System and method for automatically retrieving relevant digital advertisements from multiple channels

1. A system, comprising:

a non-transitory memory having instructions stored thereon; and

at least one processor operatively coupled to the non-transitory memory, and configured to read the instructions to:

obtain, from a computing device, a search request identifying a query and seeking items to be displayed on a webpage of a website to a user,

search, based on the query, a first database to retrieve a first set of sponsored items associated with the website, wherein the first set of sponsored items are retrieved and ranked based on an optimization of conversion rate,

search, based on the query, a second database to retrieve a second set of sponsored items associated with the website, wherein the second set of sponsored items are retrieved and ranked based on an optimization of click-through rate,

compute, for each sponsored item in the first set and the second set, a relevance score representing a degree of relevancy between the sponsored item and the query,

filter the first set of sponsored items based on their relevance scores and at least one predetermined threshold, to generate a first filtered set of sponsored items,

generate, for each of the first filtered set of sponsored items, a first ranking score representing a likelihood of conversion using a first machine learning model,

select, from the first filtered set, up to a predetermined first number of sponsored items based on their respective first ranking scores, to generate a first selected set of sponsored items,

filter the second set of sponsored items based on their relevance scores and at least one predetermined threshold, to generate a second filtered set of sponsored items,

generate, for each of the second filtered set of sponsored items, a second ranking score representing a likelihood of click using a second machine learning model,

select, from the second filtered set, up to a predetermined second number of sponsored items based on their respective second ranking scores, to generate a second selected set of sponsored items,

generate, based on the first selected set of sponsored items and the second selected set of sponsored items, a ranked list of recommended items based on an advertisement auction mechanism, and

transmit, to the computing device, the ranked list of recommended items in response to the search request.

2. The system of claim 1 , wherein the second database is searched based on a determination that the query is not associated with an explicit intent or an implicit intent of the user.

3. The system of claim 1 , wherein the at least one processor is further configured to read the instructions to:

search, based on the query, the second database to retrieve a third set of sponsored items associated with the website, wherein the third set of sponsored items are retrieved and ranked based on an optimization of click-through rate.

4. The system of claim 3 , wherein:

each of the second set of sponsored items is retrieved based on a literal match to the query; and

each of the third set of sponsored items is retrieved based on a semantic match to the query.

5. The system of claim 3 , wherein the at least one processor is further configured to read the instructions to:

filter the third set of sponsored items based on their relevance scores and at least one predetermined threshold, to generate a third filtered set of sponsored items;

generate, for each of the third filtered set of sponsored items, a third ranking score representing a likelihood of click using a third machine learning model; and

select, from the third filtered set, up to a predetermined third number of sponsored items based on their respective second ranking scores, to generate a third selected set of sponsored items.

6. The system of claim 1 , wherein:

the click-through rate is computed by the computing device based on a ratio between a number of clicks of users on an item and a number of times the item is displayed to the users;

the first machine learning model is assigned to be executed on a first processing device to generate the first ranking score; and

the second machine learning model is assigned to be executed on a second processing device to generate the second ranking score.

7. The system of claim 5 , wherein the at least one processor is further configured to read the instructions to:

merge the three selected sets of sponsored items to generate a unified list of sponsored items; and

compute, for each of the unified list of sponsored items, a unified ranking score based on a weighted average of the first ranking score and a maximum of the second ranking score and the third ranking score.

8. The system of claim 7 , wherein the at least one processor is further configured to read the instructions to:

compare the unified ranking score of each sponsored item in the unified list with a dynamic threshold, wherein the dynamic threshold is determined based on a machine learning model and a traffic segment of the query; and

select sponsored items beyond the dynamic threshold from the unified list to generate a relevant list of sponsored items.

9. The system of claim 8 , wherein the at least one processor is further configured to read the instructions to:

obtain, based on the advertisement auction mechanism, cost per click (CPC) data associated with each of the relevant list of sponsored items; and

modify, based on the CPC data and a predetermined price squashing parameter, the unified ranking score to generate a re-ranking score for each sponsored item in the relevant list.

10. The system of claim 9 , wherein the at least one processor is further configured to read the instructions to:

re-rank the relevant list of sponsored items based on their respective re-ranking scores; and

filter, based on item availability in local inventory for the user, the re-ranked sponsored items to generate the ranked list of recommended items, wherein each of the ranked list of recommended items has a corresponding re-ranking score and is recommended to be displayed at a corresponding position in the webpage based on its corresponding re-ranking score.

