Cache-enabled Knowledge Base Construction Strategy in Semantic Communications
Yue Che, Huachao Xiong, Shujun Han, Xiaodong Xu
Abstract
With the increasing demand for intelligence in wireless networks, semantic communication is considered as a enabling technology to achieve goal-oriented communication. Deploying semantic knowledge base is seen as a critical technology for support semantic communication in resource-constrained systems. However, the current network architecture does not consider the way of deploying and constructing knowledge bases. To fill this gap, we propose a knowledge base deployment mechanism based on edge caching to support the application of semantic communication. Further, for the semantic understanding-oriented requests of intelligent devices, the knowledge base is considered to be deployed in the edge cache in the form of a knowledge graph. Considering the reusability of triples in requested subgraphs, we define a subgraph value function to measure the translatability and effectiveness of subgraphs, and establish a cached semantic knowledge base value maximization model. Besides, we propose two semantic knowledge base caching strategies, including a subgraph-based dynamic programming caching strategy and a triple-based greedy caching strategy. Compared with other schemes, the recovery accuracy of semantic information and the system delay of the proposed schemes both achieve better performances.