1,200 Urdu news documents categorized into three levels of manual text reuse annotation: wholly derived, partially derived, and non-derived. The collection focuses on real-world journalistic content to facilitate research in text derivation and plagiarism detection in the Urdu language.
Use Cases
- Train supervised classification models to detect text reuse levels using the manual document-level labels
- Benchmark Urdu-specific natural language processing tools for identifying 'wholly derived' versus 'partially derived' news articles
- Analyze patterns of content propagation in Urdu journalism using the provided derivation categories
Strengths
- 1,200 documents sourced from real-world Urdu journalism
- Manual document-level annotations for three specific reuse classes
- Classification labels include 'wholly derived', 'partially derived', and 'non derived'