Enhancing sentiment analysis of Online product reviews using Machine learning techniques

Abstract

The dramatic increase in Internet-based applications, including newlinesocial networking media platforms and blogs, has resulted in reviews newlineand comments with respect to daily activities. Sentiment Analysis (SA) newlineis the way of analysing and gathering opinions, thoughts, and newlineimpressions of people concerning different products, topics, services, newlineand subjects. The opinions of individuals, corporations, and newlinegovernments can be valuable for decision-making and information newlinegathering. But the SA and evaluation process faces a large number of newlineproblems. These challenges create impediments to accurately newlineinterpreting sentiments and determining the appropriate sentiment newlinepolarity. Sentiment analysis identifies and extracts subjective newlineinformation from the text using natural language processing and data newlinemining. Many real-time applications need SA for comprehensive study. newlineFor instance, product analysis, discover which qualities or components newlineof the product appeal to customers with respect to product quality. One newlineof the key challenges in SA is feature extraction. One of the most newlinecommonly used approaches include the Bag-of-Words (BoW) model, newlinewhere each review is represented as a vector of word frequencies, and newlineTF-IDF (Term Frequency-Inverse Document Frequency), which newlineconsidered the importance of words based on their frequency in the newlinereview and the entire dataset. Once the feature vectors are generated, a newlineMachine Learning (ML) algorithm is trained on labeled data to create a newlinepredictive model. Several algorithms have been successfully applied to newlineSA of online product reviews, including Naive Bayes (NBs), Support newlineVector Machines (SVM), Decision Trees (DTs), Random Forests (RFs), newlineand Logistic Regression (LRs). These algorithms learn patterns from the newlineextracted features and can classify new reviews into positive, negative, newlineor neutral sentiment categories.