This paper investigates semantic communication in high-speed mobile Internet of Vehicles (IoV) environments, focusing on spectrum sharing between vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications. The authors address spectrum scarcity and network traffic by proposing a semantic-aware spectrum sharing algorithm (SSS) based on deep reinforcement learning (DRL) soft actor-critic (SAC) approach. They introduce metrics such as high-speed semantic spectrum efficiency (HSSE) and semantic transmission rate (HSR) tailored for high-speed mobile IoV spectrum sharing. The SSS algorithm optimizes decision-making based on semantic data, including link selection, transmission power, and semantic symbol length, to maximize HSSE and enhance the success rate of effective semantic information transmission (SRS). Experimental results show that the SSS algorithm outperforms traditional communication-based spectrum sharing methods and other reinforcement learning approaches, achieving a 15% increase in HSSE and a 7% increase in SRS. The paper also discusses the system model, DeepSC transceivers, novel metrics for spectrum sharing, and the optimization problem. The training and testing phases of the SSS algorithm are detailed, and simulation results demonstrate its effectiveness.This paper investigates semantic communication in high-speed mobile Internet of Vehicles (IoV) environments, focusing on spectrum sharing between vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications. The authors address spectrum scarcity and network traffic by proposing a semantic-aware spectrum sharing algorithm (SSS) based on deep reinforcement learning (DRL) soft actor-critic (SAC) approach. They introduce metrics such as high-speed semantic spectrum efficiency (HSSE) and semantic transmission rate (HSR) tailored for high-speed mobile IoV spectrum sharing. The SSS algorithm optimizes decision-making based on semantic data, including link selection, transmission power, and semantic symbol length, to maximize HSSE and enhance the success rate of effective semantic information transmission (SRS). Experimental results show that the SSS algorithm outperforms traditional communication-based spectrum sharing methods and other reinforcement learning approaches, achieving a 15% increase in HSSE and a 7% increase in SRS. The paper also discusses the system model, DeepSC transceivers, novel metrics for spectrum sharing, and the optimization problem. The training and testing phases of the SSS algorithm are detailed, and simulation results demonstrate its effectiveness.