ENG
INA
85th Doctoral Promotion
The Study Program of Doctoral Engineering Science (PSDIT) held its 85th doctoral promotion on Thursday, 27 February 2025, at the Wismakarma Hall, Faculty of Engineering, Udayana University on Jl. P.B. Sudirman Denpasar. Promovendus I Putu Gede Hendra Suputra, S.Kom., M.Kom with the dissertation titled Detection and Classification of Cognitive Distortion Using Indobert, Keyword Extraction and Part-Of-Speech Tagging for Indonesian Text received a very satisfactory predicate.
Dr. I Putu Gede Hendra Suputra, S.Kom., M.Kom describes that Cognitive distortions are often expressed in a person's speech and writing. Distorted thinking is very important to detect early; otherwise, it will put a person at risk of developing major depression. Cognitive Behavioral Therapy (CBT) is the most popular domain in psychotherapy research for depression. Various studies on detecting and classifying cognitive distortions have been conducted with machine learning models. Cognitive distortion detection is a type of short text classification that has the weakness of the lack of features that can represent the meaning or intent of a text. BERT is still one of the models that works consistently in various cases. This study proposes a new model to detect and classify cognitive distortions in Indonesian texts. The model is built by combining IndoBERT, Class Topic Modeling (CTM) and Part-of-Speech (POS) Rule or what in this study will be referred to as the IndoBERT-CTM-POSR model. The CTM and POSR proposed in this study are representations of the feature enrichment process shown to overcome problems in short text classification.
The combination of the two IndoBERT+CTM features in detecting distortion showed positive results, namely achieving an average accuracy value of 0.787 and an F1 value of 0.769. Accuracy and F1 values increased by 3.39% and 3.45% compared to the IndoBERT detection model. This increase in detection results will be useful as an early detection tool to support the online CBT program. In classifying 10 distortion classes + 1 non-distortion class, the model developed based on the combination of IndoBERT+CTM or IndoBERT+POSR features can increase the accuracy and F1 values, but these values do not increase significantly. The results of the 11-class classification evaluation tend to be small, with an average accuracy of 0.523 and an F1 of 0.528. The addition of CTM and POSR features to the IndoBERT model always provides an increase in accuracy and F1 values in the classification of cognitive distortion. However, it should be noted that the accuracy value tends to be small and insignificant, so that for further development, it is necessary to conduct trials on features from text/sentences such as the extraction of Named Entity Recognition (NER) or Relation Extraction (RE) features.
As chairman of the session is the Coordinator of PSDIT, Prof. Dewa Made Priyantha Wedagama, ST., MT., MSc., Ph.D. The promoter, co-promoters I and II are Prof. Ir. Linawati, M.Eng.Sc., Ph.D., Dr. Gede Sukadarmika, S.T., M.Sc, and Dr. Nyoman Putra Sastra, S.T., M.T, respectively. Acting as examiners are Prof. Dr. Ir. I Made Oka Widyantara, S.T., M.T. IPU., ASEAN.Eng., Prof. Dr. I Made Sukarsa, S.T., M.T., Prof. Dr. Ir. Nyoman Gunantara, S.T., M.T., Prof. Ir. Ni Made Ary Esta Dewi Wirastuti, S.T., M.Sc., Ph.D., IPU, and Dr. Dewa Made Wiharta, S.T., MT. Academic invitees are Dr. Ngurah Indra ER, S.T., M.Sc., Ir. Komang Oka Saputra, S.T., M.T., Ph.D., and Dr. Ir. I Ketut Gede Suhartana, S.Kom., M.Kom., IPM , ASEAN Eng (Faculty of Math and Natural Science, Udayana University)
FACULTY OF ENGINEERING