Sentiment Analysis and Topic Modelling for Academic Integrity in the Era of AI

Abstract

This study explores the sentiments and discussion topics of X/Twitter users regarding academic integrity in the era of artificial intelligence (AI). The approach incorporates sentiment analysis and topic modelling to reveal the public perspective on academic integrity issues, including plagiarism, online exams, and AI usage. Our study aims to provide a framework for exploring topics and findings related to the trend of academic integrity in the era of AI. In sentiment classification, Naive Bayes, support vector machine (SVM), and Random Forest algorithms are combined with vectorization techniques such as Count Vectorizer, Word Level TF-IDF, N-Gram TF-IDF, and Character Level TF-IDF. The results show that Naive Bayes with Count Vectorizer provides the best performance on imbalanced data. For the topic modelling, NMF proved to be the most effective in generating specific topics, such as plagiarism and AI detection, with the highest coherence scores. This study also examines the crucial role of each preprocessing step in enhancing data quality, which significantly impacts classification and topic modelling performance. The findings are expected to provide new insights into sentiment analysis and a deeper understanding of academic integrity issues in the era of artificial intelligence.