Consumer review analysis using NLP and data mining

Abstract

The popularity and development of the Internet have greatly increased the amount of information that is easily available. For a sizable portion of the population, it has evolved into the primary forum for opinion expression. These viewpoints can be discussed in reviews, online forums, and social media websites like Twitter and Facebook. Sentiment analysis and information extraction are made possible by the ease of access to all of these viewpoints, which is free. Sentiment analysis is concerned with automatically ascertaining, through the use of computer tools, if a reviewer's purpose is positive, negative, or neutral with regard to specific goods, services, and problems. Since they have such a positive impact on every particular organization, these strategies are rapidly gaining favor. As a result, we have researched multiple methods for classification, data preprocessing, and feature extraction in this study and compared the outcomes using the logistic regression, multinomial naive bayes, and other accuracy measures. The results of the simulation demonstrate that, when we have applied to a dataset with features that the model has retrieved, SVM is the most accurate classifier for the given data set.