Klasifikasi Moral Tokoh Cerita Rakyat Bali Menggunakan Named Entity Recognition (NER)

Upadhana, Nyoman Satyawikrama (2026) Klasifikasi Moral Tokoh Cerita Rakyat Bali Menggunakan Named Entity Recognition (NER). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221194-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Cerita rakyat merupakan warisan budaya dan mengandung nilai moral serta etika yang relevan dengan kehidupan sehari-hari di Bali. Analisis nilai moral dan etika dalam cerita rakyat Bali terhambat oleh ketidakkonsistenan dalam penamaan tokoh. Minimnya sumber daya atau resource yang membahas nilai moral dalam cerita rakyat Bali, baik dalam bentuk dataset beranotasi, kamus moral, maupun pengembangan model. Penelitian ini berfokus pada klasifikasi moral tokoh dalam cerita rakyat Bali menggunakan Named Entity Recognition (NER) yang akan membantu proses ekstraksi tokoh. Penerapan NER membantu ekstraksi nama-nama tokoh yang sering muncul dalam berbagai bentuk, seperti gelar, epitet, dan variasi ejaan. Untuk menyederhanakan variasi nama tokoh digunakan metode alias clustering yang menyatukan berbagai penyebutan tokoh. Selanjutnya hasil cluster digunakan untuk klasifikasi tipe protagonis atau antagonis karakter pada tingkat kalimat. Kemudian Moral Foundations Theory (MFT) dengan enam polaritas, Care/Harm, Fairness/Cheating, Loyalty/Betrayal, Authority/Subversion, Sanctity/Degradation, dan Liberty/Oppression, digunakan dalam klasifikasi moral untuk mengkategorikan tokoh. Metode pengklasifikasi memanfaatkan machine learning dan deep learning, termasuk model BERT untuk mengenali tipe dan nilai moral tokoh dalam cerita rakyat. Evaluasi dilakukan menggunakan metrik akurasi, presisi, recall, dan F1-score untuk menilai model klasifikasi moral. Pada tahap ekstraksi tokoh menggunakan NER, model terbaik adalah mBERT dengan F1-score sekitar 0,7760. Kemudian pada alias clustering, metode Jaro Winkler dengan word sense mapping (WSM) pada threshold 0,85 mencapai F1-score 0,5864. SVM dengan empat fitur mencapai F1-score 0,8100 pada tahap klasifikasi tipe tokoh. Pada tahap klasifikasi moral, performa terbaik diperoleh oleh Logistic Regression penambahan fitur keyword dengan F1-score 0,3150.
====================================================================================================================================
Folktales are cultural heritage that contain moral and ethical values relevant to everyday life in Bali. The analysis of moral and ethical values in Balinese folktales is hindered by inconsistencies in character naming. In addition, there are limited resources that discuss moral values in Balinese folktales, including annotated datasets, moral lexicons, and model development. This research focuses on the moral classification of characters in Balinese folktales using Named Entity Recognition (NER) to support the character extraction process. The application of NER helps extract character names that often appear in various forms, such as titles, epithets, and spelling variations. To simplify these variations, alias clustering is used to unify different references that refer to the same character. The resulting clusters are then used to classify character types as protagonists or antagonists at the sentence level. Furthermore, Moral Foundations Theory (MFT), consisting of six polarities—Care/Harm, Fairness/Cheating, Loyalty/Betrayal, Authority/Subversion, Sanctity/Degradation, and Liberty/Oppression—is applied in the moral classification process to categorize characters. The classification methods utilize machine learning and deep learning approaches, including BERT-based models, to identify character types and moral values in folktales. Evaluation is conducted using accuracy, precision, recall, and F1-score to assess the performance of the moral classification model. In the character extraction stage using NER, the best-performing model is mBERT, achieving an F1-score of approximately 0.7760. In the alias clustering stage, the Jaro-Winkler method with Word Sense Mapping (WSM) at a threshold of 0.85 achieves an F1-score of 0.5864. For character type classification, SVM with four features achieves an F1-score of 0.8100. In the moral classification stage, the best performance is obtained by Logistic Regression with the addition of keyword features, achieving an F1-score of 0.3150.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Named Entity Recognition, Moral Foundations Theory, Aliases Clustering, Cerita Rakyat Bali, Machine Learning, Deep Learning, Named Entity Recognition, Moral Foundations Theory, Aliases Clustering, Balinese Folktales, Machine Learning, Deep Learning
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Nyoman Satyawikrama Upadhana
Date Deposited:	24 Jul 2026 07:05
Last Modified:	24 Jul 2026 07:05
URI:	http://repository.its.ac.id/id/eprint/137506

Actions (login required)

View Item