Please download the dossier by clicking on the dossier button x
×

Method for estimating user interests

  • US 9,740,782 B2
  • Filed: 02/02/2015
  • Issued: 08/22/2017
  • Est. Priority Date: 09/12/2014
  • Status: Active Grant
First Claim
Patent Images

1. A computer-implemented method for estimating user interests, the method executable by a computing device in communication with an output device, the method comprising:

  • determining, by the computing device, a first input vector corresponding to a first user event and a second input vector corresponding to a separate second user event;

    configuring a first vector-mapping module to map given input vectors to output vectors in a first multidimensional space such that a respective distance separating the output vectors from each other is correlated to a difference between a user event context associated with the first input vector having been mapped to its output vector and a user event context associated with the second input vector having been mapped to its output vector, the configuring the first vector-mapping module including configuring a first neural network and a second neural network by;

    connecting the first neural network and the second neural network in a coupled Siamese neural network arrangement, andtraining the first and second neural networks to minimize a cross-modal loss between the first neural network and the second neural network;

    mapping, using the first vector-mapping module of the computing device, the first input vector to a first output vector in the first multidimensional space and the separate second input vector to a second output vector in the first multidimensional space,the first vector-mapping module being enabled to map both the first output vector and the second output vector to the first multidimensional space even if the first user event and the second user event are of different types;

    determining, by the computing device, a third input vector based at least in part on the first output vector and the separate second output vector;

    configuring a second vector-mapping module to map input vectors to output vectors in a second multidimensional space such that a respective distance separating the output vectors from each other is correlated to a difference between a user event context associated with one input vector, the one input vector being a respective output vector of the first vector-mapping module and a user event context associated with another input, the other input vector being another respective output vector of the first vector-mapping module, the configuring the second vector-mapping module including configuring a third neural network;

    mapping, using the second vector-mapping module of the computing device, the third input vector to a third output vector in the second multidimensional space;

    determining, by the computing device, a message to be provided to a user based at least in part on an analysis of at least one of the first output vector and the third output vector; and

    causing, by the computing device, the output device to provide the message to the user.

View all claims
  • 4 Assignments
Timeline View
Assignment View
    ×
    ×