Explanation videos are increasingly common on media websites such as YouTube and are used by school students, university students, and members of the general public alike. Such videos cover all areas of knowledge and aim to provide viewer-appropriate explanations concerning a large variety of topics. It is, however, still far from clear how such videos work and under what conditions they are effective. In this paper, we consider how this can be measured and whether guidelines can be determined empirically for their improvement. Building on the notion of semantic waves developed in Legitimation Code Theory, we discuss the design of cumulative knowledge-building processes and how to isolate cases where this fails to operate, selecting examples from a number of videos of this kind. To support this, we introduce a detailed annotation framework that fully reflects the multimodally-rich extent of our data. This framework systematically defines coding categories for use with the ELAN annotation tool and uses these for the adjacent construction of multimodal cohesive chain diagrams. We first motivate and describe the application of this annotation framework and then show through subsequent multimodal cohesion analyses how constructed cohesive chains can be interpreted within the scope of pedagogically relevant semantic wave patterns.

解释性视频在YouTube等媒体网站上越来越普遍，在校学生，大学生和一般公众都在使用。这些视频涵盖了所有知识领域，旨在为观众提供有关各种主题的适当解释。但是，仍不清楚此类视频如何工作以及在什么条件下有效。在本文中，我们考虑了如何对其进行衡量以及是否可以凭经验确定准则以对其进行改进。基于语义波的概念我们使用“合法性代码理论”进行了开发，我们讨论了累积知识构建过程的设计，以及如何隔离失败的案例，并从许多此类视频中选择示例。为了支持这一点，我们引入了一个详细的注释框架，该框架完全反映了我们数据的多模式丰富范围。该框架系统地定义了与ELAN注释工具一起使用的编码类别，并将这些编码类别用于多模态内聚链图的相邻构造。我们首先激发并描述此注释框架的应用，然后通过后续的多模式衔接分析显示如何在教学相关的语义波动模式范围内解释构造的衔接链。