Conventional ideological and political high-quality teaching resource sharing methods mainly use Oracle VM VirtualBox to generate sharing channels, which is vulnerable to changes in resource similarity threshold, resulting in poor teaching resource sharing effect. Certainly, developing a novel approach for sharing high-quality ideological and political teaching resources, leveraging deep learning and graph neural networks, is crucial. This paper endeavors to establish a sharing model for such resources, utilizing graph neural networks and deep learning techniques. Additionally, we devise a sharing algorithm and establish a sharing center, thereby facilitating the dissemination of high-quality ideological and political teaching resources. Experimental outcomes demonstrate that our proposed method exhibits a high resource utilization rate and scalability, while maintaining a low packet loss rate and reduced MC bifurcation probability. These findings attest to the effectiveness, reliability, and practical value of our approach in enhancing the sharing of teaching resources. Consequently, our work contributes significantly to improving the efficiency of ideological and political education and advancing the reform of related curricula.