System, method, and software application for targeted advertising via behavioral model clustering, and preference programming based on behavioral model clusters

• US 8,046,797 B2
• Filed: 01/09/2002
• Issued: 10/25/2011
• Est. Priority Date: 01/09/2001
• Status: Expired due to Fees

• Current Assignee: Thomson Licensing SAS (Vantiva SA)
• Original Assignee: Thomson Licensing (Vantiva SA)
• Inventors: Zhou, Yiming, Ismail, Labeeb K., Bentolila, Isaac, Humpleman, Richard
• Primary Examiner(s): NGUYEN BA, HOANG VU A
• Application Number:
  US10/043,714
• Publication Number:
  US 20030101451A1
• Time in Patent Office: 3,576 Days
• Field of Search: 725/35, 725/36, 725/45, 725/46, 725/34, 705/14.49, 705/14.52, 705/14.53, 705/14.66, 705/14.67, 707/899
• US Class Current: 725/46
• CPC Class Codes: G06Q 30/0251 Targeted advertisements
• CPC Class Codes: G06Q 30/0254 based on statistics
• CPC Class Codes: G06Q 30/0255 based on user history
• CPC Class Codes: G06Q 30/0269 based on user profile or attribute
• CPC Class Codes: G06Q 30/0271 Personalized advertisement
• CPC Class Codes: H04N 21/252 Processing of multiple end-users' preferences to derive collaborative data
• CPC Class Codes: H04N 21/2668 Creating a channel for a dedicated end-user group, e.g. insertion of targeted commercials based on end-user profiles information retrieval from the Internet by querying with filtering and personalisation G06F16/9535; arrangements for replacing or switching information during the broadcast H04H20/10; push services over packet-switching network H04L12/1859; adaptation of message content in packet-switching networks H04L51/063
• CPC Class Codes: H04N 21/44222 Analytics of user selections, e.g. selection of programs or purchase activity monitoring of user selections in data processing systems G06F11/34; arrangements for monitoring the user's behaviour or opinions in broadcast systems H04H60/33
• CPC Class Codes: H04N 21/4532 involving end-user characteristics, e.g. viewer profile, preferences monitoring of user activities for profile generation for accessing a video database G06F16/739; user profiles in network data switching protocols H04L67/306; processing of user preferences or user profiles in wireless networks H04W8/18
• CPC Class Codes: H04N 21/454 Content or additional data filtering, e.g. blocking advertisements
• CPC Class Codes: H04N 21/466 Learning process for intelligent management, e.g. learning user preferences for recommending movies services using the results of monitoring in broadcast systems H04H60/61
• CPC Class Codes: H04N 21/4663 involving probabilistic networks, e.g. Bayesian networks
• CPC Class Codes: H04N 21/4667 Processing of monitored end-user data, e.g. trend analysis based on the log file of viewer selections
• CPC Class Codes: H04N 21/4826 using recommendation lists, e.g. of programs or channels sorted out according to their score
• CPC Class Codes: H04N 7/163 by receiver means only
• CPC Class Codes: H04N 7/17309 Transmission or handling of upstream communications

Abstract

The method and system for TV user profile data prediction and modeling allows accurate and narrowly focused behavioral clustering. A client-side system classifies television consumers into representative user profiles. The profiles target individual user advertising and program preference category groups. A contextual behavioral profiling system determines the user'"'"'s monitor behavior and content preferences, and the system may be continually updated with user information. A behavioral model database is queried by various system modules. The programming, including targeted advertising for television and interactive television is based on the profile data prediction, modeling and preference determination. The system is enabled to present a complete program sequence to the viewer based on the preference determination and stored programming. The latter is referred to as automatic program sequence (virtual channel) creation and the virtual channel can be presented as a separate channel in an electronic programming guide (EPG).

Claims

1. A television rating computer system for targeted program delivery, comprising:

a server-side computer system for evaluating television behavioral viewing data from a plurality of users and for categorizing data into non-demographically classifiable category groups;

a clustering engine included in the server-side computer system for;

receiving the television behavioral viewing data, processing the television behavioral viewing data;

using a category training set for clustering the television behavioral viewing data into the non-demographically classifiable category groups over a predetermined training period;

generating category behavioral profiles targeting the non-demographically classifiable category groups; and

generating advertising category prototypes by removing television behavioral viewing data parameters most common between the category behavioral profiles;

a client-side computer system coupled to the server-side computer system and configured to classify a television user into at least one of the non-demographically classifiable category groups based on advertising category prototypes received from the clustering engine;

a contextual behavioral profiling agent included in the client-side system for deriving profiling information related to a television user'"'"'s viewing behavior with content and usage-related preferences; and

a behavioral model database for storing in the client-side system the profiling information derived by the profiling agent.

2. The television rating system according to claim 1, wherein said clustering engine is a software agent residing in a central computer system at a television distribution head-end in the server side system and is programmed to create template behavioral profiles each corresponding to an associated one of the targeted category groups.

3. The television rating system according to claim 2, wherein said clustering engine is trained substantially exclusively on tagged viewing data from a given target group to learn a most general profile of the given target group.

4. The television rating system according to claim 2, wherein said clustering engine is programmed to generalize user profiles in a targeted category group into an aggregation representative of all dimensions most strongly in common for the targeted group and all dimensions most unique across several of the targeted category groups.

5. The television rating system according to claim 1, which further comprises an advertisement manager residing at the server-side system and connected to query said behavioral model database in the client-side system, said advertisement manager being programmed to parameterize behavioral profiles of said behavioral model database and to download the parameterized behavioral profiles to an advertising category membership agent residing at said client-side system.

