Hugging Face Datasets - Hotel Search

One search box over the machine-learning community's hotel shelf. A full-text query for 'hotel' against the Hugging Face Hub returned roughly 20-40 publicly listed dataset repositories when this record was cataloged in August 2026, contributed by companies, university labs and individual practitioners rather than one curator. The material splits into two working clusters and one long tail. The first cluster is booking demand: gemuchu/hotel_bookings and SIEP/hotel_bookings, both derived from the widely used resort-and-city-hotel booking demand corpus - about 119,000 bookings across 32 columns, carrying cancellation flags, lead times, average daily rate, market segment and reservation outcome. The second is reviews: argilla/tripadvisor-hotel-reviews, ashraq/hotel-reviews and patrickbdevaney/tripadvisor_hotel_reviews, ranging from 10K to 100K+ reviews each. The tail is robot-services and assistant datasets of marginal financial relevance, and honest inventory says so.

For hotel & resort REIT work the booking-demand subset earns its place: its columns mirror the operating metrics REITs disclose - ADR, segment mix, cancellation behavior, stay length. Get a sample of this dataset and we return rows shaped exactly like the dictionary below.

Two shapes share one shelf, so the sample shows both. One booking-level row from the 119,000-row demand corpus, then the catalog card of a review corpus:

record             : gemuchu/hotel_bookings
shape              : booking-demand corpus, 119,000 rows x 32 columns
hotel              : Resort Hotel
is_canceled        : 0            lead_time: 342 days
arrival            : July 1, 2015 (week 27)
stays_weekend_nights: 0

record             : argilla/tripadvisor-hotel-reviews
shape              : per-review text corpus, 10K<n<100K band
downloads          : 1,711        likes: 7        format: parquet

The first shape feeds models; the second feeds language work. One reading rule applies throughout: download counters are cumulative popularity signals, not freshness indicators - recency reads off modification dates, never off the counter.

Three core fields carry verified definitions and verified example values, captured during cataloging against live records rather than inferred from documentation. Everything else in the booking-demand schema is named, typed and defined in the second table below, with its example values folded under additional fields on request - confirmed against live records when your sample is prepared. The same rule covers the review corpora, whose text-plus-annotation structures vary by uploader, and the repository-level counters (downloads, likes, size bands) visible in the sample above.

Three chips summarize the footprint:

• Geography: the booking-demand corpus covers two European properties - one resort, one city; the review corpora span global TripAdvisor and hotel properties.
• Time frame: booking arrivals span 2015-2017; review corpora reflect their respective crawl dates; Hub listings shift continuously as uploads land or vanish.
• Granularity: per-booking records in the demand corpus, per-review records in the review corpora - nothing aggregates above the single reservation or the single review.

Scored honestly: this slice grades 6/10 against a Datadory catalog averaging 7.81 across 1,744 datasets. The deduction is stated plainly - it is a search-results surface rather than a curated corpus, the booking corpus reflects two properties rather than a portfolio, and its ADR and segment mixes therefore do not map one-to-one onto U.S. lodging REIT disclosures. As directional modeling material it is strong; as portfolio evidence it needs company-level sources alongside it.

API, files, or your warehouse. Daily, weekly, or hourly.

You pick the channel and the cadence; the field dictionary travels unchanged through all three. Record-level payloads suit pipelines watching for new hotel repositories entering the slice, bulk files suit overnight loads into a modeling environment, and warehouse tables suit teams joining booking behavior against their own RevPAR and occupancy series. Cadence changes are a settings conversation, not a re-integration project.

Ranked by how directly this slice answers the day job:

1. Lodging REIT and hotel equity analysts. Cancellation rates, booking windows, segment mix and ADR distributions built from transaction-level rows mirror the operating metrics management teams disclose - a way to stress-test reported numbers against observed booking behavior.
2. Revenue-management and asset-management teams. Lead-time curves and stay-length patterns drawn from 119,000 real reservations benchmark seasonality and booking-window assumptions for individual properties.
3. Machine-learning practitioners. A 32-column tabular corpus with a binary target is the textbook setup for cancellation prediction and demand classification, and the community has already built on it.
4. Market researchers. Guest-origin countries and distribution channels quantify how travelers reach hotels - inputs to channel-mix studies that pair naturally with OTA-side datasets.

Investors and quant researchers get transaction-level hospitality fundamentals to set against REIT disclosures - the closest thing to bottom-up lodging demand in the catalog's community-data tier. Data scientists and ML engineers get a clean, widely understood tabular target plus review text for language work, keyed by one uniform repository identifier. Market researchers and consultants get a census of what the open ML community has already assembled around hotels, useful for scoping bespoke collection budgets. Journalists and academics get citable, timestamped artifacts with visible provenance and per-record counters.

• The hotel & resort REITs data hub collects every dataset in the industry, this one included.
• Nareit Lodging/Resorts Sector Overview supplies sector-level REIT performance - the top-down series to join against this corpus's bottom-up demand rows.
• SEC EDGAR Company Filing Search grounds the operating metrics this booking corpus models in the disclosures themselves.
• 515K Hotel Reviews in Europe (Kaggle) is the larger review corpus at scale - compare it against the Hub's review sets before choosing one.
• Amadeus Hotel Search & Booking APIs adds commercial-grade rate and availability signals to set against community-built demand data.
• Yahoo Finance Quote Pages (Hotel REITs) prices the REITs whose fundamentals the booking-demand columns echo.
• Ranked placement lives in best hotel & resort REITs datasets.