Abstract
A challenge in digital content personalization can be sequencing multiple content components in a manner that is relevant to user intent and internally coherent. Existing methods may result in disjointed user experiences or introduce undesirable latency. A described system may address this by representing user intent and discrete content units as semantic embeddings in a vector space. A search algorithm, such as a beam search, can then iteratively construct and score candidate sequences of these units based on their collective semantic relevance to the user's intent, while also incorporating diversity constraints to help reduce redundancy. This process can facilitate the dynamic generation of ordered, cohesive, and personalized content sequences from a pre-approved inventory, which can be delivered to a user with potentially low latency.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Pandit, Sharbani; Rangan, Vijai Kasthuri; Solis, David; and Weiss, Maximilian, "Coherent Content Sequencing via Semantic Embeddings and Beam Search", Technical Disclosure Commons, (February 26, 2026)
https://www.tdcommons.org/dpubs_series/9406