Multilingual sentiment analysis of university social media posts using BERT and machine learning classifiers
Zainab Jameel, Abbas Abdulazeez Abdulhameed, Yasmin Makki Mohialden · IET conference proceedings. · 2025
BERT (Bidirectional Encoder Representations from Transformers) and machine learning classifiers are used in this study to improve sentiment classification accuracy for Arabic and English university social media postings. Applying BERT-based sentiment analysis and classical machine learning techniques to synthetic posts and comments to categorize words is presented. It uses Naive Bayes, Random Forest, SVM, and K-Nearest Neighbors. Every classifier was trained on a vocabulary of positive and negative words to predict new word attitudes. The BERT model was utilized in order to assign ratings to the sentiments expressed in the comments.Sentiment recognition accuracy is improved with BERT and machine learning classifiers in comparison to more traditional methods. It was possible to compute and visualize post sentiment using the way that was suggested. More validation of the combined method was achieved by the analysis of sentiment score distribution and frequency. It is clear from these findings that the hybrid technique is capable of providing extensive sentiment analysis in situations that involve more than one language.