Accommodation Matching via Embedding Similarity

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing accommodation management systems are unable to identify equivalent homestay accommodations for bookings due to their unique nature, leading to inadequate alternatives for guests in case of cancellations, especially last-minute cancellations.

Innovation Solution

An accommodation management system that creates mappings between attributes of canceled bookings and equivalent listings in a database, allowing for automatic rebooking or recommendation of comparable listings with similar attributes, ensuring seamless replacement options for guests.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional database search is used to find equivalent accommodations, then exact attribute matching is achieved, but no suitable alternatives are found due to the unique nature of homestay accommodations

Engineering Contradiction:
Improveattribute matching accuracyVSAvoidaccommodation alternative identification
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system changes the approach from exact attribute matching to semantic similarity matching by transforming attribute data into embeddings that capture the essence of homestay features. This allows the system to find accommodations with similar characteristics even when exact attribute matches don't exist, resolving the contradiction between precise matching and adaptable alternative identification.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent introduces an intermediary layer (embedding space) between the database storage and the search query. Instead of directly comparing raw attributes, the system transforms both the stored accommodation data and search queries into embedding representations, enabling semantic similarity search that bridges the gap between exact matching requirements and the unique nature of homestays.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If detailed attribute tracking is implemented for each homestay, then accurate booking management is achieved, but system complexity increases due to the unique features of each accommodation

Engineering Contradiction:
Improvebooking management accuracyVSAvoiddata structure complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system segments the complex attribute data into manageable components by separating the detailed homestay attributes from the booking management logic. The embedding layer handles the complexity of unique feature representation, while the database stores simplified structured data, dividing the system into distinct functional layers that reduce overall complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Instead of storing every possible unique attribute combination for each homestay, the system creates a simplified representation (embedding) that captures the essential characteristics. This copying approach allows the system to manage detailed attributes without storing them all explicitly, reducing data structure complexity while maintaining booking management accuracy.

Inventive Principle:
Principle #26Copying

3Ease of operation

If manual search for replacement accommodations is required, then guest choice is maintained, but time consumption increases especially for last-minute cancellations

Engineering Contradiction:
Improveguest booking controlVSAvoidreplacement accommodation search time
Core Design Contradiction:
Ease of operationVSLoss of time

Solution Approach 1:

The system performs preliminary action by pre-computing embeddings for all homestay accommodations and maintaining an updated index of available properties. When a cancellation occurs, the system can immediately query the embedding space for similar accommodations without requiring manual search, significantly reducing search time while preserving guest choice through ranked recommendations.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system implements feedback mechanisms where guest preferences and booking histories are continuously fed back into the embedding generation process. This allows the system to learn and adapt to guest preferences over time, providing increasingly accurate and relevant accommodation recommendations that reduce search time while maintaining ease of operation.

Inventive Principle:
Principle #23Feedback

Data Source

PatentUS11580460B2Database systems for similar accommodation determination
Publication Date: 2023.02.14 AIRBNB INC
  • US11580460B2 patent drawing
  • US11580460B2 patent drawing
  • US11580460B2 patent drawing

AI summary

A system and a method are disclosed for replacing a listing of a canceled booking for a guest user. In an embodiment, an accommodation management system receives an indication that a booking by the guest user has been canceled. The system retrieves, from an attribute database, attributes of the canceled listing and maps each attribute to a set of similar attributes in an equivalence table. The system retrieves, from a listing database, a set of comparable listings, each of which includes one or more of the mapped similar attributes and available during the same time period as the canceled booking. The system books one of the comparable listings for the guest user during the time period by populating a booking data structure with a connection between the comparable listing and the guest user.