Identifying propaganda from online social networks during COVID-19 using machine learning techniques

Area of Science:

• Computer Science
• Social Sciences
• Public Health

Background:

• The COVID-19 pandemic necessitated social distancing, leading to increased reliance on online social networks for information dissemination.
• The rapid spread of information, including propaganda, on social media platforms during the pandemic poses significant challenges.
• Propaganda, deliberately shared for political or religious influence, can shape public opinion and distort understanding of critical health issues.

Purpose of the Study:

• To identify and classify propaganda within tweets related to the COVID-19 pandemic.
• To evaluate the effectiveness of machine learning algorithms in detecting propaganda on social media during a global health crisis.

Main Methods:

• Data extraction from Twitter via its Application Program Interface (API).
• Manual annotation of tweets to label propaganda content.
• Hybrid feature engineering to select the most relevant features for classification.
• Binary classification of tweets using various machine learning algorithms, including decision trees.

Main Results:

• The decision tree algorithm demonstrated superior performance in classifying propaganda tweets compared to other tested algorithms.
• Hybrid feature engineering proved effective in identifying key features for propaganda detection.
• Manual annotation provided a crucial dataset for training and evaluating the machine learning models.

Conclusions:

• Machine learning, particularly decision trees, can effectively identify propaganda in social media discourse during public health emergencies like COVID-19.
• Further improvements in feature engineering and the application of deep learning methods could enhance the accuracy of propaganda detection.
• Understanding and mitigating the spread of online propaganda is vital for maintaining informed public discourse during pandemics.