# US 2020/0394199 A1 · Search result optimization using machine learning models

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2020/0394199 A1
- Document ID: US20200394199A1
- Status: Published application
- Kind: A1
- Application number: 16/441,952
- Filed: 2019-06-14
- Published: 2020-12-17
- Earliest claimed: 2019-06-14
- Assignees: Airbnb, Inc.
- Applicants: Airbnb, Inc.
- Inventors: Malay Haldar
- Web page: https://patentgrep.com/patent/US20200394199A1

## CPC

- G06F 16/24578: {using ranking}
- G06F 16/9535: Search customisation based on user profiles and personalisation
- G06F 16/248: Presentation of query results
- G06N 3/045: Combinations of networks
- G06N 3/0499: Feedforward networks
- G06N 3/08: Learning methods
- G06N 3/09: Supervised learning
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Systems and methods are provided for search result optimization using machine learning models. A search system uses machine learning models generate a target vector based on query features of a search query and a set of listing vectors based on listing features of listings identified as part of the search query. The target vector represents an estimated optimal listing for the search query and each listing vector represents a corresponding listing identified as part of the search query. The search system determines distances (e.g., Euclidian distance) between each listing vector and the target vector. The determined distances indicate how similar each listing is to the estimated optimal listing for the search query. The search system ranks the listings based on the distances such that listings that are similar to the estimated optimal listing are ranked higher than listing that are not similar to the estimated optimal listing.

## Claims

### Claim 1

A method comprising:
receiving, from a client device, search parameters defining a first search query;
generating, based on query features of the first search query, a first target vector representing an estimated optimal listing for the first search query, the first target vector generated using a first machine learning model;
identifying at least a first listing and a second listing based on the first search query;
generating, based on listing features of the first listing, a first listing vector representing the first listing, the first listing vector generated using a second machine learning model that is different than the first machine learning model;
generating, based on listing features of the second listing, a second listing vector representing the second listing, the second listing vector generated using the second machine learning model;
determining a first distance measured between the first listing vector and the first target vector;
determining a second distance measured between the second listing vector and the first target vector; and
ranking, based on the first distance and the second distance, a set of listings for the first search query, the set of listing including the first listing and the second listing.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
generating the first machine learning model based on query features of historical search queries and listing features of selected listings of the historical search queries.

### Claim 3 (depends on 2)

The method of claim 2, further comprising:
generating the second machine learning model based on target vectors generated by the first machine learning model for the historical search queries, listing features of the selected listings of the historical search queries, and listing features of unselected listings of the historical search queries.

### Claim 4 (depends on 3)

The method of claim 3, wherein the second machine learning model includes a first neural network and a second neural network, and generating the second machine learning model comprises:
generating, based on listing features of a first selected listing, a selected listing vector representing the first selected listing, the selected listing vector generated using the first neural network, the first selected listing corresponding to a first historical search query;
generating, based on listing features of a first set of unselected listings, an unselected listing vector representing the first set of unselected listings, the unselected listing vector generated using the second neural network, the first set of unselected listings corresponding to the first historical search query;
generating, based on query features of the first historical search query, a training target vector representing an estimated optimal listing for the first historical search query, the training target vector generated using a first machine learning model;
determining a first training distance measured between the selected listing vector and the training target vector;
determining a second training distance measured between the unselected listing vector and the training target vector; and
adjusting weights of the first neural network and the second neural network based on the first training distance and the second training distance.

### Claim 5 (depends on 4)

The method of claim 4, wherein adjusting weights of the first neural network and the second neural network comprises:
determining that the first training distance is longer than the second? training distance; and
adjusting the weights of the first neural network and the second neural network to result in a subsequent iteration of the first training distance being shorter than a subsequent iteration of the second training distance.

### Claim 6 (depends on 1)

The method of claim 1, wherein ranking the set of listings for the first search query comprises:
determining that the first distance is shorter than the second distance; and
ranking the first listing higher than the second listing based on determining that the first distance is shorter than the second distance.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
causing presentation of the listing results on a display of the client device, the listing results presented according to the ranking.

### Claim 8

A system comprising:
one or more computer processors; and
one or more computer-readable mediums storing instructions that, when executed by the one or more computer processors, cause the system to perform operations comprising:
receiving, from a client device, search parameters defining a first search query;
generating, based on query features of the first search query, a first target vector representing an estimated optimal listing for the first search query, the first target vector generated using a first machine learning model;
identifying at least a first listing and a second listing based on the first search query;
generating, based on listing features of the first listing, a first listing vector representing the first listing, the first listing vector generated using a second machine learning model that is different than the first machine learning model;
generating, based on listing features of the second listing, a second listing vector representing the second listing, the second listing vector generated using the second machine learning model;
determining a first distance measured between the first listing vector and the first target vector;
determining a second distance measured between the second listing vector and the first target vector; and
ranking, based on the first distance and the second distance, a set of listings for the first search query, the set of listing including the first listing and the second listing.

### Claim 9 (depends on 8)

The system of claim 8, the operations further comprising:
generating the first machine learning model based on query features of historical search queries and listing features of selected listings of the historical search queries.

### Claim 10 (depends on 9)

The system of claim 9, the operations further comprising:
generating the second machine learning model based on target vectors generated by the first machine learning model for the historical search queries, listing features of the selected listings of the historical search queries, and listing features of unselected listings of the historical search queries.

### Claim 11 (depends on 10)

The system of claim 10, wherein the second machine learning model includes a first neural network and a second neural network, and generating the second machine learning model comprises:
generating, based on listing features of a first selected listing, a selected listing vector representing the first selected listing, the selected listing vector generated using the first neural network, the first selected listing corresponding to a first historical search query;
generating, based on listing features of a first set of unselected listings, an unselected listing vector representing the first set of unselected listings, the unselected listing vector generated using the second neural network, the first set of unselected listings corresponding to the first historical search query;
generating, based on query features of the first historical search query, a training target vector representing an estimated optimal listing for the first historical search query, the training target vector generated using a first machine learning model;
determining a first training distance measured between the selected listing vector and the training target vector;
determining a second training distance measured between the unselected listing vector and the training target vector; and
adjusting weights of the first neural network and the second neural network based on the first training distance and the second training distance.

### Claim 12 (depends on 11)

The system of claim 11, wherein adjusting weights of the first neural network and the second neural network comprises:
determining that the first training distance is longer than the second? training distance; and
adjusting the weights of the first neural network and the second neural network to result in a subsequent iteration of the first training distance being shorter than a subsequent iteration of the second training distance.

### Claim 13 (depends on 8)

The system of claim 8, wherein ranking the set of listings for the first search query comprises:
determining that the first distance is shorter than the second distance; and
ranking the first listing higher than the second listing based on determining that the first distance is shorter than the second distance.

### Claim 14 (depends on 8)

The system of claim 8, the operations further comprising:
causing presentation of the listing results on a display of the client device, the listing results presented according to the ranking.

### Claim 15

A non-transitory computer-readable medium storing instructions that, when executed by one or more computer processors of a computing system, cause the computing system to perform operations comprising:
receiving, from a client device, search parameters defining a first search query;
generating, based on query features of the first search query, a first target vector representing an estimated optimal listing for the first search query, the first target vector generated using a first machine learning model;
identifying at least a first listing and a second listing based on the first search query;
generating, based on listing features of the first listing, a first listing vector representing the first listing, the first listing vector generated using a second machine learning model that is different than the first machine learning model;
generating, based on listing features of the second listing, a second listing vector representing the second listing, the second listing vector generated using the second machine learning model;
determining a first distance measured between the first listing vector and the first target vector;
determining a second distance measured between the second listing vector and the first target vector; and
ranking, based on the first distance and the second distance, a set of listings for the first search query, the set of listing including the first listing and the second listing.

### Claim 16 (depends on 15)

The non-transitory computer-readable medium of claim 15, the operations further comprising:
generating the first machine learning model based on query features of historical search queries and listing features of selected listings of the historical search queries.

### Claim 17 (depends on 16)

The non-transitory computer-readable medium of claim 16, the operations further comprising:
generating the second machine learning model based on target vectors generated by the first machine learning model for the historical search queries, listing features of the selected listings of the historical search queries, and listing features of unselected listings of the historical search queries.

### Claim 18 (depends on 17)

The non-transitory computer-readable medium of claim 17, wherein the second machine learning model includes a first neural network and a second neural network, and generating the second machine learning model comprises:
generating, based on listing features of a first selected listing, a selected listing vector representing the first selected listing, the selected listing vector generated using the first neural network, the first selected listing corresponding to a first historical search query;
generating, based on listing features of a first set of unselected listings, an unselected listing vector representing the first set of unselected listings, the unselected listing vector generated using the second neural network, the first set of unselected listings corresponding to the first historical search query;
generating, based on query features of the first historical search query, a training target vector representing an estimated optimal listing for the first historical search query, the training target vector generated using a first machine learning model;
determining a first training distance measured between the selected listing vector and the training target vector;
determining a second training distance measured between the unselected listing vector and the training target vector; and
adjusting weights of the first neural network and the second neural network based on the first training distance and the second training distance.

### Claim 19 (depends on 18)

The non-transitory computer-readable medium of claim 18, wherein adjusting weights of the first neural network and the second neural network comprises:
determining that the first training distance is longer than the second? training distance; and
adjusting the weights of the first neural network and the second neural network to result in a subsequent iteration of the first training distance being shorter than a subsequent iteration of the second training distance.

### Claim 20 (depends on 15)

The non-transitory computer-readable medium of claim 15, wherein ranking the set of listings for the first search query comprises:
determining that the first distance is shorter than the second distance; and
ranking the first listing higher than the second listing based on determining that the first distance is shorter than the second distance.
