Neural network training utilizing loss functions reflecting neighbor token dependencies

Abstract

Systems and methods for neural network training utilizing loss functions reflecting neighbor token dependencies. An example method comprises: receiving a training dataset comprising a plurality of labeled tokens; determining, by a neural network, a first tag associated with a current token processed by the neural network, a second tag associated with a previous token which has been processed by the neural network before processing the current token, and a third tag associated with a next token to be processed by the neural network after processing the current token; computing, for the training dataset, a value of a loss function reflecting a first loss value, a second loss value, and a third loss value, wherein the first loss value is represented by a first difference of the first tag and a first label associated with the current token by the training dataset, wherein the second loss value is represented by a second difference of the second tag and a second label associated with the previous token by the training dataset, and wherein the third loss value is represented by a third difference of the third tag and a third label associated with the next token by the training dataset; and adjusting a parameter of the neural network based on the value of the loss function.

What is claimed is: 
     
         1 . A method, comprising:
 receiving, by a processing device, a training dataset comprising a sequence of labeled tokens comprising a first token, a second token, and a third token, wherein the second token follows the first token, and the third token follows the second token;   determining, by a neural network, a first tag corresponding to the first token, a second tag corresponding to the second token, and a third tag corresponding to the third token;   computing a value of a loss function reflecting a first loss value, a second loss value, and a third loss value, wherein the first loss value is represented by a first difference of the first tag and a first label associated with the first token by the training dataset, wherein the second loss value is represented by a second difference of the second tag and a second label associated with the second token by the training dataset, and wherein the third loss value is represented by a third difference of the third tag and a third label associated with the third token by the training dataset; and   adjusting a parameter of the neural network based on the value of the loss function.   
     
     
         2 . The method of  claim 1 , further comprising:
 performing, using the neural network, a natural language processing task.   
     
     
         3 . The method of  claim 1 , further comprising:
 repeating the determining, computing, and adjusted operations until the value of the loss function falls below a predetermined threshold.   
     
     
         4 . The method of  claim 1 , wherein the first token is represented by a natural language word, and wherein the first tag identifies a part of speech associated with the first token. 
     
     
         5 . The method of  claim 1 , wherein the first token is represented by a natural language word, and wherein the first tag identifies one or more grammatical attributes associated with the first token. 
     
     
         6 . The method of  claim 1 , wherein the value of the loss function is represented by a linear combination of the first loss value, the second loss value, and the third loss value. 
     
     
         7 . The method of  claim 1 , wherein adjusting the parameter of the neural network further comprises:
 back-propagating an error reflected by the value of the loss function to one or more previous layers of the neural network; and   adjusting an edge weight in order to minimize the loss function.   
     
     
         8 . The method of  claim 1 , wherein the neural network comprises a feature extraction layer, a bi-directional long-short term memory (BiLSTM) layer, and a prediction layer, and wherein the BiLSTM layer further comprises a BiLSTM, a backward LSTM and a forward LSTM, such that a first output of the backward LSTM and a second output of the forward LSTM is fed to the BiLSTM. 
     
     
         9 . A system, comprising:
 a memory; and   a processing device coupled to the memory, the processing device configured to:
 receive a training dataset comprising a sequence of labeled tokens comprising a first token, a second token, and a third token, wherein the second token follows the first token, and the third token follows the second token; 
 determine, by a neural network, a first tag corresponding to the first token, a second tag corresponding to the second token, and a third tag corresponding to the third token; 
 compute a value of a loss function reflecting a first loss value, a second loss value, and a third loss value, wherein the first loss value is represented by a first difference of the first tag and a first label associated with the first token by the training dataset, wherein the second loss value is represented by a second difference of the second tag and a second label associated with the second token by the training dataset, and wherein the third loss value is represented by a third difference of the third tag and a third label associated with the third token by the training dataset; and 
 adjust a parameter of the neural network based on the value of the loss function. 
   
     
     
         10 . The system of  claim 9 , wherein the processing device is further configured to:
 perform, using the neural network, a natural language processing task.   
     
     
         11 . The system of  claim 9 , wherein the processing device is further configured to:
 repeat the determining, computing, and adjusted operations until the value of the loss function falls below a predetermined threshold.   
     
     
         12 . The system of  claim 9 , wherein the first token is represented by a natural language word, and wherein the first tag identifies a part of speech associated with the first token. 
     
     
         13 . The system of  claim 9 , wherein the first token is represented by a natural language word, and wherein the first tag identifies one or more grammatical attributes associated with the first token. 
     
     
         14 . The system of  claim 9 , wherein the value of the loss function is represented by a linear combination of the first loss value, the second loss value, and the third loss value. 
     
     
         15 . The system of  claim 9 , wherein adjusting the parameter of the neural network further comprises:
 back-propagating an error reflected by the value of the loss function to one or more previous layers of the neural network; and   adjusting an edge weight in order to minimize the loss function.   
     
     
         16 . A computer-readable non-transitory storage medium comprising executable instructions that, when executed by a computer system, cause the computer system to:
 receive a training dataset comprising a sequence of labeled tokens comprising a first token, a second token, and a third token, wherein the second token follows the first token, and the third token follows the second token;   determine, by a neural network, a first tag corresponding to the first token, a second tag corresponding to the second token, and a third tag corresponding to the third token;   compute a value of a loss function reflecting a first loss value, a second loss value, and a third loss value, wherein the first loss value is represented by a first difference of the first tag and a first label associated with the first token by the training dataset, wherein the second loss value is represented by a second difference of the second tag and a second label associated with the second token by the training dataset, and wherein the third loss value is represented by a third difference of the third tag and a third label associated with the third token by the training dataset; and   adjust a parameter of the neural network based on the value of the loss function.   
     
     
         17 . The computer-readable non-transitory storage medium of  claim 16 , wherein the first token is represented by a natural language word, and wherein the first tag identifies a part of speech associated with the first token. 
     
     
         18 . The computer-readable non-transitory storage medium of  claim 16 , wherein the first token is represented by a natural language word, and wherein the first tag identifies one or more grammatical attributes associated with the first token. 
     
     
         19 . The computer-readable non-transitory storage medium of  claim 16 , wherein the value of the loss function is represented by a linear combination of the first loss value, the second loss value, and the third loss value. 
     
     
         20 . The computer-readable non-transitory storage medium of  claim 16 , wherein adjusting the parameter of the neural network further comprises:
 back-propagating an error reflected by the value of the loss function to one or more previous layers of the neural network; and   adjusting an edge weight in order to minimize the loss function.