Sentiment Analysis Using Facebook Comments

Abstract

Sentiment analysis of Bangla text on social media platforms like Facebook offers valuable insights into public opinion, sentiment trends, and behavior. Despite growing interest, challenges such as intricate morphology, script variations, and lack of wellannotated datasets restrict precise analysis. This study is interested in categorizing Bangla Facebook comments into positive, negative, and neutral sentiments using a dataset of 46,655 comments. Various machine learning and deep learning techniques like Logistic Regression, Random Forest, XGBoost, AdaBoost, SVM, and Neural Networks were employed in conjunction with text preprocessing methods like cleaning, tokenization, and feature extraction to enhance model performance. The research highlights the effectiveness of these techniques in managing linguistic complexity and improving prediction accuracy. The findings are of benefit to computational linguistics and social media analysis, offering practical applications to businesses, policymakers, and further research in data-driven decision-making. The study highlights the promise of sentiment analysis as a tool for understanding public opinion in Bangla and addressing core challenges in NLP for low-resource languages.