Enforcing diversity in ranked relevance results returned from a universal relevance service framework

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 API including:

a relevance application layer, configured to provide the relevance service API, wherein the relevance application layer includes at least one aggregator node that is operable to execute a portion of the relevance service processing algorithms; and

a distributed search cluster, configured to be implemented on one or a combination of distributed search servers, wherein the distributed search cluster is configured to implement relevance functions including deal ranking and scoring; and

wherein the relevance application layer performs operations comprising:

receiving, by the relevance service API, a relevance search request from a relevance API client, wherein the relevance search request is a search query that was generated on behalf of a particular consumer for deals describing promotion offerings that are currently available to the consumer;

receiving, from the distributed search cluster, an ordered list of deals, the ordering based on each deal's user-item relevance score representing a likelihood that the deal will be purchased by consumers; and

adjusting the ordering based in part on enforcing diversity in the ordered list of deals;

wherein there are N deals in the ordered list of deals, and wherein adjusting the ordering comprises:

receiving a group of N filters, each filter representing a constraint and being respectively associated with a desired mix percentage and a weight;

generating a set of N filtered lists by applying each filter to the sorted list;

instantiating an empty output sorted list; and

populating the output sorted list using deals from the set of filtered lists, wherein the output sorted list mix percentage is calculated based on attributes of the deals and the output sorted list ordering is calculated based on the respective desired mix percentages of each of the filters.

2. The system of claim 1 , wherein populating the output sorted list is preceded by:

in an instance in which at least one filter is associated with at least one seed deal,

adding the seed deal to the output sorted list.

3. The system of claim 1 , wherein adjusting the ordering includes a co-ranking personalization adjustment based on data describing signals representing activity of the particular consumer.

4. 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 provide a relevance service API including:

a relevance application layer, configured to provide the relevance service API, wherein the relevance application layer includes at least one aggregator node that is operable to execute a portion of the relevance service processing algorithms; and

a distributed search cluster, configured to be implemented on one or a combination of distributed search servers, wherein the distributed search cluster is configured to implement relevance functions including deal ranking and scoring; and

wherein the relevance application layer performs operations comprising:

receiving, by the relevance service API, a relevance search request from a relevance API client, wherein the relevance search request is a search query that was generated on behalf of a particular consumer for deals describing promotion offerings that are currently available to the consumer;

receiving, from the distributed search cluster, an ordered list of deals, the ordering based on each deal's user-item relevance score representing a likelihood that the deal will be purchased by consumers; and

adjusting the ordering based in part on enforcing diversity in the ordered list of deals;

wherein there are N deals in the ordered list of deals, and wherein adjusting the ordering comprises:

receiving a group of N filters, each filter representing a constraint and being respectively associated with a desired mix percentage and a weight;

generating a set of N filtered lists by applying each filter to the sorted list;

instantiating an empty output sorted list; and

populating the output sorted list using deals from the set of filtered lists, wherein the output sorted list mix percentage is calculated based on attributes of the deals and the output sorted list ordering is calculated based on the respective desired mix percentages of each of the filters.

5. The computer program product of claim 4 , wherein populating the output sorted list is preceded by:

in an instance in which at least one filter is associated with at least one seed deal,

adding the seed deal to the output sorted list.