Sentiment Polarity Detection on Bengali Book Reviews By Word2vec and Hybrid Machine Algorithm

Abstract

Despite sentiment analysis's significance in gauge public opinion, current methods frequently fail to account for Bangla and other languages sufficiently. Specifically, this work develops a Sentiment Polarity Detection on Bengali Book Reviews By Word2vec and Hybrid Machine Algorithm. A total of 2012 reviews, including both good and unfavorable assessments, were compiled from various sources. The study assesses several ML models, including Word2Vec, for feature extraction, RF, SVM, and a hybrid of the two. With superior recall, accuracy, and precision, the best model was the one that combined RF with SVM. Results showed that the most effective RF and SVM hybrid models scored 0.92%. On top of that, they took first place in the categorization report with an exceptional accuracy rate of 0.93%. These outcomes prove that it can easily handle complex Bangla text. Improved processing of the Bangla natural language is the goal of the suggested effort. It paves the way for more advancements, such as creating an AI-driven mobile app that can analyze sentiment in real time and make it more accessible than a web app