Research Article

Vietnamese Sentiment Analysis under Limited Training Data Based on Deep Neural Networks

Table 3

The accuracy results of various data augmentation techniques for Vietnamese sentiment analysis based on machine learning classifiers.

DatasetsClassifiers(1)(2)(3)(4)(5)(6)(7)(8)

Dataset 1LR0.8500.8640.8480.8650.8630.8560.8580.858
SVM0.8590.8520.8120.8600.8650.8560.8610.831
OVO0.8540.8570.8540.8630.8640.8560.8620.858
OVR0.8540.8570.8540.8630.8640.8560.8620.858

Dataset 2LR0.7420.7430.7520.7540.7520.7510.7490.748
SVM0.7210.7410.7110.7540.7360.7460.7200.732
OVO0.7510.7450.7450.7560.7590.7470.7510.750
OVR0.7340.7310.7380.7410.7470.7380.7390.738

Dataset 3LR0.8260.8310.8320.8380.8310.8250.8280.826
SVM0.8240.8290.8250.8320.8310.8200.8260.826
OVO0.8230.8290.8270.8310.8290.8190.8250.826
OVR0.8230.8290.8270.8310.8290.8190.8250.826

Dataset 4LR0.8160.8300.8290.8290.8200.8140.8220.821
SVM0.8280.8260.8330.8310.8340.8170.8260.827
OVO0.8260.8260.8330.8290.8340.8160.8260.827
OVR0.8260.8260.8330.8290.8340.8160.8260.827

(1) Preprocessing techniques; (2) EDA; (3) sentence shuffling; (4) back translation; (5) syntax-tree transformation; (6) contextual substitution (w2v); (7) contextual substitution ( + ); (8) masked language model (PhoBERT).