METHOD AND APPARATUS FOR MACHINE-LEARNING BASED PROFILING

• US 20100138370A1
• Filed: 11/23/2009
• Published: 06/03/2010
• Est. Priority Date: 11/21/2008
• Status: Active Grant

• Current Assignee: WSOU Investments, LLC (WSOU Holdings, LLC)
• Original Assignee: Kindsight Incorporated
• Inventors: Lee, Denny Lung Sun, Gassewitz, Michael, Wu, Wang, Gaudet, Robert, Tse, Dorothy
• Granted Patent: US 9,135,348 B2
• Time in Patent Office: Days
• US Class Current: 706/12
• CPC Class Codes: G06F 16/337 Profile generation, learning or modification
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: H04H 60/31 Arrangements for monitoring the use made of the broadcast services
• CPC Class Codes: H04H 60/33 Arrangements for monitoring the users' behaviour or opinions
• CPC Class Codes: H04H 60/46 for recognising users' preferences

Abstract

A method and system for profiling a user based upon a user'"'"'s previous on-line actions is provided. The profile provides a characterization of the user'"'"'s preferences based upon a received user event. The user event identifying event identification information and a user identifier. A look-up in a cached web map is performed to retrieve classification information associated with the event identification information. A user profile is retrieved or created for the user identifier. Profile update information is generated based upon the retrieved classification information for the user event, to identify how the user is to be updated based upon the retrieved classification information and defined profiling rules. The user profile is updated and stored for access by an external advertising server. The classification information provides a text-score record comprising a text string and a score defined in relation to a lexical ontology comprising a hierarchy of categories.

Claims

1. A method of profiling a user of a computing device, connected to a network, based upon a user'"'"'s on-line actions, the method comprising:

receiving a user event generated by actions performed by the user on the computing device, the user event identifying event identification information indicating content requested by the user and a user identifier;

performing a look-up in a cached web map using the event identification information to determine classification information associated with the content requested by the user, the classification information providing a text-score record comprising at least a text string and a score associated with the text string, the text string defined in relation to a lexical ontology comprising a hierarchy of categories, and the score representing a strength of association of the text string to the content;

retrieving classification information associated with the event identification information from the cached web map if classification information is present for the event identification information;

retrieving a user profile associated with the user identification information from a profile database;

generating profile update information by applying profiling rules defining known associations between user action events and user behaviours to the retrieved classification information for the user event, wherein the profile update information identifies how to modify the user profile by adding classification identification information or modifying existing classification identification information scores stored in the profile;

updating the user profile based upon the update information, the user profile providing a characterization of the user'"'"'s preferences; and

storing the updated user profile wherein the user profile can be accessed by an external advertising server.

2. The method of claim 1 wherein the profiling rules are provided by a modelling system for modelling user behaviour from user action events to predict user preferences, the profiling rules generated by:

removing the user identification information from the received user event;

aggregating a plurality of de-identified user events;

sending the aggregated de-identified user events periodically to the modelling system for use as the user action events; and

receiving updated profiling rules in response to the sent aggregated de-identified user events, the updated profiling rules used for generating the profile update information.

3. The method of claim 2 further comprising creating a new user profile when a profile associated with the user identification information is not present in the profile database.

4. The method of claim 3 wherein score is a value within a scalar range with a decimal value between 0 and 1.

5. The method of claim 4 wherein the value of each score in a user profile associated to one or more categories is decreased by a defined amount on a periodic basis in order to provide a time decay function.

6. The method of claim 2 wherein the event identification information is a universal resource locator (URL).

7. The method of claim 6 further comprising:

sending the URL to a modelling system to perform a mapping function of the location associated with the URL and determine classification information for the URL when the URL is not present in the web map cache; and

adding the classification information to the web map.

8. The method of claim 2 wherein each user profile further comprises fine grain category activities for each category identified in the user profile, each fine grain category activity entry comprising a text string and one or more timestamps corresponding to classification information, the timestamp represents the time when the classification information containing the text string was received.

