PERSPECTIVESCOPE: A CONTRASTIVE FRAMEWORK FOR MULTI-PERSPECTIVE NEWS ANALYSIS WITH OPEN AND CLOUD MODELS
Pakistan Journal of Scientific Research · 25.03.2026 · 6 (01) · 85-92
DOI: 10.57041/vxr0nr13 ↗Kurz gefasst
Fachartikel in der Zeitschrift Pakistan Journal of Scientific Research, online erschienen am 25.03.2026. Verfasst von Aatif Hussain und 5 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).
Abstract (Originalsprache)
The proliferation of digital news media has created a critical paradox: while any major real-world event is now reported globally within minutes, most readers encounter only a single perspective – typically local or national – and remain unaware of how the same event is framed, quantified, or attributed by other outlets. This paper introduces Perspective Scope, a computational framework for Multi-Perspective News Analysis that systematically compares local (domestic) and global (international) coverage of identical events. Unlike conventional multi-document summarization, which seeks consensus, our approach is explicitly contrastive: given a target event, we pair articles from a domestic outlet and an international outlet and produce, in a single structured generation, both commonalities – facts corroborated across perspectives – and contradictions – discrepancies in factual reporting (e.g., casualty figures, statistics) and in narrative framing (e.g., actor characterization, causal attribution). We fine-tune an open Mixture-of-Experts model (Qwen3.6-35B A3B) for the task and compare it against a closed cloud model (GPT-4.1-mini) fine-tuned on the same training data and output schema on 229 held-out events. Each generated analysis is scored 0–100 for quality by an independent LLM-as-judge (Claude Opus 4.8). The open model attains a mean quality of 82.14 versus 80.51 for the cloud model (+1.63) and wins the per-event head-to-head on 125 of 216 decisive events (57.9%) against 91 for the cloud model – a modest but statistically reliable advantage (sign test, p = 0.025). We contribute a perspective-divergence taxonomy, an event-aligned evaluation methodology, and a controlled open-vs-cloud comparison, with implications for media-literacy education, automated fact-checking, and misinformation detection.
Zitieren
APA
Hussain, A., Ullah, N., Ullah, S., Khalid, A. R., Ahmad, M. U., & Shabbir, I. (2026). PERSPECTIVESCOPE: A CONTRASTIVE FRAMEWORK FOR MULTI-PERSPECTIVE NEWS ANALYSIS WITH OPEN AND CLOUD MODELS. Pakistan Journal of Scientific Research, 6(01), 85-92. https://doi.org/10.57041/vxr0nr13
Vancouver
Hussain A, Ullah N, Ullah S, Khalid AR, Ahmad MU, Shabbir I. PERSPECTIVESCOPE: A CONTRASTIVE FRAMEWORK FOR MULTI-PERSPECTIVE NEWS ANALYSIS WITH OPEN AND CLOUD MODELS. Pakistan Journal of Scientific Research. 2026;6(01):85-92. doi:10.57041/vxr0nr13
BibTeX
@article{Hussain2026,
author = {Hussain, Aatif and Ullah, Naeem and Ullah, Saleem and Khalid, Abdul Rehman and Ahmad, Muhmmad Usman and Shabbir, Imran},
title = {PERSPECTIVESCOPE: A CONTRASTIVE FRAMEWORK FOR MULTI-PERSPECTIVE NEWS ANALYSIS WITH OPEN AND CLOUD MODELS},
journal = {Pakistan Journal of Scientific Research},
year = {2026},
volume = {6},
number = {01},
pages = {85--92},
doi = {10.57041/vxr0nr13}
}
EUVIMED