Vec4Cred: a model for health misinformation detection in web pages

Area of Science:

• Computational Linguistics
• Health Informatics
• Machine Learning

Background:

• The proliferation of non-genuine information online, including fake news and opinion spam, is a growing concern.
• The COVID-19 pandemic exacerbated the spread of health misinformation, highlighting the need for automated detection solutions.
• Existing approaches often rely on handcrafted features and traditional Machine Learning models for content analysis.

Purpose of the Study:

• To propose and evaluate Vec4Cred, a novel health misinformation detection model.
• To assess the effectiveness of using embedded representations of structural and content features for classifying health information.
• To advance previous models by incorporating new features and architectural designs.

Main Methods:

• Developed a health misinformation detection model (Vec4Cred) leveraging embedded representations of web page characteristics.
• Utilized an embedding model pre-trained on medical data to extract features.
• Employed a deep learning classification model to categorize content as genuine health information or misinformation.

Main Results:

• The Vec4Cred model demonstrated effectiveness in distinguishing genuine health information from misinformation.
• The use of embedded representations of web page features proved advantageous over traditional methods.
• Evaluated the model's performance, highlighting improvements from previous iterations.

Conclusions:

• Vec4Cred offers a promising deep learning-based approach for automated health misinformation detection.
• The model's reliance on pre-trained embeddings for structural and content features enhances its classification capabilities.
• This research contributes to developing more robust tools for combating the spread of inaccurate health information online.