Uncovering the Bigger Picture: Comprehensive Event Understanding Via Diverse News Retrieval

TL;DR

NEWSCOPE is a two-stage news retrieval framework that improves event understanding by explicitly modeling semantic variation at the sentence level, promoting diversity without sacrificing relevance.

Contribution

It introduces a novel two-stage retrieval method with sentence-level diversity modeling and new metrics, enhancing event coverage in news retrieval systems.

Findings

Outperforms baselines in diversity metrics

Achieves higher event coverage without relevance loss

Demonstrates effectiveness of fine-grained diversity modeling

Abstract

Access to diverse perspectives is essential for understanding real-world events, yet most news retrieval systems prioritize textual relevance, leading to redundant results and limited viewpoint exposure. We propose NEWSCOPE, a two-stage framework for diverse news retrieval that enhances event coverage by explicitly modeling semantic variation at the sentence level. The first stage retrieves topically relevant content using dense retrieval, while the second stage applies sentence-level clustering and diversity-aware re-ranking to surface complementary information. To evaluate retrieval diversity, we introduce three interpretable metrics, namely Average Pairwise Distance, Positive Cluster Coverage, and Information Density Ratio, and construct two paragraph-level benchmarks: LocalNews and DSGlobal. Experiments show that NEWSCOPE consistently outperforms strong baselines, achieving significantly higher diversity without compromising relevance. Our results demonstrate the effectiveness of fine-grained, interpretable modeling in mitigating redundancy and promoting comprehensive event understanding. The data and code are available at https://github.com/tangyixuan/NEWSCOPE.