Construction and Evolutionary Analysis of Dynamic Event Logic Graphs for the Field of Ideological and Political Education

Abstract

Addressing the challenge of inadequately mining the dynamic and correlational nature of event timelines within massive ideological and political education data, this paper proposes a method for constructing and analyzing a dynamic event logic graph for this domain, named the Temporal Event Graph Attention Network (TEGAT). This method fuses multi-source heterogeneous texts, identifies core event elements through natural language processing techniques, and constructs a dynamic event logic graph incorporating a temporal dimension to represent the sequential and causal relationships between events. Furthermore, this paper designs an evolutionary analysis algorithm based on a graph attention network and introduces the L2 norm of node embedding vectors as a proxy for event influence to quantitatively analyze the propagation paths and influence evolution of ideological and political hotspots. Experimental results on a self-constructed corpus in the ideological and political domain show that the proposed TEGAT model achieves an F1-score of 0.881 in event logic relation extraction, a 3.9% improvement over the next-best model, and significantly reduces the error in the event influence prediction task. This research provides a new technical path and theoretical support for the quantitative analysis and intelligent decision-making in ideological and political education.