A Multiplex Network Approach for Longitudinal Frame Detection in Newspaper and Press Language

Computational approaches to framing analysis continue to face substantial difficulties in modelling temporal evolution without conflating substantive narrative change with methodological artefacts. This paper proposes a multiplex network framework in which temporal windows are represented as structurally coupled layers and analysed by optimising a unified multislice modularity function. By embedding temporal continuity within the network architecture, the approach enables discursive communities to persist across periods while remaining sensitive to local contextual variation. Structural validation based on Matrix Similarity further ensures that inferred continuities reflect internal relational consistency rather than superficial lexical overlap. An empirical application of the framework to AI news illustrates its capacity to identify persistent framing structures and their systematic reconfiguration over time. The proposed framework provides a structured approach to addressing alignment issues in dynamic discourse analysis, thereby supporting the preservation of semantic coherence in evolving media environments.