| Business | Reviewer | Rating | Fake Score | Review Text | Top Engines |
|---|---|---|---|---|---|
| No flagged reviews yet. Run a scan first. | |||||
Six detection engines analyze each review. Scores combine into a weighted composite (0-100). Based on Robert Dove's thesis research on sentiment analysis and text mining (NW Missouri State).
Uses Bing Liu lexicon (6,789 words). Real reviews have mixed sentiment. Fake reviews are 100% positive or negative.
Uses DOVE Dictionary. Checks for hedging, discourse particles, hesitators, pronouns, time adverbials. Real reviews have 3+ categories.
Groups reviews by date. Flags 5+ reviews on same day or 3+ in a 2-day window vs normal rate.
Analyzes reviewer patterns: generic phrases, minimal text with 5 stars, excessive punctuation, all-caps.
TF-IDF vectorization + cosine similarity. Flags review pairs with >0.7 similarity (copy/paste or templates).
Checks sentence length uniformity, vocabulary diversity (TTR), punctuation patterns, formal transition words.