Universal relevance service framework

Patent metadata

Field	Value
Publication number	US-10572925-B1
Application number	US-201514824912-A
Country	US
Kind code	B1
Filing date	Aug 12, 2015
Priority date	Aug 15, 2014
Publication date	Feb 25, 2020
Grant date	Feb 25, 2020

Abstract

In general, embodiments of the present invention provide systems, methods and computer readable media for a universal relevance service framework for ranking and personalizing items.

First claim

What is claimed is:

1. A universal relevance service framework system comprising:
one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to provide a relevance service comprising:
a relevance application layer, configured to provide a relevance service application programming interface (API), wherein the relevance application layer includes at least one aggregator node that is operable to execute a portion of one or more relevance service processing algorithms associated with the relevance service, and wherein the at least one aggregator node is configured to include at least one plugin; and
a distributed search cluster, configured to be implemented on a plurality of distributed search servers, wherein the distributed search cluster includes a plurality of distributed search nodes, and wherein each distributed search node of the plurality of distributed search nodes is configured to include at least one distributed search server plugin that implements relevance functions including item ranking and scoring; and
wherein the relevance service performs operations comprising:
receiving, by the relevance service API, a relevance search request from a relevance API client associated with a consumer device associated with a particular consumer, wherein the relevance search request is a search query that was generated on behalf of the particular consumer for one or more products that are currently available to the particular consumer;
generating, by the relevance application layer, a distributed search query based on the search request, wherein the distributed search query includes at least one document describing attributes of a product-specific search;
executing, by the distributed search cluster, the distributed search query, wherein executing the distributed search query comprises:
retrieving, using at least one search index, data describing a set of items that satisfy distributed search query terms;
calculating a user-item relevance score associated with each item of the set of items based on one or more user-item attribute vectors, the one or more user-item attribute vectors generated based on a combination of a user-item attribute graph and historical data, wherein the user-item attribute graph describing user-item relationships between one or more personal attributes of the particular consumer and one or more product attributes for a product of the one or more products, and wherein the historical data describing interactions of a personal attribute space comprising the one or more personal attributes and a product space comprising each product attribute of the one or more product attributes for a product of the one or more products;
ranking the set of items based on their respective user-item relevance scores;
providing to the relevance service API a top-ranked ordered list of the set of items; and
transmitting, by the relevance service API and to the consumer device, a response to the relevance search request, wherein the response includes the top-ranked ordered list of the set of items.

2. The system of claim 1 , wherein the distributed search query is a multi-query that includes a payload including at least two different product-specific searches, each product-specific search including a document respectively describing one of the products that are currently available to the consumer, and wherein generating the distributed search query comprises:
for each of the products, selecting a product-specific model that is associated the product;
retrieving features described by the selected product-specific models;
implementing feature computation of document-independent feature data;
generating an ordered list of the feature data; and
including an ordered list of the feature data in the distributed search query payload.

3. The system of claim 2 , wherein selecting the product-specific model is implemented by a custom selection plugin associated with the product.

4. The system of claim 2 , wherein generating the ordered list of the feature data includes deduplicating features across the selected product-specific models.

5. The system of claim 2 , wherein generating the different product-specific searches is implemented in parallel.

6. The system of claim 2 , wherein calculating each user-item relevance score and ranking the set of items is based on feature data retrieved from a group of data sources including document-independent data from the distributed search payload and document-dependent features retrieved from one or more of the search indexes.

7. The system of claim 6 , wherein calculating each user-item relevance score and ranking the set of items is implemented by a custom scoring plugin.

8. The system of claim 1 , wherein returning the ordered list of the top-ranked items from the set of items includes adjusting, by the aggregator node, the ordered list ranking by per-product and/or cross-product re-ranking and co-ranking.

9. The system of claim 1 , wherein the products are promotion offerings.

10. The system of claim 9 , wherein the search query that was generated on behalf of a particular consumer includes search terms describing at least one promotion attribute and at least one attribute describing the particular consumer.

11. A computer program product, stored on a computer readable medium, comprising instructions that when executed on one or more computers cause the one or more computers to perform operations implementing a relevance service, the operations comprising:
receiving, by a relevance service application programming interface (API), a relevance search request from a relevance API client associated with a consumer device associated with a particular consumer, wherein the relevance search request is a search query that was generated on behalf of the particular consumer for one or more products that are currently available to the particular consumer;
generating, by a relevance application layer, a distributed search query based on the search request, wherein the distributed search query includes at least one document describing attributes of a product-specific search;
executing, by a distributed search cluster, the distributed search query, wherein executing the distributed search query comprises:
retrieving, using at least one search index, data describing a set of items that satisfy distributed search query terms;
calculating a user-item relevance score associated with each item of the set of items based on one or more user-item attribute vectors, the one or more user-item attribute vectors generated based on a combination of a user-item attribute graph and historical data, wherein the user-item attribute graph describing user-item relationships between one or more personal attributes of the particular consumer and one or more product attributes for a product of the one or more products, and wherein the historical data describing interaction of a personal attribute space comprising the one or more personal attributes and a product space comprising each product attribute of one or more product attributes for a product of the one or more products;
ranking the set of items based on their respective user-item relevance scores;
providing to the relevance service API a top-ranked ordered list of the set of items; and
transmitting, by the relevance service API and to the consumer device, a response to the relevance search request, wherein the response includes the top-ranked ordered list of the set of items.

12. The computer program product of claim 11 , wherein the distributed search query is a multi-query that includes a payl

Assignees

• Groupon Inc

Inventors

• Roy Chowdhury Amber
• Casey Steven
• Subramaniam Sri

Classifications

• G06Q30/0625 Primary

  by formulating product or service queries, e.g. using keywords or predefined options · CPC title

• G06F16/9535

  Search customisation based on user profiles and personalisation · CPC title

• G06F16/24578

  using ranking · CPC title