airbnb-aspect-based-improvement-extractor

Airbnb Review Improvement Extractor

This model extracts actionable improvement suggestions from Airbnb guest reviews, per predefined aspect (e.g. Cleanliness, Noise, Communication).

It is designed to help hosts, property managers, and analytics tools quickly understand what needs to be improved, without generating hallucinated positives.

---

What this model does

Given:

• a full Airbnb review
• a specific aspect (e.g. Cleanliness, Noise / Quietness)

The model returns:

• a short list of concrete improvements, or
• an explicit empty result if the review contains no actionable issue for that aspect.

The model is conservative by design:
if no clear problem is mentioned, it returns nothing.

---

Supported Aspects

The model operates on the following canonical Airbnb-style aspects:

• Cleanliness
• Communication
• Location
• Value / Price
• Accuracy / Description Match
• Check-in / Convenience
• Amenities / Comfort
• Safety / Security
• Noise / Quietness
• Wifi
• Host Behavior / Professionalism / Respect

Each aspect is evaluated independently.

---

Example Usage (Python)

from train.train_canonical_classifier_improve import ImprovementGenerator

CANONICAL_ISSUES = [
    "Cleanliness",
    "Communication",
    "Location",
    "Value / Price",
    "Accuracy / Description Match",
    "Check-in / Convenience",
    "Amenities / Comfort",
    "Safety / Security",
    "Noise / Quietness",
    "Wifi",
    "Host Behavior / Professionalism / Respect",
]

model = ImprovementGenerator()
model.load("insightful-stays/airbnb-review-improvement-extractor")

review = """
Not a bad place, but there were no curtains in the shower, the bathroom was often wet,
and communication was difficult because the hosts did not speak English.
"""

for aspect in CANONICAL_ISSUES:
    result = model.generate(review=review, aspect=aspect)
    print(aspect, result)