CoVerifi

CoVerifi evaluates the credibility of COVID-19-related news by combining machine learning classification with crowdsourced human feedback to detect misinformation and generate labeled data for research.

Key Features:

• Machine Learning Integration: Uses machine learning algorithms to classify the veracity of COVID-19-related news and identify patterns indicative of false information.
• Human Feedback Mechanism: Incorporates user votes as crowdsourced labels that provide real-time feedback to inform and refine model predictions.
• Open Source Data Release: Produces labeled datasets from user annotations and model outputs and releases them as open source for research on COVID-19 misinformation.

Scientific Applications:

• Misinformation Detection: Automated classification of news content to identify and monitor COVID-19-related falsehoods during the infodemic.
• Public Health Communication Support: Informing interventions aimed at reducing stress and mitigating panic-induced behaviors such as panic buying or riots by identifying misleading information.
• Research Dataset Provision: Supplying open-source labeled datasets to enable development and evaluation of new methods for misinformation prevention and analysis.

Methodology:

Combines automated machine learning model training on existing datasets to detect patterns of false information with user votes that provide real-time crowdsourced labels used to refine the models.