Title: Revealing Temporal Framing in News Media

Abstract: Beyond simply situating events on a timeline, temporal language serves as a rhetorical tool in news discourse, utilizing references to past, present, and future moments to influence interpretation and drive persuasion. This study investigates "temporal framing," which we define as the strategic application of time-related vocabulary to construct meaning rather than to establish chronological order. Building on existing research into temporality and framing, we introduce a taxonomy comprising eight distinct temporal frames. We operationalized this framework through expert annotation of a multilingual news corpus. The resulting dataset comprises 458 news articles in English and German, featuring over 2,000 sentences exhibiting temporal framing and approximately 3,000 specific temporal framing annotations, all extracted from a larger pool of more than 20,000 sentences. Our analysis covers the prevalence of these frames, their co-occurrence dynamics, and associated lexical indicators. Furthermore, we assess the efficacy of detecting temporal framing via supervised fine-tuning and zero-shot classification techniques. Experimental results indicate that temporal framing patterns are learnable at the sentence level, with supervised models demonstrating significant superiority over zero-shot methods. To facilitate ongoing research in this area, we have made the corpus publicly available at: https://mbzuai-nlp.github.io/temporal-framing/.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC