Fake Business Review Detector
Model Description
This model was fine-tuned on DistilBERT to detect AI-generated fake restaurant reviews.
It classifies reviews as either real (0) or fake (1).
Developed by: Rakiya Nesbit
Model type: DistilBERT fine-tuned for Text Classification
Language: English
Finetuned from: distilbert-base-uncased
Training Data
137 labeled restaurant reviews — 40 real reviews collected from Yelp and Google,
and 97 AI-generated fake reviews created using ChatGPT.
Results
- Accuracy: 71%
- F1 Score: 0.83
- Recall: 1.0 (caught every fake review)
GitHub
