺陷，利用数据挖掘和机器学习等技术，进行缺陷知识自动识别、缺陷知识间关联关系抽取以及缺陷知识精化，以更好地组织、管理缺陷知识；在上述工作基础上，针对新缺陷，通过缺陷知识生成和事件抽取，建立以知识为中心的缺陷分析与理解集成服务，从而为开发人员全面、准确的软件缺陷分析与理解提供理论和技术支持。

Developers need to perform an in-depth analysis and accurate understanding of software bugs during the bug resolution process. However, existing bug analysis technology suffers from serious challenges in mining the hidden knowledge in the bugs, exploiting implicit associations among the bug knowledge, and characterizing the procedural events for bug occurrence and fix, thereafter, cannot provide effective support for developers’ comprehensive and accurate bug understanding. To overcome the above challenges, this project aims to study the key techniques in knowledge-driven software bug analysis and understanding. First, the hidden knowledge and implicit associations are defined to establish a bug knowledge classification system, laying a foundation for the automatic recognition of bug knowledge from the bug data. Second, data mining and machine learning techniques are employed to better organize and manage the bug knowledge by identifying the hidden knowledge and implicit associations as well as refining the bug knowledge from the historical bug repository. Finally, for the incoming bug reports at hand, bug knowledge generation and bug event extraction are developed to realize knowledge-centric bug analysis and understanding integration services, providing theoretic and technical supports for developers’ comprehensive and accurate bug understanding.