Abstract
Conventional personalization systems that utilize centralized, server-side user data collection may present privacy challenges and can result in profile contamination from transient user intents. A decentralized, client-side framework can shift data storage and processing to a user's local computing device (e.g., a smartphone, smart watch, wearable device, etc.). This framework can utilize a private, on-device knowledge graph to manage user context. For a given user action, a client-side process can extract a relevant sub-graph of contextual data, convert it into a transient, privacy-enhanced vector, and transmit it to a remote service. The service may return a generalized set of results that the device then privately re-ranks using the detailed local context. This approach can enable contextual personalization while potentially limiting the exposure of a user's broader personal information and reducing reliance on persistent server-side profiles.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
N/A, "Decentralized Contextual Personalization Using an On-device Personal Knowledge Graph", Technical Disclosure Commons, (September 07, 2026)
https://www.tdcommons.org/dpubs_series/11618