Sentimental Analysis on IMDb Movies Review using BERT

Abstract

The sentiment analysis is a system that is used to perform automated analysis processes on various services and product reviews. The Internet Movie Database (IMDb), which is categorized using Bidirectional Encoder Transformers (BERT), is included in this scheme. However, the sentiment analysis system is segregated into three stages: pre-processing, feature extraction, and sentiment classification. To extract features, Word Embedding using Word to Vector (Word2Vec) is employed. The data is then classified using a BERT-based test confusion matrix with settings for accuracy, recall, precision, and F1-Score. The test results showed that, the proposed scheme attain precision of 96.10%, recall of 88.40% and F1-Score of 92.10% for the negative class. Subsequently, precision of 89.20%, recall of 96.40% and F1-Score of 92.60% for the positive class using confusion matrix. Also overall 92.40% of accuracy is achieved to depict that the proposed scheme is an effective and reliable technique to detect sentiments for movie reviews.