Supervised ensemble learning methods towards automatically filtering Urdu fake news within social media

Area of Science:

• Natural Language Processing
• Machine Learning
• Information Science

Background:

• The proliferation of online misleading information, including fake news and reviews, poses a global challenge across various platforms.
• Existing fake news detection research predominantly focuses on English, neglecting resource-poor languages due to a lack of labeled datasets.
• Urdu language news faces a significant gap in automated fake news detection methodologies.

Discussion:

• This study addresses the challenge of fake news detection in the Urdu language, a resource-poor linguistic context.
• It introduces a novel annotated corpus of Urdu news articles specifically designed for fake news detection tasks.
• The research explores the efficacy of individual machine learning models and enhances performance through ensemble learning methods.

Key Insights:

• Ensemble learning methods, specifically Ensemble Selection and Vote models, demonstrate superior performance compared to individual machine learning models for Urdu fake news detection.
• The developed annotated Urdu corpus serves as a valuable resource for future research in low-resource fake news detection.
• Balanced accuracy, area under the curve, and mean absolute error metrics confirm the effectiveness of the proposed ensemble approaches.

Outlook:

• Future research should focus on expanding annotated corpora for more low-resource languages to combat global misinformation.
• Investigating advanced deep learning architectures within ensemble frameworks could further improve detection accuracy.
• Developing real-time fake news detection systems for Urdu news platforms is a critical next step.