A collection of French news articles across 5 main categories and 8 subcategories scraped from the Orange Actu website between 2011 and 2020. It features professionally written single-sentence titles and brief abstracts for each article, supporting two distinct summarization tasks.
Use Cases
- Fine-tune French language models for abstractive summarization using the 'OrangeSum Title' field.
- Train text classifiers to predict article categories like 'politics' or 'automotive' based on article content.
- Benchmark summarization algorithms on the 'OrangeSum Abstract' task to generate professional-style news summaries.
- Analyze shifts in French news topics over time using the decade-long temporal coverage.
Strengths
- Spans a 9-year period from February 2011 to September 2020.
- Categorized into 5 primary sections: France, world, politics, automotive, and society.
- Society category includes 8 specific subcategories such as 'insolite', 'culture', and 'high-tech'.
- Provides two ground-truth targets: a single-sentence title and a brief abstract.