Combining neural and semantic features in the analysis of being supportive in online feedback from customers

Abstract

Over the past ten years, there has been a notable increase in the number of individuals accessing the internet. Positive evaluations serve as social evidence, convincing future purchasers of the product's quality and advantages. They can impact purchase decisions by offering real-world user information. Good reviews increase a product's or brand's trust and reputation. Customers are more inclined to buy from a firm that has received excellent feedback since it demonstrates dependability and contentment. Reviews can be considered user-generated content since they emphasise different applications, features, or advantages associated with a product. This material has the potential to persuade indecisive shoppers. The Yelp website was utilised to scrape feedback data for all Asian restaurants in New York City, which was then trained and assessed using three different models like Navie Bayes, next one is Logistic Regression, and then finally with Support Vector Classifiers. The Logistic Regression classifier outperformed the others by having the lowest proportion of mistakes and the highest Area under the ROC Curve noted as AUC on the receiver operating characteristic curve ROC curve. Commercial insights were gathered by recognising the existence of highly significant phrases while contrasting how they performed to the universal probabilities when the machine learning system was given review data from my restaurant.