Clustering Urdu News Using Headlines
This paper that proposes and evaluates a new algorithm to automatically cluster Urdu news from different news agencies. The task is challenging because there are no language processing libraries for the Urdu language. The authors' experimental dataset consists of news from famous Pakistani media houses, including Jang, BBC Urdu, Express, UrduPoint, and Voice of America Urdu (VOA). The proposed algorithm only uses headlines to cluster the news. The authors argue that news headlines provide a concise summary of the news, which motivates them to use it instead of using the entire news story. Their experimental evaluation shows micro and macro averages for precision of 0.45 and 0.48 respectively for identifying similar news using headlines.
PDFDatasets
Introduced in the Paper:
Urdu News Headlines DatasetResults from the Paper
Task | Dataset | Model | Metric Name | Metric Value | Global Rank | Benchmark |
---|---|---|---|---|---|---|
Text Clustering | Urdu News Headlines Dataset | Vector Space Model | Related Headlines | 85 | # 1 |