Abstract
<title>Abstract</title> <p>Climate change has emerged as a central site of political discourse across multilingual governance systems, yet the mechanisms by which different language communities construct, adapt, and diverge in their framing of this shared issue remain poorly understood. This study investigates how semantic frames are cross-linguistically constructed in Chinese, Spanish, and German climate-change political discourse, asking whether and how convergence and divergence in framing can be quantitatively modelled through distributional semantics. A trilingual corpus of official political speeches (2015–2023; 187,952–260,711 tokens per language) was analysed using TF-IDF weighting, co-occurrence network modelling, and contextual embedding techniques (Word2Vec and BERT), targeting three recurring frames: global responsibility, national interest, and technological innovation. Results reveal systematic convergence in issue definition and causal attribution, reflecting shared lexical repertoires and semantic-space alignment shaped by transnational discourse circulation. Cross-linguistic similarity in contextual embeddings is statistically significant across all three language pairs and frames (permutation tests, p ≤ 0.031). However, frame instantiation diverges markedly in moral evaluation and policy recommendation, reflecting culturally embedded strategies of semantic adaptation: Chinese discourse constructs collective responsibility through institutionalised formulations such as “Community of Shared Future for Mankind”; Spanish discourse foregrounds climate justice and civic proximity; and German discourse prioritises legal-institutional accountability. These findings demonstrate that framing can be operationalised as a computationally tractable, measurable construct, extending Entman’s framework into cross-linguistic empirical analysis. The study contributes to computational discourse linguistics, political communication research, and multilingual climate governance studies by providing a replicable methodological pipeline for cross-linguistic semantic framing analysis applicable to other policy domains and language pairs.</p>