Psychological-Based Opinion Mining for Fake Review Detection Using Advanced NLP and Machine Learning Techniques

The recent adoption of online review systems has had an enormous impact on consumer behavior, especially within the scope of the online marketplace. Even though such reviews are useful to prospective users, the increasing amount of fake reviews present great problems to both e-commerce sites and consumers. In this paper, an original method of fighting fake reviews is proposed through the application of psychological and linguistic patterns in addition to standard methods of machine learning. Although there are some methods which are based on a review or word-frequency analysis done in a checklist format, our system uses state-of-the-art natural-language processing (NLP) techniques and learns algorithms to penetrate the more generalized meaning and emotion of a review. The system uses language analysis and sentiment analysis and metric of cognition and emotion consistency provided by Linguistic Inquiry and Word Count LIWC for enhancement of accuracy. Testing on various datasets showed substantial gains in detection accuracy over conventional techniques. Our approach managed to outdo the previous best solutions with accuracy level of between 82.5 to 77.5. The application of psychological analysis to the integration of fake review detection is a significant development in this technology.