Detecting and classifying online health misinformation with 'Content Similarity Measure (CSM)' algorithm: an...

Area of Science:

• Digital Media Analysis
• Computational Linguistics
• Machine Learning

Background:

• Misinformation in the digital world, particularly in healthcare, poses significant risks.
• Existing automated fact-checking methods often rely on linguistic features and similarity measures, which struggle with sparse data and lack factual knowledge.

Purpose of the Study:

• To propose a novel Content Similarity Measure (CSM) algorithm for automated fact-checking of URLs in the healthcare domain.
• To address limitations of previous methods by incorporating factual knowledge and improving accuracy with sparse data.

Main Methods:

• Developed a 'Content Similarity Measure (CSM)' algorithm incorporating novel content similarity, domain-specific, and sentiment polarity score features.
• Employed information retrieval, natural language processing, and machine learning techniques for automated fact-checking.
• Validated the algorithm's robustness on three datasets: CoAID, FakeHealth, and ReCOVery.

Main Results:

• The CSM algorithm achieved 91.06% accuracy in its algorithmic approach, significantly outperforming Jaccard similarity (74.26%).
• The 'content similarity score' feature alone reached 88.26% accuracy, surpassing other features.
• CSM demonstrated high accuracy across datasets, with the algorithmic approach achieving up to 88.83% and the feature-based approach up to 86.80%.

Conclusions:

• The proposed CSM algorithm offers a superior approach to automated fact-checking for healthcare information.
• The algorithmic CSM approach demonstrates enhanced performance and robustness compared to feature-based methods and traditional similarity measures.
• CSM effectively addresses the challenge of misinformation detection in the digital media landscape.