Analisis Tren Pemikiran Islam melalui Publikasi Ilmiah Berbasis Teknik LDA dan BERTopic

This study uses topic modeling techniques, specifically Latent Dirichlet Allocation (LDA) and BERTopic, to examine trends in Islamic thought in Indonesian scholarly publications from 2019 to 2023. The findings demonstrate that the LDA model frequently produces general and recurring themes like learning, research, and education that show little semantic variation across years. On the other hand, BERTopic finds themes that are more contextually specific, such as local educational institutions, religious tourism, and halal products. Performance-wise, BERTopic continuously produces higher coherence scores, reaching a peak of 0.979 in 2023, while LDA's highest score was 0.679. These results demonstrate the superiority of BERTopic in capturing the changing discourse of Islamic studies over time, in addition to its superiority in generating organized and interpretable topics.