9. The method of claim 8 wherein when multiple instances of classification information containing the same text string have been received, each timestamp corresponds to the receipt of a different instance of classification information, and wherein the profiling rules decay the significance of the text-timestamp records over time.

10. The method of claim 9 wherein a timestamp older than a predetermined threshold is removed.

11. A system for profiling a user of a computing device, connected to a network based upon a user'"'"'s on-line actions, the system comprising:

an event receiver module for receiving a user event generated by the actions performed by the user on the computing device, the user event identifying event identification information indicating content requested by the user and a user identifier;

a web map module for storing classification information associated a plurality of event identification information instances, the classification information providing a text-score record comprising at least a text string and a score associated with the text string, the text string defined in relation to a lexical ontology comprising a hierarchy of categories and fine grain text, and the score representing a strength of association of the text string to the content;

an event processor module for processing the received user event to determine classification information from the web map based upon the event identification information;

a profile database comprising a plurality of user profiles, each profile identifying classification information associated with a unique user identifier and providing a characterization of the user'"'"'s preferences; and

a profiling engine module for retrieving a user profile form the profile database based upon the received user identifier and for generating update information by applying profiling rules defining known associations between user action events and user behaviours to the retrieved classification information to update the retrieved user profile, wherein the profile update information identifies how to modify the user profile by either adding classification identification information or modifying existing classification identification information score stored in the profile.

12. The system of claim 11 wherein the profiling rules are provided by a modelling system for modelling user behaviour from the user action events to predict user preferences, the system further comprising:

a clickstream collection unit for removing the user identification information from the received user event; and

a modelling data aggregator for aggregating a plurality of de-identified user events and sending the aggregated de-identified user events periodically to the modelling system for use as the user action events;

wherein the profiling engine receives updated profiling rules in response to the sent aggregated de-identified user events, the updated profiling rules used in generating profile update information.

13. The system of claim 12 wherein the value of the score is a scalar value with a decimal value between 0 and 1.

14. The system of claim 13 wherein the value of each score in a user profile associated to one or more categories is decreased by a defined amount on a periodic basis in order to provide a time decay function.

15. The system of claim 14 wherein the event identification information is a universal resource locator (URL).

16. The system of claim 15 wherein when the URL is not present in the web map cache, the URL is sent to a modelling system to perform a mapping function of the location associated with the URL and assign a classification to the URL, the classification being added to the web map.

17. The system of claim 13 wherein each user profile further comprises fine grain category activities for each category identified in the user profile, each fine grain category activity entry comprising a text string and one or more timestamps corresponding to classification information, the timestamp represents the time when the classification information containing the text string was received.

18. The system of claim 17 wherein when multiple instances of classification information containing the same text string have been received, each timestamp corresponds to the receipt of a different instance of classification information, and wherein the profiling rules decay the significance of the text-timestamp records over time.

19. The system of claim 18 wherein a timestamp older than a predetermined threshold is removed.

20. A computer-readable storage medium containing instructions for execution on a processor for performing a method of profiling a user of a computing device, connected to a network, based upon a user'"'"'s on-line actions, the instructions comprising:

receiving a user event generated by actions performed by the user on the computing device, the user event identifying event identification information indicating content requested by the user and a user identifier;

performing a look-up in a cached web map using the event identification information to determine classification information associated with the content requested by the user, the classification information providing a text-score record comprising at least a text string and a score associated with the text string, the text string defined in relation to a lexical ontology comprising a hierarchy of categories, and the score representing a strength of association of the text string to the content;

retrieving classification information associated with the event identification information from the cached web map if classification information is present for the event identification information;

retrieving a user profile associated with the user identification information from a profile database;

generating profile update information by applying profiling rules defining known associations between user action events and user behaviours to the retrieved classification information for the user event, wherein the profile update information identifies how to modify the user profile by adding classification identification information or modifying existing classification identification information scores stored in the profile;

updating the user profile based upon the update information, the user profile providing a characterization of the user'"'"'s preferences; and

storing the updated user profile wherein the user profile can be accessed by an external advertising server.