Personalized Language-Oriented Semantic Communication

In the emerging research field of Semantic Communication, the focus is on making network communications more effective and efficient. While prior research has primarily concentrated on improving transmission efficiency, limited efforts have addressed the overall effectiveness of communication. In this work, we propose a novel framework for personalized content dissemination in broadcast communication systems. In the envisioned system, the sender transmits either a textual description or a compressed latent representation of the content. At the receiver side, a personalization module leverages large language models to dynamically manipulate the received semantics, exploiting a user interest database. Subsequently, generative models recreate the final content conditioned on the manipulated semantics, ensuring that the regenerated content is both semantically consistent with the original and tailored to individual user interests. This communication paradigm aims to enhance overall system effectiveness and maximize user engagement through relevance and personalization. Potential applications include targeted advertising, personalized news or media delivery, and adaptive educational content.