The Role of Automated Classification in Preserving Indonesian Folk and National Songs

Aji Prasetya Wibawa, AH. Rofi'uddin, Rafal Drezewski, Ilham Ari Elbaith Zaeni, Irfan Zuhdi Abdillah, Triyanti Simbolon, Fabyan Raif Erlangga, Agung Bella Putra Utama

Abstract: The preservation of Indonesia's rich tapestry of language and culture is a significant concern, particularly in response to claims made by other countries about cultural treasures, such as folk music. Preserving cultural aspects is now easier in our tech-savvy era with abundant information. A potentially effective strategy entails the methodical categorization of traditional and patriotic songs, thereby facilitating the determination of their respective sources. The present work utilizes a dataset of 480 folk and 90 national songs, categorized into situations involving 2, 4, and 31 classes. A meticulous pre-processing pipeline is utilized, including cleaning, case folding, tokenization, and word weighting. The selected word weighting strategy is the TF-IDF. In order to mitigate the issue of class imbalance, the SMOTE is proposed as a supplementary approach. This study assesses three classification algorithms, namely KNN, SVM, and Naïve Bayes, using the 10-fold validation technique. The SVM, known for its utilization of hyperplane-based classification, demonstrates exceptional performance by reaching a remarkable accuracy rate of 99.69% while utilizing the RBF kernel function. The present study observed that the KNN demonstrated superior performance when the value of k was set to 2, and no SMOTE was applied. A binary classification methodology was employed. Notably, this strategy yielded an impressive accuracy rate of 97.02%. It is worth mentioning that Naïve Bayes demonstrates ideal performance in a two-label scenario when SMOTE is applied, yielding an accuracy rate of 93.75%. This comprehensive investigation proves that the SVM performs better than KNN and Naïve Bayes when classifying Indonesian song lyrics. This finding highlights the significance of SVM as a beneficial instrument for preserving Indonesia's culturally diverse legacy. This initiative connects technology and culture by providing educational resources and enlisting tribe elders, linguists, and other experts to demonstrate the integration of technology and tradition and preserve Indonesia's cultural heritage.