Domain Knowledge Catalog

Travel & Hospitality Knowledge Catalog for LLMs

Equip travel agents, concierge bots, and booking LLMs with authoritative, structured context covering global aviation, transit hierarchies, visa policies, lodging taxonomies, and booking API schemas.

LLM Optimization: Token-compressed markdown chunks and JSON-LD entity graphs designed for high-precision retrieval with zero hallucinated airport codes or transit regulations.

Core Catalog Modules

Entities

Aviation & Transit Ontologies

IATA/ICAO airport codes, timezone offsets, terminal transfer thresholds, baggage limits by carrier alliance, and codeshare networks.

Policies

Immigration & Border Regulations

Passport validity rules (e.g., 6-month rule), transit visa waivers, reciprocity agreements, and biometric entry systems indexed by passport country of issuance.

Hospitality

Accommodation & Amenity Classifications

Standardized room typology (Deluxe, Suite, Connecting), accessibility ratings (ADA compliant features), pet policies, and resort fee disclosures.

Itinerary Logic

Seasonal & Geographic Constraints

Monsoon seasons, high-altitude acclimatization windows, national holidays, peak tourist closures, and realistic multi-city routing topologies.

Sample Context Representation (JSON-LD)

How entities in this catalog are formatted for minimal context-window consumption and maximum tool-calling clarity:

{
  "@context": "https://tensorknowledge.io/schema/travel/v1",
  "@type": "AirportEntity",
  "iata": "HND",
  "icao": "RJTT",
  "name": "Tokyo Haneda International Airport",
  "location": { "city": "Tokyo", "country": "JP", "tz": "Asia/Tokyo" },
  "terminals": ["T1", "T2", "T3"],
  "minimum_connection_time_minutes": {
    "domestic_to_domestic": 35,
    "domestic_to_international": 70,
    "international_to_international": 60
  },
  "ground_transit": ["Keikyu Line", "Tokyo Monorail", "Limousine Bus"]
}

Retrieval & Tool Specs

Tool-call Schema: get_itinerary_feasibility(origin, destination, dates, traveler_profile)
RAG Partitioning: Pre-chunked into 256-token discrete semantic units with parent document breadcrumbs
Evaluations: 1,200 curated multi-hop benchmark questions testing ambiguous layovers and visa transits