Fig 1.
Methodology steps.
Table 1.
Evaluation of machine translation quality for back-translated datasets using BLEU and METEOR scores.
Fig 2.
Training strategy.
Table 2.
Average runtime per fold (in seconds) for each model and training strategy.
Fig 3.
Box plot comparing F1-scores for the Logistic Regression classifier.
Fig 4.
Box plot comparing F1-scores for the Random Forest classifier.
Fig 5.
Box plot comparing F1-scores for the Support Vector Machine classifier.
Fig 6.
Box plot comparing F1-scores for the RNN-LSTM classifier.
Table 3.
Wilcoxon matched pairs test (Logistic Regression classifier).
Table 4.
Wilcoxon matched pairs test (Random Forest classifier).
Table 5.
Wilcoxon matched pairs test (Support Vector Machine classifier).
Table 6.
Wilcoxon matched pairs test (RNN-LSTM classifier).
Table 7.
Descriptive statistics of classifier with the RoBERTa model performance on original and back-translation datasets by F1-score metric.
Fig 7.
Box plot comparing F1-scores for the fully connected classifier with the RoBERTa model.
Table 8.
Wilcoxon matched pairs test (RoBERTa model classifier).