Table 14 Best performing model based on each learning approach.
From: Multilingual identification of nuanced dimensions of hope speech in social media texts
Learning Approach | English | Spanish | German |
|---|---|---|---|
Binary Hope Speech Detection | |||
Traditional ML | 0.8006 SVM(RBF) unigrams | 0.7779 SVM(linear) uni+bigrams | 0.8076 SVM(linear) uni+bi+trigrams |
DL | 0.7869 BiLSTM | 0.7567 CNN | 0.7847 CNN |
Transformers | 0.8623 xlm-roberta-base | 0.8387 bert-base-spanish-wwm-uncased | 0.8704 bert-base-german-dbmdz-uncased |
Multiclass Hope Speech Detection | |||
Traditional ML | 0.5169 SVM(linear) unigrams | 0.4896 SVM(linear) unigrams | 0.4472 DT uni+bigrams |
DL | 0.5703 BiLSTM | 0.5190 CNN | 0.5028 BiLSTM |
Transformers | 0.7081 roberta-base | 0.6801 xlm-roberta-base | 0.7007 bert-base-german-dbmdz-uncased |
