Shengyin Gu

US Patent:

8560549, Oct 15, 2013

Filed:

May 14, 2012

Appl. No.:

13/471391

Inventors:

Elin R. Pedersen - Portola Valley CA, US
Karl A. Gyllstrom - Chapel Hill NC, US
Shengyin Gu - Sunnyvale CA, US
Peter Jin Hong - San Francisco CA, US

Assignee:

Google Inc. - Mountain View CA

International Classification:

G06F 17/30

US Classification:

707739

Abstract:

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating action trails from web history are described. In one aspect, a method includes receiving a web content access history of a user, the content access history including one or more user actions, each user action being associated with a content item upon which the user action is performed and identifying one or more action trails from the content access history, each action trail including a sequence of user actions performed one content items relating to a topic. Identifying a particular action trail includes clustering the user actions into a series of segments using temporal criteria; calculating semantic similarities between the content items, and adding a segment of the series of segments to the action trail when the semantic similarities between the segment and another segment satisfy a similarity threshold.

US Patent:

8180778, May 15, 2012

Filed:

Feb 5, 2010

Appl. No.:

12/701444

Inventors:

Elin Pedersen - Portola Valley CA, US
Karl A. Gyllstrom - Chapel Hill NC, US
Shengyin Gu - Sunnyvale CA, US
Peter Jin Hong - San Francisco CA, US

Assignee:

Google Inc. - Mountain View CA

International Classification:

G06F 17/30

US Classification:

707739

Abstract:

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating action trails from web history are described. In one aspect, a method includes receiving a web content access history of a user, the content access history including one or more user actions, each user action being associated with a content item upon which the user action is performed and identifying one or more action trails from the content access history, each action trail including a sequence of user actions performed one content items relating to a topic. Identifying a particular action trail includes clustering the user actions into a series of segments using temporal criteria; calculating semantic similarities between the content items, and adding a segment of the series of segments to the action trail when the semantic similarities between the segment and another segment satisfy a similarity threshold.