Automatic detection of latent software component relationships from online Q&A sites
Proceedings of the 7th International Workshop on Realizing Artificial Intelligence Synergies in Software Engineering(2019)
摘要
Modern software system stacks are composed of large numbers of software components. These components may include a broad range of entities such as services, libraries, and frameworks, all intended to address specific requirements. It is not only necessary that these components satisfy respective functional and non-functional concerns, but also that the combinations of selected components work well together. The space of component combinations to explore is huge. Together with the almost universal lack of formal documentation suggesting desirable combinations and cautioning against undesirable ones, this renders the proper selection of combinations very challenging. For this reason, software engineers often solicit advice and document their experience on online forums such as community Q&A sites. In this paper, we show that these Q&A sites contain valuable knowledge about inter-component relations. We develop an approach using information extraction techniques to automatically identify three different types of compatibility relations from unstructured text on Q&A site postings. Our work demonstrates that identifying such relations is valuable for the design of component-based systems and that automatic relation extraction is a promising technique to systematically harness such community knowledge.
更多查看译文
关键词
Component Based Software Development,Relation Extraction,Information Extraction,Deep Learning
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络