Abstract
Techniques are described for recommendation pipelines that track per-user per-topic satiation state and modulate topic diversity and quality policies accordingly. For each user-topic pair, the system maintains exposure counts and an engagement quality trend computed from a composite of dwell time, save rate, and completion rate. The user-topic pair is classified into one of six phases: discovery, rising, plateau, declining, saturated, or recovery, using exposure, trend, and time since last exposure, with thresholds normalized by the user’s consumption rate. Phase-dependent multipliers are computed to scale a baseline minimum quality threshold and a base per-topic diversity cap, including relaxing diversity and lowering the quality bar in rising and tightening diversity and raising the quality bar in saturated. For declining or saturated topics, semantically adjacent topic substitution may be applied, and recovery may be tested via a probe item with backoff on negative response.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Anonymous, "Systems and Methods for Phase-Specific Content Diversity Modulation Based on Per-User Per-Topic Satiation State Tracking", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/10753