User profile filtering based upon sensitive topics

1 . A method executing on a processor of a computing device that causes the computing device to perform operations comprising:

identifying a set of candidate topics labeled with sensitivity labels corresponding to whether candidate topics are sensitive topics or non-sensitive topics;

applying the sensitivity labels to an unknown entity space of entities to label the entities with the sensitivity labels to create a set of labeled topics labeled as either the sensitive topics or the non-sensitive topics, wherein the applying comprises:

crawling a plurality of remote content sources to identify a first unknown entity of the entities and a second unknown entity of the entities, wherein the crawling is performed using a crawler configured to retrieve one or more pages and provide the one or more pages to a text cleaner configured to at least one of remove redundant text or add one or more tags;

labeling the first unknown entity with a sensitivity label based upon a determination that the first unknown entity has one or more similar characteristics as a sensitive topic; and

labeling the second unknown entity with a non-sensitivity label based upon a determination that the second unknown entity does not have one or more similar characteristics as a sensitive topic, wherein at least one of the determination that the first unknown entity has one or more similar characteristics as a sensitive topic or the determination that the second unknown entity does not have one or more similar characteristics as a sensitive topic is performed by one or more machine learning models configured to (i) project one or more entities into a semantic embedding space and (ii) compare one or more embedding distances between the one or more entities and one or more candidate topics;

storing the set of labeled topics and metadata for the set of labeled topics within a central sensitivity data store configured to store one or more sensitivity mappings and associated metadata;

detecting a modification to a regulatory compliance policy;

updating the central sensitivity data store based upon the modification to the regulatory compliance policy by reclassifying one or more entities within the central sensitivity data store with one or more updated sensitivity labels based upon the modification to the regulatory compliance policy;

filtering, utilizing a user profile updater configured to retrieve the set of labeled topics determined by crawling the plurality of remote content sources and the metadata within the central sensitivity data store, user profiles of users of a first geographical region to exclude content corresponding to sensitive topics without excluding at least some other content from the user profiles in compliance with a first regulatory compliance policy associated with what information about users is allowed to be tracked in the first geographical region; and

filtering, utilizing a user profile updater configured to retrieve a second set of labeled topics determined by crawling the plurality of remote content sources and second metadata, user profiles of users of a second geographical region to exclude content corresponding to second sensitive topics without excluding at least some other content from the user profiles in compliance with a second regulatory compliance policy associated with what information about users is allowed to be tracked in the second geographical region.

2 . The method of claim 1 , wherein:

the plurality of remote content sources are from an online encyclopedia website.

3 . The method of claim 1 , wherein comprising:

the regulatory compliance policy subjected to the modification is the first regulatory compliance policy associated with what information about users is allowed to be tracked in the first geographical region.

4 . The method of claim 1 , comprising:

periodically generating new unknown entity spaces of new entities to classify with sensitivity labels for updating the central sensitivity data store with a new sensitive topic; and

in response to updating the central sensitivity data store, performing a subsequent filtering operation to filter the new sensitive topic from the user profiles.

5 . The method of claim 1 , comprising:

crawling one or more content sources to identify at least one of the entities.

6 . The method of claim 1 , comprising:

identifying a set of content items available to provide to a user;

filtering the set of content items to exclude content items corresponding to the sensitive topics to create a filtered set of content items; and

providing a content item selected from the filtered set of content items to a computing device for display to the user.

7 . The method of claim 1 , comprising:

crawling one or more content sources to identify one or more entities;

updating the central sensitivity data store based upon the one or more entities; and

filtering one or more new sensitive topics from one or more user profiles utilizing the central sensitivity data store.

8 . The method of claim 1 , comprising:

providing a user with access to the central sensitivity data store through an interface; and

populating the interface with a list of sensitive topics for review and editing.

9 . The method of claim 8 , comprising:

in response to receiving, through the interface, user input indicating that a sensitive topic within the list of sensitive topics is a non-sensitive topic, reclassifying the sensitive topic as the non-sensitive topic.

10 . A non-transitory machine-readable medium having stored thereon processor-executable instructions that when executed cause performance of operations, the operations comprising:

identifying a set of candidate topics labeled with sensitivity labels corresponding to whether candidate topics are sensitive topics or non-sensitive topics;

applying the sensitivity labels to an unknown entity space of entities to label the entities with the sensitivity labels to create a set of labeled topics labeled as either the sensitive topics or the non-sensitive topics, wherein the applying comprises:

crawling a plurality of remote content sources to identify a first unknown entity of the entities and a second unknown entity of the entities, wherein the crawling is performed using a crawler configured to retrieve one or more pages and provide the one or more pages to a text cleaner configured to at least one of remove redundant text or add one or more tags;

labeling the first unknown entity with a sensitivity label based upon a determination that the first unknown entity has one or more similar characteristics as a sensitive topic; and

labeling the second unknown entity with a non-sensitivity label based upon a determination that the second unknown entity does not have one or more similar characteristics as a sensitive topic, wherein at least one of the determination that the first unknown entity has one or more similar characteristics as a sensitive topic or the determination that the second unknown entity does not have one or more similar characteristics as a sensitive topic is performed by one or more machine learning models configured to (i) project one or more entities into a semantic embedding space and (ii) compare one or more embedding distances between the one or more entities and one or more candidate topics;

storing the set of labeled topics and metadata for the set of labeled topics within a central sensitivity data store configured to store one or more sensitivity mappings and associated metadata;

detecting a modification to a regulatory compliance policy;

