Method for relevancy ranking of products in online shopping

1. A method comprising:

receiving a ranked list comprising a plurality of products, each product associated with one or more attributes;

for each product in a subset of the plurality of products:

determining a similarity between an attribute of the product on the ranked list and an attribute of a second product on the ranked list; and

diversifying the ranked list by modifying a ranking of the product based on the determined similarity;

determining a discounted cumulative gain (DCG) score for the diversified ranked list for each product in the subset of the plurality of products; and

modifying a ranking of a product in the diversified ranked list, wherein modifying the ranking includes:

determining a position adjustment for a product of the plurality of products in the diversified ranked list based on a cost of the product paid by an owner of the product in response to the cost of the product paid by the owner of the product being less than a cost paid by an owner of another product having a higher position in the diversified ranked list;

determining a loss of DCG score for the diversified ranked list for the product due to performing the position adjustment for the product, the loss of the DCG score for the diversified ranked list for the product comprising a difference between the DCG score for the diversified ranked list for the product and an alternative DGC for the diversified ranked list with the position adjustment for the product; and

responsive to determining that the loss of DCG score is less than a predetermined threshold value, modifying the ranking of the product in the diversified ranked list.

2. The method of claim 1 , further comprising:

receiving a search query, wherein the DCG score for the diversified ranked list reflects a relevance of the diversified ranked list for the received search query in comparison to a relevance of a different diversified rank list for the received search query.

3. The method of claim 1 , wherein determining the DCG score for the diversified ranked list comprises summating a plurality of marginal relevance scores, each marginal relevance score assigned to each product in the diversified ranked list.

4. The method of claim 3 , wherein determining the loss of DCG score incurred by the diversified ranked list due to performing the position adjustment for the product comprises:

determining a modified marginal relevance score for the product based on the position adjustment for the product;

adjusting the DCG score based on the determined modified marginal relevance score for the product; and

determining a difference between the DCG score and the adjusted DCG score.

5. The method of claim 3 , wherein each marginal relevance score is determined based on a measure of similarity between a product assigned the marginal relevance score and one or more products ranked higher than the product in the diversified ranked list.

6. The method of claim 3 , wherein modifying the ranking of the product further comprises:

prior to determining the position adjustment, determining that the product is associated with the cost of the product, wherein the cost of the product is paid by the owner of the product.

7. The method of claim 1 , wherein determining the position adjustment for the product comprises identifying a different product in the diversified ranked list ranked immediately above the product in the diversified list.

8. The method of claim 1 , wherein the cost of the product comprises a cost per click (CPC) of the product.

9. A non-transitory computer readable medium comprising computer code that, when executed by a processor, causes the processor to:

receive a ranked list comprising a plurality of products, each product associated with one or more attributes;

for each product in a subset of the plurality of products:

determine a similarity between an attribute of the product on the ranked list and an attribute of a second product on the ranked list; and

diversify the ranked list by modifying a ranking of the product based on the determined similarity;

determine a discounted cumulative gain (DCG) score for the diversified ranked list for each product in the subset of the plurality of products; and

modify a ranking of a product in the diversified ranked list, wherein modifying the ranking includes:

determine a position adjustment for a product of the plurality of products in the diversified ranked list based on a cost of the product paid by an owner of the product in response to the cost of the product paid by the owner of the product being less than a cost paid by an owner of another product having a higher position in the diversified ranked list;

determine a loss of DCG score for the diversified ranked list for the product due to performing the position adjustment for the product, the loss of the DCG score for the diversified ranked list for the product comprising a difference between the DCG score for the diversified ranked list for the product and an alternative DGC for the diversified ranked list with the position adjustment for the product; and

responsive to determine that the loss of DCG score is less than a predetermined threshold value, modify the ranking of the product in the diversified ranked list.

10. The non-transitory computer readable medium of claim 9 , further comprising computer code that, when executed by a processor, causes the processor to:

receive a search query, wherein the DCG score for the diversified ranked list reflects a relevance of the diversified ranked list for the received search query in comparison to a relevance of a different diversified rank list for the received search query.

11. The non-transitory computer readable medium of claim 9 , wherein the computer code that causes the processor to determine the DCG score for the diversified ranked list further comprises computer code that, when executed by the processor, causes the processor to summate a plurality of marginal relevance scores, each marginal relevance score assigned to each product in the diversified ranked list.

12. The non-transitory computer readable medium of claim 11 , wherein the computer code that causes the processor to determine the loss of DCG score incurred by the diversified ranked list due to performing the position adjustment for the product further comprises computer code that, when executed by the processor, causes the processor to:

determine a modified marginal relevance score for the product based on the position adjustment for the product;

adjust the DCG score based on the determined modified marginal relevance score for the product; and

determine a difference between the DCG score and the adjusted DCG score.

13. The non-transitory computer readable medium of claim 11 , wherein each marginal relevance score is determined based on a measure of similarity between a product assigned the marginal relevance score and one or more products ranked higher than the product in the diversified ranked list.

14. The non-transitory computer readable medium of claim 11 , wherein the computer code that causes the processor to modify the ranking of the product further comprises computer code that, when executed by the processor, causes the processor to:

prior to determine the position adjustment, determine that the product is associated with the cost of the product, wherein the cost of the product is paid by the owner of the product.

15. The non-transitory computer readable medium of claim 9 , wherein the computer code that causes the processor to determine the position adjustment for the product further comprises computer code that, when executed by the processor, causes the processor to:

identify a different product in the diversified ranked list ranked immediately above the product in the diversified list.

16. The non-transitory computer readable medium of claim 9 , wherein the cost of the product comprises a cost per click (CPC) of the product.