System and method for model derivation for entity prediction

Abstract

A method for deriving a model for a chatbot for predicting entities in a sentence. The sentence is input into a named-entity recognition module and features obtained. A LSTM RNN forward pass and backward pass is performed on the features to obtain a first and second set of results, respectively. A first concatenating is performed on the first set of results and the second set of results. A second concatenation is performed on the first concatenation using output target entities. A connected set of neurons from the second concatenation is obtained. An output is obtained, and a prediction is collected on a next output by summing the outputs previous to that output. The prediction is input into the performing of the second concatenation step, wherein the method is performed cyclically until all outputs have been processed with input predictions.

1 . A method for deriving a model for a chatbot for predicting entities in a given sentence comprising a plurality of words wherein the words are separated by space, the method comprising the steps of:
 inputting the sentence into a named-entity recognition module and truncating the sentence;   obtaining features from the truncated sentence;   performing a Long Short-Term Memory Recurring Neural Network forward pass on the features to obtain a first set of results;   performing a Long Short-Term Memory Recurring Neural Network backward pass on the features to obtain a second set of results;   performing a first concatenating on the first set of results and the second set of results;   performing a second concatenation on the first concatenation using output target entities, wherein the output target entities are shifted by one step;   obtaining a fully connected set of neurons from the second concatenation, which are shared across all encoded words;   obtaining an output;   collecting a prediction on a next output by summing the plurality of outputs previous to that output; and   inputting the prediction into the performing of the second concatenation step, wherein the method is performed cyclically until all outputs have been processed with input predictions.   
     
     
         2 . The method of  claim 1 , wherein the Long Short-Term Memory Recurring Neural Network comprises 100 memory cells. 
     
     
         3 . The method of  claim 1 , wherein the first concatenation results in 200 dimensions. 
     
     
         4 . The method of  claim 1 , wherein the second concatenation results in a number of dimensions proportionally related to a number of features from the truncated sentence. 
     
     
         5 . The method of  claim 1 , wherein the features comprise one or more of: word embeddings, parts-of-speech encoding, special character information, digit count information, digit chunks count information, and fixed-size ordinally forgetting encoding. 
     
     
         6 . A system for deriving a model for a chatbot for predicting entities in a given sentence comprising a plurality of words wherein the words are separated by space, the system comprising:
 a processor; and   a memory in communication with the processor, the memory storing instructions that, when executed by the processor, causes the processor to derive the model by:
 inputting the sentence into a named-entity recognition module and truncating the sentence; 
 obtaining features from the truncated sentence; 
 performing a Long Short-Term Memory Recurring Neural Network forward pass on the features to obtain a first set of results; 
 performing a Long Short-Term Memory Recurring Neural Network backward pass on the features to obtain a second set of results; 
 performing a first concatenating on the first set of results and the second set of results; 
 performing a second concatenation on the first concatenation using output target entities, wherein the output target entities are shifted by one step; 
 obtaining a fully connected set of neurons from the second concatenation, which are shared across all encoded words; 
 obtaining an output; 
 collecting a prediction on a next output by summing the plurality of outputs previous to that output; and 
 inputting the prediction into the performing of the second concatenation step, wherein the method is performed cyclically until all outputs have been processed with input predictions. 
   
     
     
         7 . The system of  claim 6 , wherein the Long Short-Term Memory Recurring Neural Network comprises 100 memory cells. 
     
     
         8 . The system of  claim 6 , wherein the first concatenation results in 200 dimensions. 
     
     
         9 . The system of  claim 6 , wherein the second concatenation results in a number of dimensions proportionally related to a number of features from the truncated sentence. 
     
     
         10 . The system of  claim 6 , wherein the features comprise one or more of: word embeddings, parts-of-speech encoding, special character information, digit count information, digit chunks count information, and fixed-size ordinally forgetting encoding.