updating the central sensitivity data store based upon the modification to the regulatory compliance policy by reclassifying one or more entities within the central sensitivity data store with one or more updated sensitivity labels based upon the modification to the regulatory compliance policy;

filtering, utilizing a user profile updater configured to retrieve the set of labeled topics determined by crawling the plurality of remote content sources and the metadata within the central sensitivity data store, user profiles of users of a first geographical region to exclude content corresponding to sensitive topics without excluding at least some other content in compliance with a first regulatory compliance policy associated with what information about users is allowed to be tracked in the first geographical region;

filtering, utilizing a user profile updater configured to retrieve a second set of labeled topics determined by crawling the plurality of remote content sources and second metadata, user profiles of users of a second geographical region to exclude content corresponding to second sensitive topics without excluding at least some other content in compliance with a second regulatory compliance policy associated with what information about users is allowed to be tracked in the second geographical region; and

utilizing a user profile to select content corresponding to a non-sensitive topic to display on a display of a computing device.

11 . The non-transitory machine-readable medium of claim 10 , comprising:

providing a user with access to the central sensitivity data store through an interface; and

populating the interface with at least one of a list of non-sensitive topics or a list of sensitive topics.

12 . The non-transitory machine-readable medium of claim 11 , comprising:

in response to receiving, through the interface, user input indicating that a non-sensitive topic within the list of non-sensitive topics is a sensitive topic, reclassifying the non-sensitive topic as the sensitive topic.

13 . The non-transitory machine-readable medium of claim 10 , comprising:

in response to receiving user input indicating that a non-sensitive topic within the central sensitivity data store is a sensitive topic, reclassifying the non-sensitive topic as the sensitive topic; and

modifying, based upon the reclassification, a weight or a classification process used by machine learning functionality to label entities with sensitivity labels.

14 . The non-transitory machine-readable medium of claim 10 , comprising:

in response to receiving user input indicating that a sensitive topic within the central sensitivity data store is a non-sensitive topic, reclassifying the sensitive topic as the non-sensitive topic; and

modifying, based upon the reclassification, a weight or a classification process used by machine learning functionality to label entities with sensitivity labels.

15 . The non-transitory machine-readable medium of claim 10 , comprising:

identifying a compliance domain for a location of a user having a user profile within the central sensitivity data store; and

modifying a weight or a classification process used by machine learning functionality to label entities with sensitivity labels for filtering the user profile of the user based upon the compliance domain.

16 . A computing device comprising:

a processor; and

memory comprising processor-executable instructions that when executed by the processor cause performance of operations, the operations comprising:

identifying a set of candidate topics labeled with sensitivity labels corresponding to whether candidate topics are sensitive topics or non-sensitive topics;

applying sensitivity labels to an unknown entity space of entities to label the entities with the sensitivity labels to create a set of labeled topics labeled as either sensitive topics or non-sensitive topics, wherein the applying comprises:

crawling a plurality of remote content sources to identify a first unknown entity of the entities and a second unknown entity of the entities, wherein the crawling is performed using a crawler configured to retrieve one or more pages and provide the one or more pages to a text cleaner configured to at least one of remove redundant text or add one or more tags;

labeling the first unknown entity with a sensitivity label based upon a determination that the first unknown entity has one or more similar characteristics as a sensitive topic; and

labeling the second unknown entity with a non-sensitivity label based upon a determination that the second unknown entity does not have one or more similar characteristics as a sensitive topic, wherein at least one of the determination that the first unknown entity has one or more similar characteristics as a sensitive topic or the determination that the second unknown entity does not have one or more similar characteristics as a sensitive topic is performed by one or more machine learning models configured to (i) project one or more entities into a semantic embedding space and (ii) compare one or more embedding distances between the one or more entities and one or more candidate topics;

storing the set of labeled topics and metadata for the set of labeled topics within a central sensitivity data store configured to store one or more sensitivity mappings and associated metadata;

detecting a modification to a regulatory compliance policy;

updating the central sensitivity data store based upon the modification to the regulatory compliance policy by reclassifying one or more entities within the central sensitivity data store with one or more updated sensitivity labels based upon the modification to the regulatory compliance policy;

filtering, utilizing a user profile updater configured to retrieve the set of labeled topics and the metadata within the central sensitivity data store, user profiles of users of a first geographical region to exclude sensitive topics from the user profiles in compliance with a first regulatory compliance policy associated with what information about users is allowed to be tracked in the first geographical region;

filtering, utilizing a user profile updater configured to retrieve a second set of labeled topics and second metadata, user profiles of users to exclude sensitive topics from user profiles of a second geographical region in compliance with a second regulatory compliance policy associated with what information about users is allowed to be tracked in the second geographical region; and

utilizing a user profile to select content corresponding to a non-sensitive topic to display on a display of a computing device.

17 . The computing device of claim 16 , wherein the operations comprise:

in response to an entity being within a threshold distance of a non-sensitive topic within the semantic embedding space, labeling the entity as the non-sensitive topic.

18 . The computing device of claim 16 , wherein the operations comprise:

in response to an entity being within a threshold distance of a sensitive topic within the semantic embedding space, labeling the entity as the sensitive topic.

19 . The computing device of claim 16 , wherein

modifying a weight or a classification process used by machine learning functionality to label entities with sensitivity labels for filtering a user profile of a user based upon a jurisdiction of where the user is located.

20 . The computing device of claim 16 , wherein

modifying a weight or a classification process used by machine learning functionality to label entities with sensitivity labels for filtering a user profile of a user based upon a compliance law of where the user is located.