%A FERISCA N LUSIANA 
%T ANALISIS KINERJA AUGMENTASI TEKS BACK TRANSLATION DAN
SYNONYM REPLACEMENT DALAM KLASIFIKASI STATUS
KESEHATAN MENTAL MENGGUNAKAN DISTILLED-BERT
(DISTILBERT)
%X Mental health is an important aspect that supports the overall quality of life of
individuals. In today?s digital age, social media has become a space for people to
express their psychological conditions. This phenomenon opens up opportunities
to utilize text data as a source of information in the process of identifying mental
health disorders. This study aims to analyze the performance of two text-based
data augmentation techniques, namely back translation and synonym replacement,
in improving the performance of mental health status classification using the
DistilBERT model. The study was conducted on English-language text data labeled
according to the type of mental disorder. The synonym replacement and back
translation augmentation techniques were applied to balance the data distribution
in the minority class. The dataset was divided into 80% training data and 20%
test data, with 20% of the training data used as validation data. The classification
process was performed using a fine-tuned DistilBERT model. The results showed
that the DistilBERT model with synonym replacement augmentation achieved the
highest accuracy, namely 87%, while back translation achieved an accuracy of
86%. This indicates that the synonym replacement augmentation technique is more
effective in increasing data variation and classification model performance on the
dataset used in this study.
Keywords: Mental Health, Text Augmentation, Text Classification, DistilBERT.
%C UNIVERSITAS LAMPUNG
%D 2025
%I FAKULTAS MATEMATIKA ILMU PENGETAHUAN ALAM
%L eprints89318