当前位置: X-MOL 学术Requirements Eng. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Automatic recommendation to omitted steps in use case specification
Requirements Engineering ( IF 2.8 ) Pub Date : 2018-02-07 , DOI: 10.1007/s00766-018-0288-z
Deokyoon Ko , Suntae Kim , Sooyong Park

Completeness is one of the key attributes for a high-quality software requirements specification. Although incomplete requirements frequently occur in the requirements specification, it is rarely discovered. This turns out to be one of the major causes of software project failure. In order to handle this issue, this paper proposes an automatic approach to recommending omitted steps in a use case-based requirements specification. First, we automatically extract diverse scenario patterns by using the verb clustering algorithm and scenario flow graphs. Based on the scenario patterns, our approach detects omitted steps of user’s scenarios by the pattern matching algorithm and automatically recommends appropriate steps for the omitted parts. For validation of our approach, we have developed tool support, named ScenarioAmigo, and collected 231 use case specifications composing of 1874 scenario steps from 12 academic or proprietary projects. We first carried out the preliminary study to decide appropriate thresholds and weights. Then, we conducted three experiments as a quantitative performance evaluation. First, the cross-validation for the collected scenarios shows the 76% precision and 80% recall. Second, the comparison of recall of ScenarioAmigo to that of human experts obtained the 20% higher score. As the last experiment, we compared the result of ScenarioAmigo and human experts in terms of severity of each scenario and found that our approach could recommend normal as well as important scenarios, compared to the human experts.

中文翻译:

自动推荐用例规范中省略的步骤

完整性是高质量软件需求规范的关键属性之一。尽管需求规格说明中经常出现不完整的需求,但很少被发现。结果证明这是软件项目失败的主要原因之一。为了解决这个问题,本文提出了一种自动方法来推荐基于用例的需求规范中省略的步骤。首先,我们通过使用动词聚类算法和场景流图自动提取不同的场景模式。基于场景模式,我们的方法通过模式匹配算法检测用户场景中遗漏的步骤,并自动为遗漏部分推荐合适的步骤。为了验证我们的方法,我们开发了工具支持,名为 ScenarioAmigo,并收集了 231 个用例规范,由 12 个学术或专有项目的 1874 个场景步骤组成。我们首先进行了初步研究,以确定适当的阈值和权重。然后,我们进行了三个实验作为定量的性能评估。首先,对收集到的场景的交叉验证显示了 76% 的准确率和 80% 的召回率。其次,ScenarioAmigo 的召回率与人类专家的召回率相比,得分高出 20%。作为最后一个实验,我们比较了 ScenarioAmigo 和人类专家在每个场景的严重程度方面的结果,发现与人类专家相比,我们的方法可以推荐正常和重要的场景。我们首先进行了初步研究,以确定适当的阈值和权重。然后,我们进行了三个实验作为定量的性能评估。首先,对收集到的场景的交叉验证显示了 76% 的准确率和 80% 的召回率。其次,ScenarioAmigo 的召回率与人类专家的召回率相比,得分高出 20%。作为最后一个实验,我们比较了 ScenarioAmigo 和人类专家在每个场景的严重程度方面的结果,发现与人类专家相比,我们的方法可以推荐正常和重要的场景。我们首先进行了初步研究,以确定适当的阈值和权重。然后,我们进行了三个实验作为定量的性能评估。首先,对收集到的场景的交叉验证显示了 76% 的准确率和 80% 的召回率。其次,ScenarioAmigo 的召回率与人类专家的召回率相比,得分高出 20%。作为最后一个实验,我们比较了 ScenarioAmigo 和人类专家在每个场景的严重程度方面的结果,发现与人类专家相比,我们的方法可以推荐正常和重要的场景。ScenarioAmigo 的召回率与人类专家的召回率相比,得分高出 20%。作为最后一个实验,我们比较了 ScenarioAmigo 和人类专家在每个场景的严重程度方面的结果,发现与人类专家相比,我们的方法可以推荐正常和重要的场景。ScenarioAmigo 的召回率与人类专家的召回率相比,得分高出 20%。作为最后一个实验,我们比较了 ScenarioAmigo 和人类专家在每个场景的严重程度方面的结果,发现与人类专家相比,我们的方法可以推荐正常和重要的场景。
更新日期:2018-02-07
down
wechat
bug