A Shared IPTC Topic Space for Cross-Source Topic Modelling

2026-06-25

AI Summary

This study addresses the challenge of inconsistent topic spaces arising from corpus-specific modeling, which hinders cross-source comparability in cross-media analysis. To overcome this, the authors propose an externally anchored, reproducible approach that constructs a shared topic space grounded in the IPTC Media Topics taxonomy. The method employs guided BERTopic to discover topics and maps them to 94 top-level IPTC categories via weighted keyword matching against target centroids, subsequently aggregating these into 17 parent-level themes based on maximum similarity. Experiments on the New York Times 2011 corpus demonstrate that the approach maintains high mapping coverage even under stringent thresholds. Moreover, parent-level augmentation of target construction significantly enhances both coverage and consistency, with performance degradation exhibiting a gradual rather than abrupt decline as thresholds tighten.

Abstract

Comparing topic attention across different media is hindered by a fundamental modelling problem: topic models fitted separately to each corpus produce corpus-specific topic spaces that cannot be aligned directly. This paper presents a reproducible framework that places corpora in a single shared topic space defined by a taxonomy. Discovered topics are obtained with guided BERTopic, scored against the ninety-four IPTC Media Topics' taxonomy topics (level-1) through weighted keyword and target centroids, and then collapsed upward to seventeen IPTC parent topics by a maximum-similarity rule. The framework was developed and selected on a controlled New York Times 2011 corpus through a narrowing sequence: a broad model screen, a focused mapping refinement, a strict finalist comparison, a target-construction ablation, and a threshold calibration. In this corpus, the guided family retained substantially stronger mapped coverage than a zero-shot benchmark under stricter assignment thresholds, a parent-enriched target construction improved both coverage and parent consistency, and coverage declined gradually rather than collapsing as the assignment threshold was tightened. The contribution is an externally anchored method for constructing a shared topic space that enables reproducible cross-source topic comparison.

Problem

• cross-source topic modelling
• topic space alignment
• IPTC taxonomy
• topic comparison
• corpus-specific topics

Innovation

• shared topic space
• guided BERTopic
• IPTC taxonomy
• cross-source topic modelling
• externally anchored alignment