Multi-Label Age Classification of Indonesian Song Lyrics using Deep Learning

In today's music landscape, ensuring the appropriateness of song lyrics for different age groups is paramount due to the prevalence of mature themes. This study focuses on classifying Indonesian song lyrics into four distinct age categories: Dewasa (Adult), Remaja (Adolescents), Semua Usia" (all ages) and "Anak-anak" (Children). The motivation behind this research is to safeguard young listeners from exposure to unsuitable content. The study employs a dataset of 400 songs, collected from prior research efforts, to meticulously train and evaluate various models. Deep learning techniques are explored, with specific attention given to three different neural network designs: Bidirectional Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), and Simple Recurrent Neural Network (RNN). Training involves a data split of 80% for training and 20% for validation and testing. Notably, the Bidirectional LSTM model exhibits remarkable performance, achieving an accuracy of 97% in effectively categorizing lyrics by age. Following closely, the CNN model attains an accuracy of 96%, displaying its capacity to differentiate between age groups. Conversely, the Simple RNN struggles to capture intricate relationships, resulting in a 55% accuracy.