Fig 1.
Confusion Matrix of VADER vs. RoBERTa Predictions (n = 5,000).
Diagonal cells represent agreement; off-diagonal cells indicate misclassifications. The 7.8% disagreement rate primarily reflects VADER’s tendency to classify context-dependent positive sentiment as neutral.
Table 1.
Sentiment Distribution by Theme (n = 52,874).
Table 2.
Mean Engagement Score (ES) by Sentiment and Theme.
Fig 2.
Engagement Score by Sentiment Polarity and Theme.
Fig 3.
Weekly Sentiment Trends by Theme.
Fig 4.
Most Frequent Terms in Virtual Events Tweets.
Table 3.
Linear Regression Results Predicting Engagement Score (n = 52,874).