Skip to main content

Table 6 BERTweet embedding extraction (EE) result with lr = 1e-4 and eps = 1e-4

From: A BERTweet-based design for monitoring behaviour change based on five doors theory on coral bleaching campaign

Scenario Description Maximum accuracy Maximum F1 score
EE#1 Last Layer 0.7714 0.7298
EE#2 All 12 Layers 0.7633 0.7406
EE#3 Last 4 Layers 0.7673 0.7833
EE#4 Last 2 Layers 0.7510 0.7484
EE#5 First 2 + Last 2 0.7510 0.7445
EE#6 First + Last 0.7673 0.7388
EE#7 Last 2 + Mid 2 0.7510 0.7589
EE#8 Last + Mid 0.7673 0.7214
  1. Scenario EE#1 is bolded because it has the highest accuracy while scenario EE#3 is bolded because it has the highest F1 score