Generation of Resources Adapted to Hierarchical Progressive Teaching Scenarios in Ideological and Political Resource Repository Based on Generative AI

Abstract

Aiming at the problems of content homogeneity, insufficient depth of thought and poor adaptability of teaching scenes in ideological and political education resource generation, this paper proposes a generative artificial intelligence solution based on particle swarm optimization (PSO) and Transformer fusion. By integrating 157,000 multi-modal teaching data from multi-university ideological and political curriculum resource library, a multi-dimensional dataset covering knowledge elements, teaching organization and learning feedback is constructed. PSO algorithm is used to optimize the hyperparameters of Transformer model, such as layers, attention mechanism and learning rate. The experimental results show that PSO-Transformer model outperforms traditional rule generation, SVR, RNN and GPT-3 baseline models in topic relevance, semantic coherence, depth of thought and content appeal. The score of topic relevance is 9.3 and the score of semantic coherence is 9.1. The hierarchical and progressive teaching resource adaptation strategy constructed based on the generation results provide effective methodological support for the intelligent generation and accurate push of ideological and political education resources.