Reproduce “FRaN-X: FRaming and Narratives-eXplorer” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Artur Muratov, Hana Fatima Shaikh, Vanshikaa Jani, Tarek Mahmoud, Zhuohan Xie, Daniil Orel, Aaryamonvikram Singh, Yuxia Wang and 16 more
FRaN-X is an interactive, multilingual system designed to analyze how individual entities are framed in news content by detecting mentions and classifying their narrative roles. The system employs a two-stage deep learning pipeline to assign entities to broad categories like Protagonist, Antagonist, or Innocent, and further to 22 fine-grained sub-roles, offering granular insights into media bias across five languages.
Key insights
- The two-stage neural architecture effectively combines entity mention detection with fine-grained narrative role classification, moving beyond pre-extracted entities to process raw text.
- Multilingual Transformer models, specifically mDeBERTa-v3-Large, demonstrate robust cross-lingual transfer capabilities, enabling effective performance across diverse languages such as Bulgarian, English, Hindi, Portuguese, and Russian.
- Data augmentation by propagating gold labels to co-referent entity mentions within documents leads to a 14.3 percentage point improvement in exact-match accuracy for entity detection.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.