Programming guide content collection and recommendation system for viewing on a portable device

• US 20060123448A1
• Filed: 12/02/2004
• Published: 06/08/2006
• Est. Priority Date: 12/02/2004
• Status: Active Grant

• Current Assignee: Sovereign Peak Ventures, LLC (Dominion Harbor Enterprises, LLC)
• Original Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Inventors: Chen, Wenliang, Ma, Yue, Zhu, Jingbo, Wang, Anhui, Guo, Jinhong, Yao, Tianshun
• Granted Patent: US 7,533,399 B2
• Time in Patent Office: Days
• US Class Current: 725/51
• CPC Class Codes: H04N 21/2353 specifically adapted to content descriptors, e.g. coding, compressing or processing of metadata
• CPC Class Codes: H04N 21/25891 being end-user preferences retrieval of video data in a video database based on user preferences G06F16/739; arrangements for recognizing users' preferences H04H60/46; 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 21/26283 for associating distribution time parameters to content, e.g. to generate electronic program guide data
• CPC Class Codes: H04N 21/41407 embedded in a portable device, e.g. video client on a mobile phone, PDA, laptop constructional details of equipment or arrangements specially adapted for portable computer application G06F1/1626; arrangements specially adapted for mobile receivers in broadcast systems H04H20/57
• CPC Class Codes: H04N 21/4622 Retrieving content or additional data from different sources, e.g. from a broadcast channel and the Internet web site content organization and management for information retrieval from the Internet G06F16/958; transmission by internet of broadcast information H04H60/82; stock exchange data over packet-switching network H04L12/1804; push services including data channel over packet-switching network H04L12/1859
• CPC Class Codes: H04N 21/4668 for recommending content, e.g. movies
• CPC Class Codes: H04N 21/4755 for defining user preferences, e.g. favourite actors or genre retrieval personalisation and generation of user profiles for the retrieval of video data G06F16/739; user profiles in network data switching protocols H04L67/306
• CPC Class Codes: H04N 21/8543 using a description language, e.g. Multimedia and Hypermedia information coding Expert Group [MHEG], eXtensible Markup Language [XML] information retrieval of semistructured data, the underlying structure being taken into account, e.g. mark-up language structure data G06F16/80
• CPC Class Codes: H04N 7/163 by receiver means only
• CPC Class Codes: H04N 7/17318 Direct or substantially direct transmission and handling of requests

Abstract

An EPG contents collection and recommendation system includes an EPG database of identifications of available programs. A program information acquisition module applies text classification to detailed descriptions of the available programs. An EPG recommendation module recommends an available program to a user based on the text classification. Preferably, EPG contents are collected from publicly available TV websites and parsed into a uniform format. For example, contents are vectorized, and a Maximum Entropy technique is applied. Also, user interaction with the EPG database is used to form a user profile database. Further, classifiers are trained based on contents of the user profile database, and these classifiers are used to recommend EPG contents to the user.

Claims

1. An EPG contents collection and recommendation system, comprising:

an EPG database of identifications of available programs;

a program information acquisition module applying text classification to detailed descriptions of the available programs; and

an EPG recommendation module recommending an available program to a user based on the text classification.

2. The system of claim 1, wherein said program information acquisition module tags a new program (tp) based on its detailed information by representing tp as a vector and computing P(c₁|tp), P(c₂|tp), . . . , P(c_i|tp), . . . , P(c_n|tp) for each domain c.

3. The system of claim 1, wherein said program information acquisition module computes P(c₁|tp), P(c₂|tp), . . . , P(c_i|tp), . . . , P(c_n|tp) for each domain according to:

4. The system of claim 2, wherein said program information acquisition module selects a domain c for the new program according to:

c=argmax(P(c_i|tp)).

5. The system of claim 1, further comprising a user profile acquisition module extracting information indicative of user interest in available programs from user interaction with said EPG database.

6. The system of claim 5, wherein said user profile acquisition module reformats the information indicative of user interest into user profile information that records attribute information including one or more of title, time, category, domain, and a flag indicating whether the user liked or disliked the program in question.

7. The system of claim 6, wherein said user profile acquisition module records the user profile information in a user profile database, and maintains the user profile database by removing outdated user profile information.

8. The system of claim 7, further comprising an EPG recommendation learning module training one or more classifiers based on the user profile information.

9. The system of claim 8, wherein said EPG recommendation learning module trains a category classifier by:

(a) computing probability of categories extracted from a record of user profile information according to;

10. The system of claim 8, wherein said EPG recommendation learning module trains a domain classifier by:

(a) computing probabilities of domains extracted from a record of user profile information according to;

11. The system of claim 8, wherein said EPG recommendation learning module trains a content classifier, by:

(a) obtaining a corpus that includes programs marked as liked or disliked by the user according to a record of user profile information;

(b) representing the programs as vectors; and

(c) training the content classifier by using the vectors and applying a maximum entropy technique.

12. The system of claim 8, wherein said EPG recommendation module uses a category classifier trained on user profile information to recommend an available program to the user based on categories recorded in relation to identifications of the available programs.

13. The system of claim 8, wherein said EPG recommendation module uses a domain classifier trained on user profile information to recommend an available program to the user based on domains recorded in relation to identifications of the available programs.

