AI Model Combines Text and Image Analysis to Detect Fake Reviews

A new AI model is revolutionizing the fight against fake online reviews by integrating text and image analysis. Developed by Suhasnadh Reddy Veluru, Sai Teja Erukude, and Viswa Chaitanya Marella, the system utilizes BERT for textual encoding and ResNet-50 for visual feature extraction. This multimodal approach enables the model to accurately classify review authenticity, addressing a critical gap in current detection methods.

The Multimodal Approach

Unlike traditional models that rely solely on textual data, this AI model analyzes both text and accompanying images to detect inconsistencies. This dual analysis significantly improves the detection of fake reviews, which often include manipulated or unrelated images to deceive users.

Training and Performance

The model was trained on a curated dataset of over 20,000 user-uploaded images across domains like food delivery, hospitality, and e-commerce. Experimental results demonstrate its superior performance, achieving an F1-score of 0.934 on the test set, outperforming unimodal baselines.

Implications for Digital Trust

This innovation offers a scalable solution for content moderation, helping to safeguard digital trust across online platforms. By accurately identifying fake reviews, the model can prevent misinformation and enhance the reliability of user-generated content.