Measuring Review Credibility and Deception on E-Commerce Platforms through NLP and Chaos Theory

Resumo

Customer reviews on e-commerce sites strongly influence customers’ choices of products for purchasing but these reviews often include noise, exaggeration, or misleading language that make a product look better or worse than it really is. We propose NETER (Nonlinear Entropy-based Trust Evaluation Rating), a method for evaluating reliability and authenticity of online product reviews by handling non linearities in review text. NETER combines techniques likes sentiment analysis, deception detection, entropy-based stability and chaos-theory modeling to generate a trust score of product. Model uses RoBERTa for sentiment analysis to identify context and hidden emotions in review text accurately. Deception module identifies features like extreme praise, urgency words, or promotional phrases to check reliability of reviews. NETER uses statistical methods on review count to compute confidence weight so that products with few reviews do not receive unfairly high scores. To detect unstable, exaggerated or hyped review patterns, NETER applies chaos-theory where a logistic map is computed to determine Lyapunov exponent. Reviews showing hype patterns , unusual changes in their style or tone receive penalties while consistent sentiment is rewarded. Entropy and dispersion measure sentiments’ consistency. Thus, NETER creates a clearer and reliable review score in comparison with sentiments only traditional methods