Abstract: Software evolution is a hot research field in software engineering. Because the open source software has the characteristics of group intelligence development and the process of evolution uncontrollable and not modeling, the traditional validation method can not apply to open source software. This paper proposes a software validation method from semantic function, attempts to cluster the codes in the way of the topic, and each topic presents a function module in software system. This paper transfers the evolving validation into finding the mapping relation between the function module and the requirements of software evolution. Finally, this paper takes advantage of source codes in open software, makes experiment and gets many actual data, then analyzes these data, the results make clear that the proposed method is better than the comparison method in distinguishing whether evolved or not, and it can be used to verify the software evolution.

Key words: software evolution, validation, topic model, open source software

摘要： 软件演化确认是软件工程领域的一个重点和热点的研究方向。由于开源软件具有群智开发，演化过程不可控和不可建模等特点，使得传统的确认方法不适合于开源软件，故从功能语义角度提出了一种软件演化确认方法，试图将代码按主题的方式进行聚类，每一个主题表征软件系统的一个功能集合，演化确认工作被转化为功能集合与演化需求之间的映射关系。通过对现有开源软件的源代码进行实验，获取了大量的真实可靠实验数据，对这些实验数据进行分析后得出的实验结果表明该方法相较基于文本的基线方法，更能有效区分是否已经演化，可以用于对软件演化进行确认工作。

关键词: 软件演化, 确认, 主题模型, 开源软件