Say Your Reason: Extract Contextual Rules In Situ for Context-aware Service Recommendation

TL;DR

SayRea presents an in situ, interactive approach to extract contextual rules for context-aware mobile service recommendations by asking users for a single-sentence reason at the moment of service use and using an LLM to map that reason to context attributes. Rules are accumulated in a context-rule tree to support efficient, interpretable, and personalized recommendations, with negative rules enabling refinement. A 10-day field study (n=20) shows substantial rule accumulation (average 62.4 per user), a 45% service coverage, and positive usability feedback, indicating the method reduces user burden while maintaining accuracy and relevance. The work demonstrates the feasibility and impact of combining in situ human input with LLM-driven context attribution to enable low-burden, controllable, and interpretable mobile personalization, with potential extension to broader recommendation domains.

Abstract

This paper introduces SayRea, an interactive system that facilitates the extraction of contextual rules for personalized context-aware service recommendations in mobile scenarios. The system monitors a user's execution of registered services on their smartphones (via accessibility service) and proactively requests a single-sentence reason from the user. By utilizing a Large Language Model (LLM), SayRea parses the reason and predicts contextual relationships between the observed service and potential contexts (such as setting the alarm clock deep in the evening). In this way, SayRea can significantly reduce the cognitive load on users in anticipating future needs and selecting contextual attributes. A 10-day field study involving 20 participants showed that SayRea accumulated an average of 62.4 rules per user and successfully recommended 45% of service usage. The participants provided positive feedback on the system's usability, interpretability, and controllability. The findings highlight SayRea's effectiveness in personalized service recommendations and its potential to enhance user experience in mobile scenarios.

Figures (12)

• Figure 1: When user use services on their smartphones, SayRea proactively requests user to say the reason in situ. SayRea use the reason as bridge between context causes and service results to extract contextual rule for context-basd service recommendation. Next time if the context meets rules, our system provide quick access for the services on home screen and lock screen.
• Figure 2: An example of contextual rule extraction through two steps.
• Figure 3: Prompt design for LLM-powered context cause extraction with its purpose decomposition.
• Figure 4: A context rule tree Example. Nodes of the tree represent context attributes(At home, 24:00, Stilling, A5), and some nodes also store the context reasons(Before sleep, C1, C2, C4) and service results(Set Alarm Clock, S1, S2) for contextual rules.
• Figure 5: SayRea's system flow. SayRea extracts contextual rules while users using smartphones, and supports using service recommendations.