Systems, methods, and computer program products for providing contextually-aware video recommendation

Patent metadata

Field	Value
Publication number	US-9451329-B2
Application number	US-201414509396-A
Country	US
Kind code	B2
Filing date	Oct 8, 2014
Priority date	Oct 8, 2013
Publication date	Sep 20, 2016
Grant date	Sep 20, 2016

Abstract

Methods, systems and computer program products are provided for providing content recommendation by obtaining metadata associated with a media object, extracting from the metadata a plurality of terms associated with the media object, and mapping at least a portion of the plurality of terms to buckets. A query vector having attributes corresponding to the buckets is used to perform a query on a database storing media object documents having attributes corresponding to the buckets.

First claim

What is claimed is:

1. A method of providing a database of content-level attributes associated with media objects, the method comprising:
performing on at least one computer of a content recommendation system, the at least one computer having a query interface for receiving a query containing at least one content-level attribute, the steps of:
obtaining, by crawling a network using a server adapted to gather text data, metadata associated with a media object from a plurality of data sources, wherein at least one of the plurality of data sources is an unstructured data source and the metadata includes extrinsic metadata corresponding to one or more content-level attributes of the media object;
extracting from the metadata a plurality of terms associated with the media object by applying an entity extraction model to the metadata;
mapping at least a portion of the plurality of terms to a plurality of buckets using an indexing engine in combination with a clustering framework configured to cluster the plurality of terms to categorization terms associated with each bucket;
calculating, for each term of the plurality of terms, a probability that the term is associated with the media object;
associating the probability to each term, correspondingly;
generating a vector of content-level attributes corresponding to the media object based on the associating;
storing the vector of content-level attributes in a database;
receiving via the query interface a query vector containing at least one of the content-level attributes;
searching the database for at least one content-level attribute contained in the query vector; and
providing, in response to the searching, a query result containing the media object.

2. The method according to claim 1 , further comprising:
generating the plurality of buckets by selecting categorization terms corresponding to a plurality of like-terms associated with a type of media content.

3. The method according to claim 1 , further comprising:
generating the plurality of buckets by clustering the plurality of terms using a plurality of cluster terms and a plurality of definitions including a plurality of references to a plurality of other terms, wherein the cluster terms having the highest scores correspond to the plurality of buckets.

4. The method according to claim 1 , wherein the metadata is extracted from at least one structured data source.

5. The method of claim 1 , further comprising:
for each bucket of the plurality of buckets:
correlating the plurality of terms associated with the bucket; and
weighting each bucket based on a correlation value obtained by the correlating.

6. The method according to claim 1 , further comprising:
generating a document containing at least the vector; and
storing the document in the database.

7. The method of claim 6 , further comprising:
generating a query vector having the at least one content-level attribute corresponding to at least one of the plurality of buckets; and
querying the database by using the query vector.

8. A system for providing a database of content-level attributes associated with media objects, the system comprising:
at least one computer of a content recommendation system, the at least one computer having a query interface for receiving a query containing at least one content-level attribute;
a server adapted to gather text data and operable to:
obtain, by crawling a network, metadata associated with a media object from a plurality of data sources, wherein at least one of the plurality of data sources is an unstructured data source and the metadata includes extrinsic metadata corresponding to one or more content-level attributes of the media object and

a database,
wherein the at least one computer is operable to:
extract from the metadata a plurality of terms associated with the media object by applying an entity extraction model to the metadata,
map at least a portion of the plurality of terms to a plurality of buckets using an indexing engine in combination with a clustering framework configured to cluster the plurality of terms to categorization terms associated with each bucket,
calculate, for each term of the plurality of terms, a probability that the term is associated with the media object;
associate the probability to each term, correspondingly;
generate vector of content-level attributes corresponding to the media object based on the associating;
store the vector of content-level attributes in the database;
receive via the query interface a query vector containing at least one of the content-level attributes;
search the database for at least one content-level attribute contained in the query vector; and
provide, in response to the searching, a query result containing the media object.

9. The system according to claim 8 , the at least one computer of the content recommendation system further operable to:
generate the plurality of buckets by selecting categorization terms corresponding to a plurality of like-terms associated with a type of media content.

10. The system according to claim 8 , the at least one computer of the content recommendation system further operable to:
generate the plurality of buckets by clustering the plurality of terms using a plurality of cluster terms and a plurality of definitions including a plurality of references to a plurality of other terms, wherein the cluster terms having the highest scores correspond to the plurality of buckets.

11. The system according to claim 8 , wherein the metadata is extracted from at least one structured data source.

12. The system according to claim 8 , the at least one computer of the content recommendation system being further operable to:
for each bucket of the plurality of buckets:
correlate the plurality of terms associated with the bucket; and
weight each bucket based on a correlation value obtained by the correlating.

13. The system according to claim 8 , the at least one computer of the content recommendation system being further operable to:
generate a document containing at least the vector; and
store the document in the database.

14. The system according to claim 13 , wherein the at least one computer of the content recommendation system is further operable to:
generate a query vector having the at least one content-level attribute corresponding to at least one of the plurality of buckets; and
query the database by using the query vector.

15. A non-transitory computer-readable medium having stored thereon one or more sequences of instructions for causing one or more processors to perform on at least one computer of a content recommendation system, the at least one computer having a query interface for receiving a query containing at least one content-level attribute, the steps of:
obtaining, by crawling a network using a server adapted to gather text data, metadata associated with a media object from a plurality of data sources, wherein at least one of the plurality of data sources is an unstructured data source and the metadata includes extrinsic metadata corresponding to one or more content-level attributes of the media object;
extracting from the metadata a plurality of terms associated with the media object by applying an entity extraction model to the metadata;
mapping at least a portion of the plurality of terms to a plurality of buckets using an indexing engine in combination with a clustering framework configured to cluster the plurality of terms to categorization terms associated with each bucket;
calculating, for each term of the plurality of terms, a probability that the term is

Assignees

• The Echo Nest Corp
• Spotify Ab

Inventors

• Whitman Brian
• Rodger David

Classifications

• G06F16/7867

  using information manually generated, e.g. tags, keywords, comments, title and artist information, manually generated time, location and usage information, user ratings · CPC title

• G06F17/3082

  Physics · mapped topic

• H04N21/44222

  Analytics of user selections, e.g. selection of programmes or purchase activity (monitoring of user selections in data processing systems G06F11/34; arrangements for monitoring the user's behaviour or opinions in broadcast systems H04H60/33) · CPC title

• H04N21/432

  Content retrieval operation from a local storage medium, e.g. hard-disk {(details of retrieval of video data and associated meta data in video databases G06F16/739)} · CPC title

• H04N21/435

  Processing of additional data, e.g. decrypting of additional data, reconstructing software from modules extracted from the transport stream · CPC title