System and method for implementing an artificially intelligent virtual assistant using machine learning

Abstract

Systems and methods for implementing an artificially intelligent virtual assistant includes collecting a user query; using a competency classification machine learning model to generate a competency label for the user query; using a slot identification machine learning model to segment the text of the query and label each of the slots of the query; generating a slot value for each of the slots of the query; generating a handler for each of the slot values; and using the slot values to: identify an external data source relevant to the user query, fetch user data from the external data source, and apply one or more operations to the query to generate response data; and using the response data, to generate a response to the user query.

What is claimed: 
     
         1 . A machine learning-based conversational system comprising:
 (I) a competency engine that is implemented by one or more computers, the competency engine comprising one or more trained competency machine learning algorithms that:
 outputs a prediction based on text string input, the prediction comprising (a-i) an intent classification label for the text string input and (a-ii) a confidence value for the intent classification label that indicates a probability that the intent classification label matches an intent of the text input, wherein the text string input comprises a plurality of distinct tokens defining an utterance; 
   (II) a slot engine that is implemented by the one or more computers, the slot engine comprising:
 (b-i) a slot identifier that enumerates one or more slots within the text string input, each of the one or more slots comprising a distinct combination of one or more tokens of the text string input; 
 (b-ii) one or more trained slot labeling machine learning algorithms that predict a slot label for each of the enumerated one or more slots within the text string input based on the intent classification label and the text string input; 
   (III) a slot extractor that is implemented by the one or more computers, wherein the slot extractor that:
 (c-i) extracts each distinct pairing of each enumerated slot of one or more tokens and the predicted slot label for the enumerated slot of one or more tokens of the text string input; 
 (c-ii) converts each distinct pairing of each enumerated slot of one or more tokens and the predicted slot label for the enumerated slot, respectively, to a distinct conversion operation; 
 (c-iii) builds one or more distinct pieces of response data based on an execution of each distinct conversion operation; 
   (IV) a response data generator that generates a response to the text string input by:
 (d-i) selecting one response template of a plurality of distinct response templates based on the intent classification label and the predicted slot label for each of the enumerated one or more slots within the text string input, wherein the selected one response template comprises predetermined output text and one or more input slots for response data; 
 (d-ii) interleaving the one or more distinct pieces of response data into the one or more input slots of the selected one response template. 
   
     
     
         2 . The system according to  claim 1 , wherein the one or more trained slot labeling machine learning algorithms comprises one or more recurrent neural networks. 
     
     
         3 . The system according to  claim 1 , wherein
 the slot extractor uses the intent classification label for the text string input to restrict one or more computer operations that can be used for building the one or more distinct pieces of response data.   
     
     
         4 . The system according to  claim 1 , wherein
 the one or more trained competency machine learning algorithms is trained using crowdsourced training data samples.   
     
     
         5 . The system according to  claim 1 , wherein
 the competency engine identifies whether the confidence value for the intent classification label satisfies a competency threshold, wherein the competency threshold relates to a minimum level of confidence,   when the confidence value for the intent classification label satisfies the competency threshold, the competency engine outputs the intent classification label.   
     
     
         6 . The system according to  claim 1 , wherein
 the slot engine enumerates each of a plurality of slots within the text string input in the order that each of the plurality of slots appear within the text string input.   
     
     
         7 . The system according to  claim 1 , wherein:
 the slot extraction engine accesses a reference table that includes a plurality of distinct slot labels mapped to one or more distinct conversion operations; and   the conversion of each distinct pairing of each enumerated slot of one or more tokens and the predicted slot label for the enumerated slot, respectively, to the distinct conversion operation includes:
 comparing each distinct pairing to the reference table; and 
 identifying a respective conversion operation for each distinct pairing based on the comparison. 
   
     
     
         8 . The system according to claim  7 , wherein the one or more conversion operations include one or more of:
 applying a filter operation against a dataset;   converting a given slot of the text string input to a date or a date range; and   applying an arithmetic function against a distinct dataset.   
     
     
         9 . The method according to  claim 8 , wherein
 building the one or more distinct pieces of response data includes executing one or more of:
 the applying the filter operation against a dataset; 
 converting the given slot of the text string input to the date or the data range; and 
 applying the arithmetic function against the distinct dataset; and 
   returning results of the executing as the one or more distinct pieces of response data.   
     
     
         10 . The system according to  claim 1 , further comprising:
 an automatic speech recognition system that:
 collects and processes utterance input, and 
 simultaneously transmits a copy of the text string input to each of the competency engine and the slot engine. 
   
     
     
         11 . The system according to  claim 1 , wherein
 each of the one or more trained competency machine learning algorithms comprises a plurality of distinct weights and associated features that are selectively activated or selectively deactivated based on the text string input.