Entity Framing and Role Portrayal in the News
🤖 AI Summary
This work addresses the challenge of cross-lingual, fine-grained identification of entity narrative roles (e.g., Guardian, Tyrant, Victim) in news texts. We introduce the first multilingual hierarchical corpus for Ukraine–Russia war and climate change reporting—comprising 1,378 documents across five languages and over 5,800 entity mentions. We propose a narrative-theory-informed three-tier role taxonomy (Protagonist/Antagonist/Innocent) with 22 archetypal labels, supporting annotation at document-, paragraph-, and sentence-level granularity. Our method integrates multilingual Transformer fine-tuning, hierarchical zero-shot inference using large language models (LLMs), and human-in-the-loop validation. Experiments demonstrate substantial improvements over state-of-the-art baselines; notably, we provide the first empirical validation that LLMs can effectively perform zero-shot, fine-grained narrative role classification and generalize robustly across languages.
Application Category
📝 Abstract
We introduce a novel multilingual hierarchical corpus annotated for entity framing and role portrayal in news articles. The dataset uses a unique taxonomy inspired by storytelling elements, comprising 22 fine-grained roles, or archetypes, nested within three main categories: protagonist, antagonist, and innocent. Each archetype is carefully defined, capturing nuanced portrayals of entities such as guardian, martyr, and underdog for protagonists; tyrant, deceiver, and bigot for antagonists; and victim, scapegoat, and exploited for innocents. The dataset includes 1,378 recent news articles in five languages (Bulgarian, English, Hindi, European Portuguese, and Russian) focusing on two critical domains of global significance: the Ukraine-Russia War and Climate Change. Over 5,800 entity mentions have been annotated with role labels. This dataset serves as a valuable resource for research into role portrayal and has broader implications for news analysis. We describe the characteristics of the dataset and the annotation process, and we report evaluation results on fine-tuned state-of-the-art multilingual transformers and hierarchical zero-shot learning using LLMs at the level of a document, a paragraph, and a sentence.
Problem
Research questions and friction points this paper is trying to address.
Analyzes entity framing in multilingual news articles.
Introduces a dataset for role portrayal in global issues.
Evaluates multilingual transformers on hierarchical role classification.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Multilingual hierarchical corpus
Fine-grained role taxonomy
State-of-the-art transformers
🔎 Similar Papers
No similar papers found.
