Reminder systems commonly rely on fixed schedules, location triggers, or simple rules, limiting their ability to leverage the rich sensing capabilities of modern smart homes. A key challenge lies in enabling users to specify context-aware reminders without requiring complex configurations. We present a system pipeline that supports reminder authoring through natural language and conversational interaction. The pipeline translates user requests into structured representations and executable logic, incorporating time-based, activity-based, sensor-based, and state-based conditions. We conducted two studies to examine how users express reminder intent and how conversational support influences the authoring process. In Study 1 (N=40), we analyzed 233 user-authored reminders and identified challenges in expressing reminders with diverse and complex logic. Based on these findings, we refined the system and evaluated it in Study 2 (N=10), demonstrating improved handling of time-based, activity-based, sensor-based, and state-based conditions. Our results highlight the diversity and ambiguity of user expressions and show that conversational guidance can help structure these expressions into flexible, context-aware reminders.

翻译：提醒系统通常依赖于固定时间表、位置触发或简单规则，这限制了其利用现代智能家居丰富感知能力。关键挑战在于让用户能够指定情境感知的提醒，而无需复杂配置。我们提出一个支持通过自然语言和对话交互创作提醒的系统流水线。该流水线将用户请求转化为结构化表示和可执行逻辑，融合了基于时间、活动、传感器和状态的条件。我们开展两项研究来考察用户如何表达提醒意图，以及对话支持如何影响创作过程。在研究1（N=40）中，我们分析了233条用户创作的提醒，并识别出在表达具有多样复杂逻辑的提醒时的挑战。基于这些发现，我们优化了系统，并在研究2（N=10）中进行了评估，展示了系统在基于时间、活动、传感器和状态条件处理上的改进。我们的结果凸显了用户表达的多样性与歧义性，并表明对话引导有助于将这些表达结构化为灵活的情境感知提醒。