NOTIFI: Context-aware Smartphone Notification System

Keyword: Context-aware User Interface, NLP, Ubiquitous computing
Current smartphone users are often overwhelmed with a flood of messages, many of which are unimportant and can be distracting. It requires effort to find important notifications amidst this sea of messages. To address this issue, we propose Notifi, an intelligent notifier system that is context-aware.

Our system leverages three types of contextual information: user-defined contact importance, user scenario context inferred from smartphone sensors (e.g., work, sleep, relax, public), and semantic understanding of message content using a fine-tuned BERT language model. Privacy is a also key consideration in our system. Textual information related to privacy is masked during the prediction process, and specific content is not displayed in the floating notifications.

By considering these factors, Notifi determines the importance of a new notification in the current context, enabling different levels of notification prompts for messages of varying importance. These prompts include muting, floating notifications, vibrations, and different volume levels for ringtones. Results from user experiments demonstrate that Notifi effectively accentuates important messages while disregarding unimportant ones. Compared to the default notification method, Notifi’s approach to prompting different types of messages is more satisfying for users (disagreement rating: 1.81 vs. 1.41).