ABSTRACT

Given the growing adoption of Internet of Things (IoT) technologies, several approaches have been presented to enable people to increase their control over their smart devices and provide relevant support. Recommendation systems have been proposed in many domains, but have received limited attention in the area of End-User Development (EUD). We propose a novel approach for formulating recommendations in this area, based on deconstructing trigger-action rules into sequences of elements and the links between them. For this purpose, we propose a solution inspired by methods aimed at addressing the sequence-prediction problem. We have used this approach to provide users with two different types of recommendations: full rules for the one being edited, and parts of rules relevant for the next step to take in order to complete the current rule editing. In this paper, we present the design and a first evaluation of the two different possibilities to generate and display recommendations in a block-based EUD environment for creating automations for IoT contexts.

摘要

鉴于物联网 (IoT) 技术的日益普及，已经提出了几种方法来使人们能够增强对智能设备的控制并提供相关支持。推荐系统已在许多领域提出，但在最终用户开发 (EUD) 领域受到的关注有限。我们提出了一种新方法来制定该领域的建议，该方法基于将触发-动作规则解构为元素序列及其之间的链接。为此，我们提出了一种解决方案，其灵感来自旨在解决序列预测问题的方法。我们使用这种方法为用户提供了两种不同类型的建议：正在编辑的规则的完整规则，以及与为完成当前规则编辑而采取的下一步相关的部分规则。