11. A computer-implemented method, comprising:

obtaining, from a computing device, a search request identifying a query and seeking items to be displayed on a webpage of a website to a user;

searching, based on the query, a first database to retrieve a first set of sponsored items associated with the website, wherein the first set of sponsored items are retrieved and ranked based on an optimization of conversion rate;

searching, based on the query, a second database to retrieve a second set of sponsored items associated with the website, wherein the second set of sponsored items are retrieved and ranked based on an optimization of click-through rate;

computing, for each sponsored item in the first set and the second set, a relevance score representing a degree of relevancy between the sponsored item and the query;

filtering the first set of sponsored items based on their relevance scores and at least one predetermined threshold, to generate a first filtered set of sponsored items;

generating, for each of the first filtered set of sponsored items, a first ranking score representing a likelihood of conversion using a first machine learning model;

selecting, from the first filtered set, up to a predetermined first number of sponsored items based on their respective first ranking scores, to generate a first selected set of sponsored items;

filtering the second set of sponsored items based on their relevance scores and at least one predetermined threshold, to generate a second filtered set of sponsored items;

generating, for each of the second filtered set of sponsored items, a second ranking score representing a likelihood of click using a second machine learning model;

selecting, from the second filtered set, up to a predetermined second number of sponsored items based on their respective second ranking scores, to generate a second selected set of sponsored items;

generating, based on the first selected set of sponsored items and the second selected set of sponsored items, a ranked list of recommended items based on an advertisement auction mechanism; and

transmitting, to the computing device, the ranked list of recommended items in response to the search request.

12. The computer-implemented method of claim 11 , further comprising:

searching, based on the query, the second database to retrieve a third set of sponsored items associated with the website, wherein the third set of sponsored items are retrieved and ranked based on an optimization of click-through rate, wherein

each of the second set of sponsored items is retrieved based on a literal match to the query; and

each of the third set of sponsored items is retrieved based on a semantic match to the query.

13. The computer-implemented method of claim 12 , further comprising:

filtering the third set of sponsored items based on their relevance scores and at least one predetermined threshold, to generate a third filtered set of sponsored items;

generating, for each of the third filtered set of sponsored items, a third ranking score representing a likelihood of click using a third machine learning model; and

selecting, from the third filtered set, up to a predetermined third number of sponsored items based on their respective second ranking scores, to generate a third selected set of sponsored items.

14. The computer-implemented method of claim 11 , wherein:

the click-through rate is computed by the computing device based on a ratio between a number of clicks of users on an item and a number of times the item is displayed to the users;

the first machine learning model is assigned to be executed on a first processing device to generate the first ranking score; and

the second machine learning model is assigned to be executed on a second processing device to generate the second ranking score.

15. The computer-implemented method of claim 11 , further comprising:

merging the three selected sets of sponsored items to generate a unified list of sponsored items; and

computing, for each of the unified list of sponsored items, a unified ranking score based on a weighted average of the first ranking score and a maximum of the second ranking score and the third ranking score.

16. The computer-implemented method of claim 15 , further comprising:

comparing the unified ranking score of each sponsored item in the unified list with a dynamic threshold, wherein the dynamic threshold is determined based on a machine learning model and a traffic segment of the query; and

selecting sponsored items beyond the dynamic threshold from the unified list to generate a relevant list of sponsored items.

17. The computer-implemented method of claim 16 , further comprising:

obtaining, based on the advertisement auction mechanism, cost per click (CPC) data associated with each of the relevant list of sponsored items; and

modifying, based on the CPC data and a predetermined price squashing parameter, the unified ranking score to generate a re-ranking score for each sponsored item in the relevant list.

18. The computer-implemented method of claim 17 , further comprising:

re-ranking the relevant list of sponsored items based on their respective re-ranking scores; and

filtering, based on item availability in local inventory for the user, the re-ranked sponsored items to generate the ranked list of recommended items, wherein each of the ranked list of recommended items has a corresponding re-ranking score and is recommended to be displayed at a corresponding position in the webpage based on its corresponding re-ranking score.

19. A non-transitory computer readable medium having instructions stored thereon, wherein the instructions, when executed by at least one processor, cause at least one device to perform operations comprising:

obtaining, from a computing device, a search request identifying a query and seeking items to be displayed on a webpage of a website to a user;

searching, based on the query, a first database to retrieve a first set of sponsored items associated with the website, wherein the first set of sponsored items are retrieved and ranked based on an optimization of conversion rate;

searching, based on the query, a second database to retrieve a second set of sponsored items associated with the website, wherein the second set of sponsored items are retrieved and ranked based on an optimization of click-through rate;

computing, for each sponsored item in the first set and the second set, a relevance score representing a degree of relevancy between the sponsored item and the query;

filtering the first set of sponsored items based on their relevance scores and at least one predetermined threshold, to generate a first filtered set of sponsored items;

generating, for each of the first filtered set of sponsored items, a first ranking score representing a likelihood of conversion using a first machine learning model;

selecting, from the first filtered set, up to a predetermined first number of sponsored items based on their respective first ranking scores, to generate a first selected set of sponsored items;

filtering the second set of sponsored items based on their relevance scores and at least one predetermined threshold, to generate a second filtered set of sponsored items;

generating, for each of the second filtered set of sponsored items, a second ranking score representing a likelihood of click using a second machine learning model;

selecting, from the second filtered set, up to a predetermined second number of sponsored items based on their respective second ranking scores, to generate a second selected set of sponsored items;

generating, based on the first selected set of sponsored items and the second selected set of sponsored items, a ranked list of recommended items based on an advertisement auction mechanism; and

transmitting, to the computing device, the ranked list of recommended items in response to the search request.

20. The non-transitory computer readable medium of claim 19 , wherein the second database is searched based on a determination that the query is not associated with an explicit intent or an implicit intent of the user.