Topic Modelling of Student Suggestions on SIMAK UNISMUH Using BERTopic

Abstract

This study examines the use of the BERTopic algorithm for topic modeling on student feedback data collected through the SIMAK UNISMUH. The research aims to identify and visualize thematic patterns in student feedback to improve academic services and campus facilities. This study utilizes Natural Language Processing (NLP) techniques, particularly BERTopic, which combines the advantages of Bidirectional Encoder Representations from Transformers (BERT) with clustering algorithms to produce contextually rich and easily interpretable topic representations. The research data comprises 232,430 feedback entries that were processed to remove noise and irrelevant information, resulting in 26,009 valid entries. These entries were then processed using the BERTopic algorithm, generating nine distinct topics related to various aspects of academic life, including teaching methods, campus facilities, and administrative services. The coherence score of 0.637 indicates strong internal consistency within the identified topics, while the analysis reveals key areas where the university can enhance its services. The findings from this study provide actionable insights for university administrators, enabling them to make informed decisions and improve student academic performance. Additionally, this research contributes to the field of topic modeling in educational contexts and demonstrates the effectiveness of BERTopic in processing large-scale textual data.