Optimizing Wireless Resource Allocation for Enhanced Semantic Communication System Reliability with Mismatched Background Knowledge Bases

Semantic communication is considered as a breakthrough beyond the Shannon paradigm, with the objective of achieving maximal semantic fidelity while minimizing the utilization of communication resources. In contrast to Shannon’s information theory, the fidelity of semantic communication is contingent upon shared background knowledge. Any mismatch between the background knowledge bases (KBs) of the transmitter and receiver can impede the receiver from decoding accurate semantic information. Consequently, resource allocation for semantic communication, taking into account the degree of knowledge matching, becomes a pivotal issue to ensure the reliability and efficiency of semantic transmission. In this paper, we establish a wireless semantic communication system tailored for text data transmission under conditions of mismatched background KBs. We introduce a novel metric, termed the Semantic Communication System Reliability (SCSR), which integrates semantic reliability with transmission reliability. Furthermore, we employ a deep reinforcement learning (DRL) algorithm to intelligently and dynamically allocate communication resources among users, with the ultimate goal of maximizing the system reliability of the wireless semantic communication system. Comparative experiments conducted subsequently confirm the advantages of the proposed algorithm in enhancing SCSR performance.