US 9,311,391 B2Granted

Method and system of content recommendation

AssigneeTelecom Italia S.p.A.

InventorsSkjalg Lepsoy, Gianluca Francini, Fabrizio Antonelli

Life of this patent

Estimating the term…

Loading patent…

Abstract

A method of content recommendation, includes: generating a first digital mathematical representation of contents to associate the contents with a first plurality of words describing the contents; generating a second digital mathematical representation of text documents different from the contents to associate the documents with a second plurality of words; processing the first and second pluralities of words to determine a common plurality of words; processing the first and second digital mathematical representations to generate a common digital mathematical representation of the contents and the text documents based on the common plurality of words; and providing content recommendation by processing the common digital mathematical representation.

Claims8

Dependent claims

A method of media content recommendation, comprising:

generating a first digital mathematical representation of media contents to associate the media contents with a first plurality of words describing the media contents, comprising defining content text documents describing the media contents by said first plurality of words and processing the first plurality of words to generate a first document-word matrix representing said media contents;
generating a second digital mathematical representation of context text documents, different from said content text documents, to associate said context text documents with a second plurality of words, comprising processing the second plurality of words to generate a second document-word matrix representing said context text documents, wherein the context text documents include subjects not connected to the media contents;
merging the first and second pluralities of words to create a common plurality of words;
processing the first and second digital mathematical representations to generate a common digital mathematical representation of the content and context text documents, based on the common plurality of words, comprising processing the first and second document-word matrices and the common plurality of words to generate a third document-word matrix representing the content and context text documents;
and
providing media content recommendation by processing the common digital mathematical representation,
wherein processing the first and second document-word matrices and the common plurality of words to generate a third document-word matrix comprises:
decomposing the first document-word matrix into a product of three first matrices using QR factorisation technique while incorporating modifications of term frequency factors and inverse document frequency factors due to updating of said content and context text documents;
decomposing the second document-word matrix into a product of three second matrices using QR factorisation technique while incorporating modifications of term frequency factors and inverse document frequency factors due to updating of said content and context text documents;
merging the three first matrices and three second matrices to create a common matrix by concatenating the three first matrices as a first element of the common matrix and the three second matrices as a second element of the common matrix;
decomposing the common matrix into a product of three common matrices using QR factorisation technique;
and
defining the third document-word matrix from said product of the three common matrices.

10 more elements

A media content recommendation system comprising at least one processor, memory, and database, the memory comprising:

a first generation module configured to generate a first digital mathematical representation of media contents associated with content text documents comprising a first plurality of words describing the media contents, the first digital mathematical representation comprising a first document-word matrix representing said media contents;
a second generation module configured to generate a second digital mathematical representation of context text documents different from said content text documents and associated with a second plurality of words, the second digital mathematical representation comprising a second document-word matrix representing said context text documents, wherein the context text documents include subjects not connected to the media contents;
a first processing module structured to merge the first and second pluralities of words to create a common plurality of words;
a second processing module structured to process the first and second digital mathematical representations to generate a common digital mathematical representation defining a common representation of the content and context text documents based on the common plurality of words, the common digital mathematical representation comprising a third document-word matrix representing a union of the content and context text documents;
and
a recommendation module configured to provide media content recommendation by processing the common digital mathematical representation,
wherein the second processing module comprises:
a first factorization module configured to decompose the first document-word matrix into a product of three first matrices using QR factorisation technique while incorporating modifications of term frequency factors and inverse document frequency factors due to updating of said content and context text documents;
a second first factorization module configured to decompose the second document-word matrix into a product of three second matrices using QR factorisation technique while incorporating modifications of term frequency factors and inverse document frequency factors due to updating of said content and context text documents;
and
a third factorization module configured to merge the three first matrices and three second matrices to create a common matrix by concatenating the three first matrices as a first element of the common matrix and the three second matrices as a second element of the common matrix, decompose the common matrix into the product of three common matrices using QR factorisation technique, and determine the third document-word matrix from said product of the three common matrices.

8 more elements

Dependencies are inferred from the claim text.