6. The television rating system according to claim 5, wherein said advertisement manager includes a television user'"'"'s history and is configured to reconstruct the downloaded parameterized behavior profiles in accordance with the television user'"'"'s history to determine a most likely advertising category for the television user, and to store the results as targeting category probabilities in a user category database.

7. The television rating system according to claim 5, which further comprises targeting agents and presentation agents disposed at said client-side system for creating an optimization of targeted category probabilities and relevant preference information in order to selectively capture, store, and display advertisements downloaded in accordance with the optimization.

8. The television rating system according to claim 1, wherein the television behavioral viewing data from the plurality of users is used to create the category groups.

9. The television rating system according to claim 8, wherein the profiling information is used to determine a category group to associate a user with.

10. The television rating system according to claim 1, wherein the television behavioral viewing data includes contextual transition data.

11. The television rating system according to claim 10, wherein the contextual transition data is based on day of week and time of day.

12. The television rating system according to claim 10, wherein the contextual transition data is based on a previous type of television program.

13. The television rating system according to claim 1, wherein the category training set is initially a preexisting collection of advertising categories.

14. The television rating system according to claim 13, wherein all viewed programs within a given predefined time interval are rated and sorted at once to provide a behavior model estimate associated with the client side system.

15. The television rating system according to claim 1, wherein the advertising category prototypes are formed during the predetermined training period.

16. The television rating system according to claim 1, wherein the predetermined training period is continuously adjusted.

17. The television rating system according to claim 1, wherein the predetermined training period duration is different based on the corresponding classifiable category group.

18. In an interactive display system having a head-end side for distributing program content that has been pruned for a category, and a client side receiving the program content and selectively displaying the program content in accordance with the selection of a user, a preference engine for determining a preferred program content for the user, comprising:

a user monitoring device receiving the pruned program content at the client side for recording contextual transition behaviors profiling the user to continually build a user profile of preferences and contextual transition behaviors associated with the user; and

a program distributing device at the head-end side for providing to the user the program content in accordance with the user profile, wherein a user is classified at the client-side into at least one category group based on advertising non-demographically classifiable category prototypes received from the head-end side, wherein the advertising non-demographically classifiable category prototypes are generated by removing television viewing data parameters most common between category behavioral profiles formed at the head-end side by using a category training set for clustering television behavioral viewing data from a plurality of users into non-demographically classifiable category groups over a predetermined training period.

19. The interactive display system and preference engine according to claim 18, wherein said user monitoring device models the user'"'"'s behavioral interaction with advertising program content and with entertainment program content.

20. The interactive display system and preference engine according to claim 18 wherein the program distributing device is connected to receive from the head-end metadata information describing advertising content and entertainment program content, and is programmed to adjust the user profile by combining the metadata information with the preferences and contextual transition behaviors of the user, and to build a relational knowledge base with associations among the behavior, demographics, and program content preferences of the user.

21. The interactive display system and preference engine according to claim 18 wherein the user maintaining device is programmed to model patterns of usage behaviors with a behavioral model and to extract key usage information from the behavioral model into a behavioral database having a confidence value that reflects an estimate of a structural and sampling quality of the data in the behavioral database.

22. In a program content delivery computer system having a head-end side and a client side, a system for targeting program delivery, comprising:

a central data system at the head-end side which receives viewing information from a plurality of users selected from the group consisting of watch data, watch start time data, watch duration data, and watch channel data, demographic information describing a program user, and electronic program guide information with metadata describing a program content;

a demographic cluster knowledge base acquirer receiving from the client side behavioral data of a user, the knowledge base acquirer outputting a knowledge base based on the viewing information in the form of a transition matrix with weight sets, the transition matrix used for predicting a particular category group of a plurality of classifiable category groups for classifying the user based on the behavioral data of the user; and

a program content generating module disposed at the head-end side and providing to the client side streams of program content based on the predicted category group of the user, wherein a user is classified into at least one category group based on advertising category prototypes transmitted from the head-end side, wherein the advertising category prototypes are generated at the head-end side by removing television viewing data parameters most common between category behavioral profiles formed at the head-end side by using a category training set for clustering the viewing information from the plurality of users into the plurality of classifiable category groups over a predetermined training period.

23. The program content delivery system according to claim 22, which further comprises a realtime feedback link for delivering to said central data system at the head-end side realtime information with click stream data concerning the viewing behavior of the user.

24. The program content delivery system according to claim 22, wherein said demographic cluster knowledge base acquirer is based on a hidden Markov model.

25. The program content delivery system according to claim 24, wherein said demographic cluster knowledge base acquirer is configured to parameterize the user'"'"'s behavior with a double random pseudo hidden Markov process, and to define a low-level statistical state machine modeling a behavioral cluster and a top-level statistical state machine with active behavioral clusters and an interaction among the active behavioral clusters.

26. The program content delivery system according to claim 22, wherein said demographic cluster knowledge base acquirer and said program content generating module are software modules each adapted to be stored on a machine readable medium in the form of a plurality of processor-executable instructions.

27. The program content delivery system according to claim 22, wherein said demographic cluster knowledge base acquirer generates demographic cluster information of the user in terms of statistical state machine transition models.

28. The program content delivery system according to claim 27, wherein the statistical state machine transition models are defined in a transition matrix at the head-end side, and the transition matrix contains information of program transitions initiated by a viewer at the client side.

29. The program content delivery system according to claim 22, wherein the transition matrix is one of at least two concurrent transition matrices including a channel matrix and a genre matrix.

30. The program content delivery system according to claim 22, wherein the transition matrix is a two-dimensional matrix with transitions from television channels in normal form to television channels in temporal form.

31. The program content delivery system according to claim 22, wherein said demographic cluster knowledge base acquirer is configured to define a double random process with a plurality of dimensions, and to determine parallel statistical state machine transition events in at least two of three state categories including channel, genre, and title of the program content.