COMPUTING SIMILARITY BETWEEN MEDIA PROGRAMS

• US 20150052564A1
• Filed: 10/15/2014
• Published: 02/19/2015
• Est. Priority Date: 10/30/2011
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Glowaty, Grzegorz, Brzozowski, Michal, Wielgus, Marcin
• Granted Patent: US 9,654,834 B2
• Time in Patent Office: Days
• US Class Current: 725/53
• CPC Class Codes: G06F 16/78 Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
• CPC Class Codes: H04N 21/435 Processing of additional data, e.g. decrypting of additional data, reconstructing software from modules extracted from the transport stream
• 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 21/4828 for searching program descriptors retrieval of video data G06F16/739

Abstract

System and method are provided to associate or compare media programs. A method includes: obtaining, using at least one processing circuit, first metadata for a first media program and second metadata for a second media program, wherein the first metadata are organized into a plurality of first fields, and the second metadata are organized into a plurality of second fields; extracting, using at least one processing circuit, a plurality of first tokens from one of the plurality of the first fields and a plurality of second tokens from one of the plurality of second fields; assigning a weight factor to each of the first and second tokens; cross-correlating the first and second tokens between the plurality of first fields and the plurality of second fields; and calculating a similarity score between the first and second media programs based on the cross-correlating.

Claims

1. A method for recommending media programs, the method comprising:

obtaining, using a hardware processor, metadata corresponding to a plurality of media programs, wherein the metadata corresponding to each media program comprises a plurality of metadata fields;

extracting at least one token from a first metadata field corresponding to each of the plurality of programs;

generating a first group of components based on the first metadata field for each of the plurality of media programs, wherein at least a subset of components in the first group of components for a media program of the plurality of media programs represent the tokens extracted from the first metadata field for the media program;

receiving information identifying a first media program of the plurality of media programs;

comparing the first group of components for the first media program to the first group of components for each other media program of the plurality of media programs to identify media programs for which the first metadata field includes at least one common token with the first metadata field of the first media program;

determining a similarity score between the first media program and each of the other media programs based at least in part on the comparison; and

recommending at least one of the plurality of media programs other than first media program based on the similarity score.

2. The method of claim 1, wherein the first group of components is an ordered group of components, where each position in the ordered group corresponds to one or more tokens

3. The method of claim 2, wherein the group of components is a vector.

4. The method of claim 1, wherein a first token of the at least one token comprises text in the first metadata field.

5. The method of claim 4, wherein the first token of the at least one token comprises a combination of text keywords included in the first metadata field.

6. The method of claim 1, further comprising calculating a term frequency for the at least one token equal to the sum of the logarithm of the number of the token'"'"'s occurrences in the first metadata fields corresponding to the plurality of programs and 1.

7. The method of claim 1, wherein the at least one token is extracted from the first metadata field corresponding to each of the plurality of programs using a probabilistic topic model.

8. The method of claim 1, further comprising:

identifying a token having a frequency of occurrence in the first metadata fields corresponding to the plurality of programs exceeding a predetermined threshold; and

inhibiting the identified token from being used to generate the first group of components.

9. The method of claim 1, further comprising:

calculating, for each component in the first group of components corresponding to each of the plurality of media programs, a weight factor based on the number of occurrences of the token corresponding to a particular component of the group of components in the first metadata field corresponding to a particular media program of the plurality of media programs and the number of first metadata fields corresponding to the plurality of programs in which the token corresponding to the particular component occurs; and

weighting each component in the first group of components corresponding to each of the plurality of media programs based on the calculated weight factors.

10. A system for recommending media programs, the system comprising:

a hardware processor that is programmed to;

obtain metadata corresponding to a plurality of media programs, wherein the metadata corresponding to each media program comprises a plurality of metadata fields;

extract at least one token from a first metadata field corresponding to each of the plurality of programs;

generate a first group of components based on the first metadata field for each of the plurality of media programs, wherein at least a subset of components in the first group of components for a media program of the plurality of media programs represent the tokens extracted from the first metadata field for the media program;

receive information identifying a first media program of the plurality of media programs;

compare the first group of components for the first media program to the first group of components for each other media program of the plurality of media programs to identify media programs for which the first metadata field includes at least one common token with the first metadata field of the first media program;

determine a similarity score between the first media program and each of the other media programs based at least in part on the comparison; and

recommend at least one of the plurality of media programs other than first media program based on the similarity score.

11. The system of claim 10, wherein the first group of components is an ordered group of components, where each position in the ordered group corresponds to one or more tokens

12. The system of claim 11, wherein the group of components is a vector.

13. The system of claim 10, wherein a first token of the at least one token comprises text in the first metadata field.

14. The system of claim 13, wherein the first token of the at least one token comprises a combination of text keywords included in the first metadata field.

15. The system of claim 10, wherein the hardware processor is further programmed to calculate a term frequency for the at least one token equal to the sum of the logarithm of the number of the token'"'"'s occurrences in the first metadata fields corresponding to the plurality of programs and 1.

16. The system of claim 10, wherein the at least one token is extracted from the first metadata field corresponding to each of the plurality of programs using a probabilistic topic model.

17. The system of claim 10, wherein the hardware processor is further programmed to:

identify a token having a frequency of occurrence in the first metadata fields corresponding to the plurality of programs exceeding a predetermined threshold; and

inhibit the identified token from being used to generate the first group of components.

18. The system of claim 10, wherein the hardware processor is further programmed to:

calculate, for each component in the first group of components corresponding to each of the plurality of media programs, a weight factor based on the number of occurrences of the token corresponding to a particular component of the group of components in the first metadata field corresponding to a particular media program of the plurality of media programs and the number of first metadata fields corresponding to the plurality of programs in which the token corresponding to the particular component occurs; and

weight each component in the first group of components corresponding to each of the plurality of media programs based on the calculated weight factors.

19. A non-transitory computer-readable medium containing computer executable instructions that, when executed by a processor, cause the processor to perform a method of recommending media programs, the method comprising:

obtaining metadata corresponding to a plurality of media programs, wherein the metadata corresponding to each media program comprises a plurality of metadata fields;

extracting at least one token from a first metadata field corresponding to each of the plurality of programs;

generating a first group of components based on the first metadata field for each of the plurality of media programs, wherein at least a subset of components in the first group of components for a media program of the plurality of media programs represent the tokens extracted from the first metadata field for the media program;

receiving information identifying a first media program of the plurality of media programs;

comparing the first group of components for the first media program to the first group of components for each other media program of the plurality of media programs to identify media programs for which the first metadata field includes at least one common token with the first metadata field of the first media program;

determining a similarity score between the first media program and each of the other media programs based at least in part on the comparison; and

recommending at least one of the plurality of media programs other than first media program based on the similarity score.