An Integrated Framework for Fake Review Detection and Review Summarization on Product Reviews

The transformation in shopping culture from traditional markets to online markets has altered the purchasing behavior of customers with the assistance of online reviews. Nonetheless, the credibility of these reviews has been compromised by the existence of fraudulent reviews, resulting in poor buying decisions. The proposed work improves the credibility of reviews given on the internet and enhances the shopping experience of the consumers as they are able to use summarized reviews to make immediate decisions on the products. It presents a machine learning approach that tackles this problem by combining Fake Review Detection and Text Summarization as an effective information system. Uniquely, the proposed framework employs various criteria ranging from content, behavior, and length of the review to help provide a verdict that separates genuine and fraudulent reviews. Furthermore, intelligent techniques are incorporated to present reviews in short form but contains all the central and significant aspects of the review and all other details are omitted. The Bi-LSTM network for fake review detection achieved an accuracy of 94.1%, while the review summarization employs a pretrained T5-small model and a BLEU score of 0.64 is attained.