Beyond the Surface: Uncovering Implicit Locations with LLMs for Personalized Local News

TL;DR

This work addresses implicit location detection for local-news personalization by comparing traditional location techniques (NER, Knowledge Graphs) with Large Language Models in Taboola's homepage personalization system. It demonstrates that LLMs, particularly when augmented with KG data, achieve superior local-article classification offline, and delivers meaningful online gains in local views without compromising brand identity. A production-ready pipeline combines offline and online flows, prompting strategies, and validation steps to scale LLM-based locality signals to millions of requests per second. The results show a practical, scalable path to enhancing local-news exposure, while also highlighting trade-offs such as knowledge noise, toponym ambiguity, and overhead from KG enrichment, guiding future improvements in explainability and robustness.

Abstract

News recommendation systems personalize homepage content to boost engagement, but factors like content type, editorial stance, and geographic focus impact recommendations. Local newspapers balance coverage across regions, yet identifying local articles is challenging due to implicit location cues like slang or landmarks. Traditional methods, such as Named Entity Recognition (NER) and Knowledge Graphs, infer locations, but Large Language Models (LLMs) offer new possibilities while raising concerns about accuracy and explainability. This paper explores LLMs for local article classification in Taboola's "Homepage For You" system, comparing them to traditional techniques. Key findings: (1) Knowledge Graphs enhance NER models' ability to detect implicit locations, (2) LLMs outperform traditional methods, and (3) LLMs can effectively identify local content without requiring Knowledge Graph integration. Offline evaluations showed LLMs excel at implicit location classification, while online A/B tests showed a significant increased in local views. A scalable pipeline integrating LLM-based location classification boosted local article distribution by 27%, preserving newspapers' brand identity and enhancing homepage personalization.

Figures (2)

• Figure 1: Local Articles Recall
• Figure 2: Localization with LLMs Production Pipeline