CONTENT CLASSIFICATION SYSTEM

• US 20170164011A1
• Filed: 08/18/2010
• Published: 06/08/2017
• Est. Priority Date: 08/18/2009
• Status: Active Grant

• Current Assignee: Jinni
• Original Assignee: Jinni
• Inventors: Rimon, Mori, Meshulam, Ram, Ben-Zaken, Izik, Glick, Yosi
• Granted Patent: US 10,536,728 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/353 into predefined classes
• CPC Class Codes: H04N 21/232 Content retrieval operation locally within server, e.g. reading video streams from disk arrays storage management G06F3/0604; details of querying and searching of video data from a database G06F16/739

Abstract

A method includes operating a content crawler over a machine communication network to form a set of input content; operating a mapper on a machine memory comprising the input content to form non-transitory machine logic comprising genes for the content, wherein operating the mapper comprises the application of previously mapped genes for the input content to the mapping of subsequent genes for the input content; and applying the genes to a machine memory storing a database of content from which user recommendations are formed.

Claims

1-14. -14. (canceled)

15. A method, comprising:

operating a content crawler over a machine communication network to form a set of input content;

operating one or more machine processors to execute a mapper on a machine memory comprising the set of input content to form non-transitory machine logic comprising subsequent genes for the set of input content, the mapper reducing machine memory utilization for the set of input content by applying previously mapped genes for the set of input content to the mapping of the subsequent genes for the set of input content;

operating the one or more machine processors to apply the subsequent genes for the set of input content to form a machine operable playlist identifying selected digital files for delivery over the machine communication network to a user device;

the subsequent genes for the set of input content comprising multi-valued, degreed, and polarized sentiment ratings for the set of input content, each of the sentiment ratings a measure of a human subjective mood as determined from linguistic features of the set of input content; and

operating the one or more machine processors to apply the machine operable playlist to deliver a personalized content channel over the machine communication network for display by the user device.

16. The method of claim 15, further comprising:

the mapper applying a plurality of mapping strategies to map the previously mapped genes for the set of input content and to map the subsequent genes for the input content; and

the mapper employing voting procedures to resolve differences of outcome among the plurality of mapping strategies.

17. The method of claim 15, further comprising:

the mapper distinguishing between salience scores and confidence values for each gene of the previously mapped genes for the set of input content and the subsequent genes for the set of input content.

18. The method of claim 15, further comprising:

the mapper identifying discourse structure from the linguistic features of the set of input content; and

the mapper increasing a weight applied to the linguistic features of the set of input content that appear in the discourse structure deemed to be more important than others to a relevance of each gene of the previously mapped genes for the set of input content and the subsequent genes for the set of input content.

19. The method of claim 15, further comprising:

the mapper tagging a particular input content item of the set of input content with one or more sentiment gene, reflecting subjective experience-related expressions in multiple free text sources for the particular input content item;

the mapper determining collective verdicts for the sentiment gene if there is a high degree of agreement among the free text sources in regard to the particular input content item; and

the mapper marking the sentiment gene with a controversy indicator if a consensus does not exist among the free text sources in regard to the particular input content item.

20. The method of claim 19, further comprising:

the collective verdicts reflecting sentiment negation of the particular input content item.

21. The method of claim 15, further comprising:

the mapper applying filtering to the mapping of the subsequent genes for the set of input content to resolve conflicts among the subsequent genes for the set of input content and to associate a score and a confidence value to each of the subsequent genes for the set of input content.

22. The method of claim 15, further comprising:

the mapper organizing the previously mapped genes for the set of input content and the subsequent genes for the set of input content in a taxonomy which distinguishes between content category types and comprises a hierarchy within the content category types.

23. The method of claim 15, further comprising:

the mapper organized as a cascade of machine procedures, each procedure in the cascade of machine procedures configured to map at least one content category types to gene mappings; and

the gene mappings for a first content category type produced in a first machine procedure at a first stage of the cascade of machine procedures utilized to produce next gene mappings for a second content category type in a second machine procedure at a second stage of the cascade of machine procedures latter to the first stage.

24. The method of claim 15, further comprising:

the mapper generating vectors of content genes for two content items and determining a level of semantic similarity between the two content items using a mathematical distance function between the vectors of content genes.

25. The method of claim 15, further comprising:

the mapper generating a reference semantic vector for a reference set of input content comprising content items previously rated by a human user;

the mapper grouping the reference semantic vector into taste clusters wherein each of the taste clusters represents one of multiple tastes of the user; and

the mapper building a user taste profile based on the multiple tastes of the user.

26. The method of claim 25, further comprising:

applying the user taste profile to generate recommended content for a user by matching gene preference levels of the user taste profile with gene values in the reference semantic vector.

27. The method of claim 25, further comprising:

applying taste profiles of a group of users to determine a joined taste for the group of users and applying the joined taste to unify and prioritize group recommended content for the group of users.

28. The method of claim 25, further comprising:

applying the user taste profile to predict a likely user rating for a particular content item represented as a particular content item semantic vector, by calculating a distance between the particular content item semantic vector and a user taste vector having a closest distance to the particular content item.

33. The method of claim 15, further comprising:

forming the user recommendations into a linear playlist defining a television channel to stream to the user device over the machine communication network.

29. A system, comprising:

a content crawler to operate over a machine communication network to form a set of input content from multiple third party digital sources;

non-transitory mapper logic that when applied to one or more machine processors cause the processors to operate on a machine memory comprising the set of input content to form non-transitory machine logic comprising subsequent genes for the set of input content, wherein the mapper logic reduces machine memory utilization for the set of input content by causing the one or more processors to apply previously mapped genes for the set of input content to the mapping of the subsequent genes for the set of input content;

a machine storage system comprising a database of content;

non-transitory recommendation logic that when applied to the one or more machine processors cause the processors to form user recommendations of content from the database of content, the recommendations based upon application of the subsequent genes for the set of input content formed by the mapper logic, the user recommendations comprising a machine operable playlist identifying selected digital files for delivery over the machine communication network to a user device;

the subsequent genes for the set of input content comprising multi-valued, degreed, and polarized sentiment ratings for the set of input content, each sentiment rating a measure of a human subjective mood as determined from linguistic features of the set of input content; and

channel logic that when applied to the one or more machine processors cause the processors to apply the machine operable playlist to deliver the selected digital files as a personalized content stream over the machine communication network.

30. The system of claim 29, further comprising:

the mapper logic employing a plurality of mapping strategies for the previously mapped genes for the set of input content and the subsequent genes for the set of input content; and

the mapper logic employing voting procedures to resolve differences of outcome among the plurality of mapping strategies.

31. The system of claim 29, further comprising:

the mapper logic distinguishing between salience scores and confidence values for each gene of the previously mapped genes for the set of input content and the subsequent genes for the set of input content.

32. The system of claim 29, further comprising:

the mapper logic identifying discourse structure from the linguistic features of the set of input content; and

the mapper logic applying weights to the linguistic features of the set of input content that appear in the discourse structure.

34. The system of claim 29, further comprising:

the channel logic operating responsive to a linear playlist comprising the user recommendations to stream a television channel to the user device.