Using visual dataflow programming for interactive model comparison
Using visual dataflow programming for interactive model comparison
复制标题
DOI：
10.1145/2642937.2642984
复制
发表时间：
2014-09
期刊：
影响因子：
--
通讯作者：
Rainer Lutz;S. Diehl
中科院分区：
文献类型：
--
作者：
Rainer Lutz;S. Diehl
在软件工程中，基于图的模型的比较是一个众所周知的问题。尽管已经提出了不同的比较指标，但在某些情况下，自动或预配合方法无法提供合理的结果。尤其是当模型包含语义相似性或差异时，通常需要其他人类知识。但是，只有很少的方法解决了如何在比较模型时如何支持人类的问题。在本文中，我们提出了一种用于交互模型比较的工具。我们在以前的工作中确定的一系列准则为它的设计提供了信息。特别是，我们的原型利用Visual DataFlow编程来允许用户实施自定义比较策略。为此，他们可以将指标和图形操作相结合以计算相似性和差异，并使用颜色编码可视化获得的结果。此外，一项定性用户研究允许评估我们的工具是否以及如何促进模型之间的相似性和差异的迭代探索。
In software engineering the comparison of graph-based models is a well-known problem. Although different comparison metrics have been proposed, there are situations in which automatic or pre-configured approaches do not provide reasonable results. Especially when models contain semantic similarities or differences, additional human knowledge is often required. However, only few approaches tackle the problem of how to support humans when comparing models. In this paper, we propose a tool for interactive model comparison. Its design was informed by a set of guidelines that we identified in previous work. In particular, our prototype leverages visual dataflow programming to allow users to implement custom comparison strategies. To this end, they can combine metrics and graph operations to compute similarities and differences, and use color coding to visualize the gained results. Additionally, a qualitative user study allowed to assess whether and how our tool facilitates iterative exploration of similarities and differences between models.