Generating Hints for Programming Problems Using Intermediate Output
Generating Hints for Programming Problems Using Intermediate Output
复制标题
使用中间输出生成编程问题的提示
DOI：
--
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
T. Barnes
中科院分区：
文献类型：
--
作者：
Barry W. Peddycord;Andrew Hicks;T. Barnes
在这项工作中，我们比较了简单编程游戏中学生互动的两个表示。我们将这些表示形式称为世界国家和代码群体。世界范围是程序的输出的表示，是代码图的概括，在运行程序时采取的源代码的快照。我们的目标是将智能数据驱动的反馈纳入系统，例如生成提示以指导学生解决问题。使用世界各地，即使在不需要专家分析的情况下，即使是以前看不见的问题，也可以更轻松地比较学生的方法，即使是以前看不见的问题，可以简化这项任务。在教育节目游戏的背景下，机器人，我们发现世界各地需要更少的先前数据来在大多数情况下生成提示，而不会牺牲质量或可解释性。
In this work, we compare two representations of student interactions within the context of a simple programming game. We refer to these representations as Worldstates and Codestates. Worldstates, which are representations of the output of the program, are generalizations of Codestates, snapshots of the source code taken when the program is run. Our goal is to incorporate intelligent data-driven feedback into a system, such as generating hints to guide students through problems. Using Worldstates simplifies this task by making it easier to compare student approaches, even for previously unseen problems, without requiring expert analysis. In the context of the educational programming game, BOTS, we find that worldstates require less prior data to generate hints in a majority of cases, without sacrificing quality or interpretability.