Perforvisor: a 5★ review said "Needs to be more clean." Open models caught it.

Description and use

Perforvisor turns reviews into a knowledge graph and gives restaurant owners concrete actions, uncovering hidden complaints masked by average ratings.

Problem

Restaurant owners have hundreds of reviews but lack time to read them; star averages mask specific complaints.

Outcome

Perforvisor turns raw reviews into actionable insights, revealing hidden complaints that star ratings hide, and provides a weekly to‑do list grounded in customer language.

Workflow

• Connect business via Google Maps or paste reviews
• Perforvisor reads each review with Qwen3-4B to extract aspects, sentiment, and claims
• Builds a knowledge graph merging semantically identical claims
• Generates a verdict and up to three actions prioritized by complaint frequency
• Flags reviews with rating–text contradictions or near‑copy‑paste
• Compares the business with up to three competitors on the same aspects
• Answers ad‑hoc questions using retrieved reviews and citations

Evidence

Perforvisor reads every review with a small open-weight model, turns what customers say into a knowledge graph and gives the owner a verdict, up to three concrete actions and the reviews behind each one.

Source

Perforvisor flagged this one: ★★★★★ "Needs to be more clean."

Source

The model read the text as negative and the Trust page flagged it as "Star rating contradicts the text".

Source

Connect a business (search Google Maps, load a sample or paste reviews from any platform, one per line with an optional rating) and Perforvisor:

Source