Tripin
A hotel search navigator and vacation ideas explorer powered by TripAdvisor reviews. A smart system that turns review chaos into clear recommendations.
"Choosing a hotel is about whether it fits you personally. We help you find that in minutes." — Tripin philosophy
About the project
This was our collaboration with a partner who built the project on his own life experience. For him, it was crucial to find hotels where guests write that the linen is truly clean and fresh every single time. With that knowledge, a traveler might even prefer a three-star hotel over a five-star one, if nearby five-stars have complaints about linen or amenities. Picture this: you are planning a vacation. You need a hotel. You open TripAdvisor and see... 847 reviews. Five stars — and someone writes about noise, someone about old furniture, someone is delighted. How do you know if the place fits you? Tripin solves this through review-based search. We analyze thousands of reviews, find patterns and turn them into clear information: is the hotel quiet, is the location convenient, is the service good. Real characteristics that help you decide in minutes. The hotel-choice problem is familiar to every traveler. Hundreds of reviews, dozens of sites, conflicting data — there is too much information, and much of it contradicts itself. What feels "excellent" to one person is "unbearable" to another. Time drains away on reading reviews when it could go to planning the trip itself. Review-based search uncovers things that aggregator filters simply miss. You can spot renovations, check the freshness of reviews. Situational details: when three weeks ago a guest left a note about cleanliness or frequency in the room, and a few others confirmed it — that means something. At the very least, we avoid deciding for the user; we highlight: take a look, pay attention, here are the hotels we found. Review-based search is honest, constrained, and opens a different layer of interaction with data. Things that real people notice, things that writers of paid reviews tend to miss. And when guests share from the heart — something irritates them, something delights them — the signal carries far more truthful information. That is why this service exists for precisely this kind of geeky search.
Gallery (grid)
How it works
Tripin understands reviews. The system processes reviews in different languages and extracts key themes: location (proximity to the center, transport, street noise), conditions (cleanliness, comfort, room state), service (staff, breakfast, extra amenities), surroundings (beach, view, nearby infrastructure). You set your priorities — quiet, location or service — and the system finds hotels where those traits get praised. It picks options with convenient transport access. It highlights places with outstanding staff. All of it runs on real reviews.
Two images (block two)
Technology
We gathered 500 thousand reviews for the Phuket location — far from exhaustive; coverage of the popular spots was reasonable. We parsed them, digitized them, built hotel cards, located them in other databases and pinned them to the map. The project uses hybrid search: vector search and index search (Vespa). Hybrid search compares results from these two groups and merges them, picking the most suitable matches through a weighting system. More reviews on the topic — higher coefficient. A special match on the nuance of the user's query — increased weight for that hotel. The system works in different modes: you can read reviews or simply receive a list of hotels and dive in from there. When mathematically there are more reviews in this model, the coefficient rose. When a special match appears on a query nuance, the hotel gained more weight. The prototype performed excellently: it found exactly what was needed, thanks to the system where vector search, index search and hybrid search work together.
Wide image
Project status
The project sits in development — a kind of "unfunny development," since it is not a current priority. We would still like to make the experiment itself public, so that even with this frozen collection people could see how it works. And perhaps it will grow into a commercial service. Compact CRM system in Telegram. Client, task and sales management right in the messenger without complex interfaces.
Compact CRM