The development of a sentiment analysis model for teaching and learning using text messaging based on the bert model

This study explores sentiment analysis in educational contexts, focusing on unravelling the inherent in online learning environments. It aims to develop and evaluate a sentiment analysis model for classifying students_ sentiments using data from text messaging platforms such as WhatsApp and Telegram. The challenge of sentiment analysis in educational contexts is twofold, which includes data-related issues and the effectiveness of analytical models. The primary concern is data-related issues, encompassing both the availability and the nature of the data. Educational text data are often unstructured, multilingual, and linguistically diverse, with frequent use of textisms, emojis, and mixed language expressions. These characteristics, combined with the scarcity of domain-specific datasets and the complexity of languages like Malay and English mixed language, hinder accurate sentiment detection. The research adopts systematic literature review methods, including pre-processing and developing the Bidirectional Encoder Representations from Transformers (BERT) model. This consists of pre-processing the text messenger data, finetuning sentiment analysis models, and evaluating the model. The results indicate that the proposed sentiment analysis with the BERT model dramatically enhances the accuracy rate in classifying up to 89% of the sentiments. The comparative analysis model of the BERT based sentiment classifier with other models, including Naive Bayes (84%), Support Vector Machine (83%), Random Forest (84%), and K-Nearest Neighbors (81%). Using this model educators can used it assist to analyse students_ sentiments from WhatsApp and Telegram message to support teaching and learning activities. However, future accountability may focus on fine-grained emotion detection, larger multilingual datasets, and multimodal analytics to enhance sentiment interpretation in educational environments.