Granted Patent B1

US 10,733,638 · App. 16/058,959 · Granted Aug 4, 2020

Inventors: Dinkar Jain (Menlo Park, CA); Tianshi Gao (Fremont, CA); Darshan Kantak (Palo Alto, CA)

Assignee: Facebook, Inc.

G06Q30/0277G06F16/24578G06F16/287G06F16/955G06F16/958G06F16/9535G06F16/9574G06N20/00G06Q50/01

Abstract

Claims (75)

1. A computer-implemented method comprising:

receiving, by an online system, a plurality of tracking requests from one or more client devices, each tracking request generated responsive to a client device of the one or more client devices presenting a user interface received by the client device from a third party website;

extracting one or more parameters comprising at least a parameter associated with a tracking request from the plurality of tracking requests;

determining, based on each of the one or more parameters, a category from a taxonomy of categories, the category describing an item presented on the user interface;

adding, based on the determined category, a dynamic component of a user profile, wherein the user profile dynamic component includes one or more attributes describing the determined category;

predicting, based on one or more machine learning based models configured to receive the one or more parameters:

values for the one or more attributes describing the determined category, and

a temporal relevance score for the dynamic component, the temporal relevance score indicating a period of time in which a user is likely to perform an action associated with an item of the determined category; and

directing a content item to a client device associated with the user profile, wherein the directing is based on the predicted values of the one or more attributes of the user profile dynamic component and the temporal relevance score.

2. The computer-implemented method of claim 1 , wherein the one or more parameters comprise another parameter extracted from data obtained through web crawling microdata contained within the user interface.

3. The computer-implemented method of claim 2 , further comprising:

web crawling the microdata contained within the user interface by:

extracting, from a web page from the third party website, a plurality of keywords describing the item presented on the user interface; and

cross-referencing the plurality of keywords against a plurality of stored keywords to determine the category.

4. The computer-implemented method of claim 1 , wherein the one or more parameters are extracted from a uniform resource locator (URL) of the user interface, the URL including tokens describing the one or more attributes associated with the category.

5. The computer-implemented method of claim 1 , wherein the one or more attributes associated with each category include one or more of: an item color, an item size, and/or an item ID.

6. The computer-implemented method of claim 1 , wherein the taxonomy of categories is associated with a list of items, wherein each item is mapped to one or more categories.

7. The computer-implemented method of claim 1 , wherein the tracking request is received from the client device of the one or more client devices responsive to the client device rendering the web page for display, wherein the rendering of the web page causes execution of instructions for sending a tracking request to the online system.

8. The computer-implemented method of claim 1 , further comprising:

adding the temporal relevance score to the dynamic component of the user profile, wherein the temporal relevance score is associated with the determined category;

determining, based on the period of time and a time at which the dynamic component was added, an expiry date for the dynamic component of the user profile; and

deleting the dynamic component after the expiry date.

9. The computer-implemented method of claim 8 , further comprising:

monitoring user interactions associated with the determined category; and

modifying the expiry date of the dynamic component according to the monitored user interactions associated with the determined category.

10. The computer-implemented method of claim 8 , further comprising:

storing information indicating whether a category of the dynamic component has a periodicity; and

determining whether to delete the dynamic component after the expiry date based on whether the category of the dynamic component has a periodicity.

11. The computer-implemented method of claim 10 , further comprising:

responsive to determining that the dynamic component is not likely to be needed after more than a threshold time period, deleting the dynamic component after the expiry date in spite of the category of the dynamic component being associated with a periodicity.

12. The computer-implemented method of claim 10 , further comprising:

determining the periodicity of the category of the dynamic component based on historic actions of other users having dynamic components of the category.

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

responsive to determining that the dynamic component does not have peridocity, deleting the dynamic component after the expiry date.

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

responsive to determining that the dynamic component has periodicity, marking the dynamic component after the expiry date as inactive without deleting the dynamic component.

15. The computer-implemented method of claim 1 , wherein the temporal relevance score comprises a weighted average indicative of a frequency with which a user interacts with the determined category.

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

determining a frequency of interactions with the determined category;

responsive to determining that the frequency of interactions is above a threshold frequency, increasing the temporal relevance score; and

responsive to determining that the frequency of interactions is below the threshold frequency, decreasing the temporal relevance score.

17. The computer-implemented method of claim 1 , wherein determining, based on each of the one or more parameters, the category from the taxonomy of categories comprises:

inputting the one or more parameters into the one or more machine learning models; and

receiving, as output from the one or more machine learning models, the determined category from the taxonomy of categories.

18. The method of claim 1 , further comprising:

building a taxonomy store by:

creating a test account, wherein the test account communicates with the third party website;

generating a Hypertext Transfer Protocol (HTTP) request for performing a search request for the item;

receiving, responsive to the HTTP request, a first web page from the third party website, wherein the first web page comprises a plurality of items fulfilling the search request;

extracting, from the first web page, a first plurality of keywords describing each item of the plurality of items;

for each respective item of the plurality of items:

generating an additional HTTP request for information for the respective item,

receiving, responsive to the additional HTTP request, a second web page from the third party website, and

extracting, from the second web page, a second plurality of keywords describing information for the respective item; and

adding the first plurality of keywords and second plurality of keywords to the taxonomy store, wherein the first plurality of keywords and second plurality of keywords are associated with the category from the taxonomy of categories.

19. A non-transitory computer readable storage medium, storing instructions for the steps comprising:

receiving, by an online system, a plurality of tracking requests from one or more client devices, each tracking request generated responsive to a client device of the one or more client devices presenting a user interface received by the client device from a third party website;

extracting one or more parameters comprising at least a parameter associated with a tracking request from the plurality of tracking requests;

determining, based on each of the one or more parameters, a category from a taxonomy of categories, the category describing an item presented on the user interface;

adding, based on the determined category, a dynamic component of a user profile, wherein the user profile dynamic component includes one or more attributes describing the determined category;

predicting, based on one or more machine learning based models configured to receive the one or more parameters:

values for the one or more attributes describing the determined category; and

a temporal relevance score for the dynamic component, the temporal relevance score indicating a period of time in which a user is likely to perform an action associated with an item of the determined category; and

directing a content item to a client device associated with the user profile, wherein the directing is based on the predicted values of the one or more attributes of the user profile dynamic component and the temporal relevance score.

20. A computer system comprising:

one or more electronic processors; and

a non-transitory computer readable storage medium, storing instructions for the steps comprising:

receiving, by an online system, a plurality of tracking requests from one or more client devices, each tracking request generated responsive to a client device of the one or more client devices presenting a user interface received by the client device from a third party website;

extracting one or more parameters comprising at least a parameter associated with a tracking request from the plurality of tracking requests;

determining, based on each of the one or more parameters, a category from a taxonomy of categories, the category describing an item presented on the user interface;

adding, based on the determined category, a dynamic component of a user profile, wherein the user profile dynamic component includes one or more attributes describing the determined category;

predicting, based on one or more machine learning based models configured to receive the one or more parameters:

values for the one or more attributes describing the determined category; and

a temporal relevance score for the dynamic component, the temporal relevance score indicating a period of time in which a user is likely to perform an action associated with an item of the determined category; and

directing a content item to a client device associated with the user profile, wherein the directing is based on the predicted values of the one or more attributes of the user profile dynamic component and the temporal relevance score.

Assignments (2)

Cited By (8)

US 12,522,243 US 12,536,131 US 12,554,467 US 12,591,240 US 12,618,976 US 12,623,691 US 12,632,166 US 12,709,294