A Pay-Close-Enough-Attention Ensemble Classifier for Spotlighting Bogus Reviews

As a result of the proliferation of social media platforms and online shopping, an increasing number of individuals are sharing their shopping experiences on the internet and relying on the opinions of their contemporaries to direct their decision-making when it comes to making purchases. The growing reliance that people place on reviews to help them make judgments about the products and services they buy online has contributed to an increase in the number of instances in which reviews have been falsified. The accuracy of classification in the traditional method is dependent on the label dataset, which is not only challenging to collect and analyze, but also slow, laborious, and reliant on the participation of a human in the form of feature extraction. The length of the text that may be found online makes the model less sensitive to irregularities since the classifier is more likely to overlook important facts. We provide an attention-based semi-Ensemble Classifier as a solution to the problem that was stated before. This classifier is designed to detect fake reviews. Deep neural networks use both labeled and unlabeled data to properly learn input distributions, which ultimately results in enhanced prediction accuracy for online reviews. One way that this is accomplished is by automatically identifying the features that are required for classification. The application to the labeled data set provided by TripAdvisor highlights both the efficiency of the method’s attention mechanisms in the classifier for acquiring an adequate semantic representation and the method’s dependency on a limited training set for identifying false reviews. The results of the trials show that the suggested system can detect fake reviews with a higher level of accuracy than the state-of-the-art semi-supervised algorithms, even when the label dataset is very limited.