Implicit Semantic Communication Based on Bayesian Reconstruction Framework

TL;DR

This work tackles the challenge of conveying semantic meaning when higher-order relations are essential but not directly observable. It introduces the Semantic Bayesian Reconstruction Framework (SBRF), which blends a semantic encoder/decoder pipeline with Bayesian hypergraph inference to recover implicit high-order semantics (hyperedges) from pairwise explicit semantics. The approach formulates a Bayesian posterior P(ℋ|hat{𝒢},θ) and uses Metropolis-Hastings refinement to prune redundant hyperedges, achieving linear per-iteration complexity with respect to the number of hyperedges and demonstrating up to 90% recovery accuracy on real datasets. The results show robust implicit semantic reconstruction across varying SNRs and outperform several baselines, underscoring the practical value of high-order semantic recovery in wireless semantic communications.

Abstract

Semantic communication is a novel communication paradigm that focuses on the transportation and delivery of the \emph{meaning} of messages. Recent results have verified that a graphical structure provides the most expressive and structurally faithful formalism for representing the relational semantics in most information sources. However, most existing works represent the semantics based on pairwise relation-based graphs, which cannot capture the higher-order interactions that are essential for some semantic sources. This paper proposes a novel Bayesian hypergraph inference-based semantic communication framework that can directly recover implicit semantic information involving high-order hyperedges at the receiver based on the pairwise relation-based explicit semantics sent by the transmitter. Experimental results based on real-world datasets demonstrated that the proposed SBRF achieves up to 90\% recovery accuracy of the high-order hyperedges based on the pairwise relation-based explicit semantics.

Figures (7)

• Figure 1: The framework of our proposed semantic communication architecture.
• Figure 2: The number of ground truth and recovered hyperedges with different numbers of entities based on the experiments conducted on the FB-AUTO dataset.
• Figure 3: The compression rate of the semantic compressor at the semantic encoder under different limits of maximum hyperedge lengths $L$.
• Figure 4: The entropy values of the accepting probability of our proposed MH algorithm for six different datasets: FB-AUTO, JF17K, M-FB15k, Wikipeople, NDC_C, and Walmart.
• Figure 5: The accuracy of implicit semantic information recovery under different received SNRs based on the experiments conducted on six different datasets.