METHOD AND SYSTEM FOR CONTENT CLASSIFICATION

• US 20110264699A1
• Filed: 12/30/2008
• Published: 10/27/2011
• Est. Priority Date: 12/30/2008
• Status: Active Grant

• Current Assignee: Telecom Italia S.p.A.
• Original Assignee: Telecom Italia S.p.A.
• Inventors: Cataldi, Mario, Sapino, Maria Luisa, Simeoni, Rossana, Di Caro, Luigi, Antonelli, Fabrizio, Mana, Dario, Geymonat, Marina, Candan, Selcuk Kasim
• Granted Patent: US 9,916,381 B2
• Time in Patent Office: Days
• US Class Current: 707/777
• CPC Class Codes: G06F 16/353 into predefined classes

Abstract

A processing method for classification (300) of contents (400) in a domain (500; 501) that can be represented through a taxonomy is described, the method comprising: generating (301) a first digital mathematical representation of the taxonomy; generating (302) a second digital mathematical representation of text documents (600) different from said contents and containing keywords; processing (303) the first and second digital mathematical representations for enriching the taxonomy, by associating keywords of the text documents (600) with the first digital mathematical representation; generating (304) a third digital mathematical representation of the contents (400); processing (305) the first digital enriched mathematical representation and third mathematical representation for classifying the contents (400) in the enriched taxonomy.

Claims

1. A processing method for classification (300) of contents (400) in a domain (500;

501) that can be represented through taxonomy, the method comprising;

generating (301) a first digital mathematical representation of the taxonomy;

generating (302) a second digital mathematical representation of text documents (600) different from said contents and containing keywords;

processing (303) the first and second digital mathematical representation to enrich the taxonomy, by associating keywords of the text documents (600) with the first digital mathematical representation;

generating (304) a third digital mathematical representation of the contents (400);

processing (305) the first enriched digital mathematical representation and third digital mathematical representation for classifying the content (400) in the enriched taxonomy.

2. A method according to claim 1, also comprising:

storing in a database (103) the data representative of said contents and data indicative of the relative classification in the taxonomy.

3. A method according to claim 1 wherein:

generating (301) the first digital mathematical representation includes generating first keyword vectors representative of the concepts of taxonomy (500).

generating (302) the second digital mathematical representation (600) includes generating (302) second keyword vectors representative of text documents (600).

generating (304) the third digital mathematical representation includes generating third keyword vectors representative of the contents to be classified.

enriching (305) the taxonomy includes processing the first and second keyword vectors in order to associate keywords of the second vectors to concepts of the first vectors.

4. A method according to claim 1 or 2 wherein:

said first digital mathematical representation includes weights (w_par;

w_chid) indicating a contribution to a concept of taxonomy (500) supplied by a plurality of concepts of the taxonomy;

said second digital mathematical representation includes respective weights associated with keywords of the text documents (600), each indicating a relevance of the respective keyword;

said third digital mathematical representation includes additional weights associated with additional keywords of the contents (400), each indicating a corresponding relevance of the respective additional keyword.

5. A method according to claim 4, wherein generating the first digital mathematical representation comprises determining said weights by means of an iterative calculation (700), carried out starting from descriptive propagation coefficients of a degree of propagation between concepts of the taxonomy (500;

501).

6. A method according to claim 4, wherein generating the first digital mathematical representation (800) comprises:

defining (801) a propagation matrix by adjacency (M(m,m)) associated with the taxonomy and representative of contributions of concepts toward other concepts of the taxonomy;

defining (802) a matrix of concept vectors (CV(m,m)) representative of the taxonomy (500);

defining (803) a matrix of propagation effect depending upon the matrix by adjacency of the matrix of concept vectors;

calculating (804) a matrix representing the taxonomy (CV_final) obtained from the matrix of propagation by adjacency, from the matrix of propagation effect and from the matrix of concept vectors.

7. A method according to claim 6, wherein said matrix of propagation by adjacency is defined by a plurality of propagation coefficients of keys from a child node of the taxonomy to a parent node of the taxonomy.

8. A method according to claim 1 or 4, wherein the generation (301) of the first digital mathematical representation of the taxonomy includes the application of a concept propagation method.

9. A method according to claim 1, also comprising:

defining at least one further taxonomy (501);

calculating a respective further digital mathematical representation of said at least one further taxonomy;

processing the first digital mathematical representation and the respective further digital mathematical representation determining semantic relationships between concepts of the taxonomy and the said at least a further taxonomy;

implementing the method of content classification to the taxonomy as related to said at least a further taxonomy.

10. A method according to claim 1, wherein enriching (303) the first digital mathematical representation with keywords of the text documents (600) comprises:

implementing a similarity function (360) in order to classify the context documents to the taxonomy, by associating the text documents with concepts of the taxonomy;

selecting (307, 370) representative keywords associated with said context documents to be used for enrichment;

vectorial summing of one of the first keyword vectors and a vectors of selected keywords in accordance with factors taking into account the respective statistical significance.

11. A method according to claims 3 and 10, wherein:

implementing a similarity function includes evaluating a scalar product between the first and second keyword vectors;

selecting representative keywords includes updating (374) the first vectors in such a way as to reflect a significance of the second vector keywords within the concept extensions.

12. A method according to claim 11, wherein updating the first vectors comprises:

interpreting the first vectors as a query to a document database, and the extension of a concept like the set of documents returned by the query and recognized as significant;

updating the first vectors in order to take into account the weights of keywords present in documents returned by the query and recognized as significant.

13. A method according to claim 12, wherein updating the first vectors is performed through a relevance feedback method by associating the first keyword vectors with interrogation queries of the text documents.

14. A method according to claim 1, wherein the content classification processing (305) comprises:

implementing a respective similarity function between the first digital mathematical representation resulting from said enrichment and the third digital mathematical representation of the contents.

15. A method according to claim 1, wherein:

said domain is one or more of the following domains;

cinematic genres, as well as geographical locations, historic events, art, sport, artists;

said contents are cinematic contents; and

said text documents are newspaper articles.

20. A computer program comprising a program code configured to implement the processing method according to at least one of claims from 1 to 15.

16. A computer system (100) of content classification (400) in a domain that can be represented by means of a taxonomy (500), the system comprising:

a first generation module (201) configured to generate a first digital mathematical representation of the taxonomy;

a second generation module (202) configured to generate a second digital mathematical representation of text documents (600);

an enrichment module (203) configured to enrich the first digital mathematical representation such as to process the first and second digital mathematical representation and associate keywords of the text documents with the taxonomy;

a third generation module (204) configured to generate a third digital mathematical representation of contents (400) to be classified;

a content classification module (205) configured to process the first and third digital mathematical representation for content classification (400) in the enriched taxonomy (500).

17. A system according to claim 16, also comprising:

a database (103) containing data representing said contents and their relative classification;

the database being consultable in order to identify at least one content through a link to another content associated with a keyword of said text documents.

18. A system according to claim 16, wherein the enrichment module comprises:

a module configured to apply similarity function (206) between the first and second digital mathematical representations in order to classify the context documents in said taxonomies;

a further module configured to select representative keywords (207) associated with said context documents and to be used for enrichment.

19. A system according to claim 18 wherein:

the first generation module (201) is configured to generate first vectors of keywords representative of concepts of the taxonomy (500);

the second generation module (202) is configured to generate second keyword vectors representative of concepts of text documents (600);

the enrichment module (203) is configured to enrich the first and second keyword vectors and to associate keywords of the second vectors to concepts of the first vectors;

the third generation module (204) is configured to generate third keyword vectors representative of the contents to be classified.