14. The system of claim 8, wherein said EPG recommendation module uses a content classifier trained by applying text classification to detailed descriptions of programs indicated as liked or disliked by the user in user profile information to recommend an available program to the user based on detailed textual descriptions recorded in relation to identifications of the available programs.

15. The system of claim 1, wherein said program information acquisition module obtains contents of multiple EPGs from publicly available TV websites.

16. The system of claim 15, wherein said program information acquisition module parses the contents based on known structures of the EPGs.

17. The system of claim 16, wherein said program information acquisition module determines the known structures based on categorized key phrases.

18. The system of claim 17, wherein said program information acquisition module extracts categorized key phrases based on the known structures of the EPGs.

19. The system of claim 16, wherein said program information acquisition module parses the contents based on known structures by extracting text from predefined locations of documents and tagging the text based on predefined categories associated with the locations in a predefined EPG structure template for the TV website from which the contents are obtained.

20. The system of claim 16, wherein said program information acquisition module parses the contents based on known structures by retagging meta-tagged text based on known correspondence between tags used by a TV website and the predefined categories.

21. An EPG contents collection and recommendation method, comprising:

accessing an EPG database of identifications of available programs;

applying text classification to detailed descriptions of the available programs; and

recommending an available program to a user based on the text classification.

22. The method of claim 21, wherein applying text classification includes tagging a new program (tp) based on its detailed information by representing tp as a vector and computing P(c₁|tp), P(c₂|tp), . . . , P(c_i|tp), . . . , P(c_n|tp) for each domain c.

23. The method of claim 21, wherein P(c₁|tp), P(c₂|tp), . . . , P(c_i|tp), . . . , P(c_n|tp) are computed for each domain according to:

24. The method of claim 22, wherein applying text classification includes selecting a domain c for the new program according to:

c=argmax(P(c_i|tp)).

25. The method of claim 21, further comprising extracting information indicative of user interest in available programs from user interaction with the identifications of the available programs.

26. The method of claim 25, further comprising reformatting the information indicative of user interest into user profile information that records attribute information including one or more of title, time, category, domain, and a flag indicating whether the user liked or disliked the program in question.

27. The method of claim 26, further comprising recording the user profile information.

28. The method of claim 27, further comprising training one or more classifiers based on the user profile information.

29. The method of claim 28, further comprising training a category classifier, including:

computing probability of categories extracted from a record of user profile information according to;

$P<br><br>\text{\hspace{1em}<br><br>}<br><br>\left(c_i\right)=\frac{N<br><br>\text{\hspace{1em}<br><br>}<br><br>\left(c_i\right)}{\underset{j=1}{\overset{<br><br>C<br><br>}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>N<br><br>\text{\hspace{1em}<br><br>}<br><br>\left(c_j\right)},$

where C denotes a set of categories, c_i denotes a category, and N(c_i) denotes frequency of c_i; and

sorting categories according to the probabilities to obtain a list indicating user interest relating to categories of available programs.

30. The method of claim 28, further comprising training a domain classifier, including:

computing probabilities of domains extracted from a record of user profile information according to;

$P<br><br>\text{\hspace{1em}<br><br>}<br><br>\left(d_i\right)=\frac{N<br><br>\text{\hspace{1em}<br><br>}<br><br>\left(d_i\right)}{\underset{j=1}{\overset{<br><br>D<br><br>}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>N<br><br>\text{\hspace{1em}<br><br>}<br><br>\left(d_j\right)},$

where D denotes a set of domains, d_i denotes a domain, and N(s_i) denotes frequency of s_i; and

sorting the domains according to the probabilities, to obtain a list indicating user interest relating to domains of available programs.

31. The method of claim 28, further comprising training a content classifier, including:

obtaining a corpus that includes programs marked as liked or disliked by the user according to a record of user profile information;

representing the programs as vectors;

training the content classifier by using the vectors and applying a maximum entropy technique.

32. The method of claim 28, further comprising recommending an available program to the user based on categories recorded in relation to identifications of the available programs, including using a category classifier trained on user profile information.

33. The method of claim 28, further comprising recommending an available program to the user based on domains recorded in relation to identifications of the available programs, including using a domain classifier trained on user profile information.

34. The method of claim 28, further comprising recommending an available program to the user based on detailed textual descriptions recorded in relation to identifications of the available programs, including using a content classifier trained by applying text classification to detailed descriptions of programs indicated as liked or disliked by the user in user profile information.

35. The method of claim 21, further comprising obtaining contents of multiple EPGs from publicly available TV websites.

36. The method of claim 35, further comprising parsing the contents based on known structures of the EPGs.

37. The method of claim 36, further comprising, determining the known structures based on categorized key phrases.

38. The method of claim 37, further comprising extracting categorized key phrases based on the known structures of the EPGs.

39. The method of claim 36, wherein parsing the contents based on known structures includes extracting text from predefined locations of documents and tagging the text based on predefined categories associated with the locations in a predefined EPG structure template for a TV website from which the contents are obtained.

40. The method of claim 36, wherein parsing the contents based on known structures includes retagging meta-tagged text based on known correspondence between tags used by a TV website and the predefined categories.