Reproduce “A Dataset for Analysing News Framing in Chinese Media” 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
Owen Cook, Yida Mu, Xinye Yang, Xingyi Song, Kalina Bontcheva
A new dataset tailored for analysing news framing in Chinese media enables automated detection of generic news frames. Integrating this dataset with existing multilingual resources improves XLM-RoBERTa's F1-micro score for Chinese news framing detection to 0.753, an improvement of 0.171 over prior multilingual baselines.
Key insights
- A dedicated Chinese news framing dataset is essential for improving automatic detection performance in Chinese, as multilingual models alone generalize poorly.
- Large Language Models like GPT-4o, in a zero-shot setting, currently underperform fine-tuned transformer models for this specific multi-label classification task.
- Integrating language-specific datasets (Chinese) with existing multilingual datasets (SemEval) can enhance performance for the target language without degrading performance on others.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.