Feature-Augmented Neural Networks and Applications of Same
First Claim
1. A method for mapping input information into output information using a neural network, comprising:
- receiving an input vector at an input layer of the neural network, the input vector representing at least part of the input information;
receiving a feature vector at the input layer of the neural network, the feature vector representing supplemental information pertaining to the input information; and
using the neural network to generate an output vector at an output layer of the neural network, based on the input vector and the feature vector,the output vector representing the output information.
3 Assignments
0 Petitions
Accused Products
Abstract
A system is described herein which uses a neural network having an input layer that accepts an input vector and a feature vector. The input vector represents at least part of input information, such as, but not limited to, a word or phrase in a sequence of input words. The feature vector provides supplemental information pertaining to the input information. The neural network produces an output vector based on the input vector and the feature vector. In one implementation, the neural network is a recurrent neural network. Also described herein are various applications of the system, including a machine translation application.
-
Citations
20 Claims
-
1. A method for mapping input information into output information using a neural network, comprising:
-
receiving an input vector at an input layer of the neural network, the input vector representing at least part of the input information; receiving a feature vector at the input layer of the neural network, the feature vector representing supplemental information pertaining to the input information; and using the neural network to generate an output vector at an output layer of the neural network, based on the input vector and the feature vector, the output vector representing the output information. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)
-
-
17. A recurrent neural network, comprising:
-
an input layer for receiving; an input vector, the input vector representing at least part of input information; a feature vector, the feature vector representing supplemental information pertaining to the input information; and a time-delayed hidden-state vector that represents an output from a hidden layer in a prior time instance; the hidden layer, the hidden layer for operating on; the input vector, as modified by a first matrix; the feature vector, as modified by a second matrix; the time-delayed hidden-state vector, as modified by a third matrix; and an output layer for providing an output vector by operating on; an output hidden-state vector provided by the hidden layer, as modified by a fourth matrix; and the feature vector, as modified by a fifth matrix, the output vector providing output information, as conditioned by the supplemental information. - View Dependent Claims (18, 19)
-
-
20. A system for processing linguistic information, comprising:
-
an input information-providing module configured to provide an input vector, the input vector representing at least one identified word in a sequence of words; a feature information-providing module configured to provide a feature vector, the feature vector describing one or more of; at least one aspect of a block of words in the sequence of words; and information that pertains to the sequence of words obtained from a source other than the sequence of words; and a recurrent neural network configured to generate an output vector based on the input vector and the feature vector.
-
Specification