# US 11,907,863 B2 · Natural language enrichment using action explanations > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,907,863 B2 - Document ID: US11907863B2 - Status: Granted - Kind: B2 - Application number: 16/937,704 - Filed: 2020-07-24 - Published: 2022-01-27 - Granted: 2024-02-20 - Earliest claimed: 2020-07-24 - Assignees: International Business Machines Corporation - Applicants: International Business Machines Corporation - Inventors: Daniela Rosu; Ruchi Mahindru - Web page: https://patentgrep.com/patent/US11907863B2 ## CPC - G06F 40/35: Discourse or dialogue representation - G06N 5/045: Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence - G06F 18/22: Matching criteria, e.g. proximity measures - G06F 40/216: using statistical methods - G06F 40/247: Thesauruses; Synonyms - G06F 40/30: Semantic analysis - G06F 40/44: Statistical methods, e.g. probability models - G06F 40/56: Natural language generation - G06N 3/006: based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO] - G06N 5/022: Knowledge engineering; Knowledge acquisition - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract Embodiments are provided that relate to a computer system, a computer program product, and a computer-implemented method for improving performance of a dialog system employing an automated virtual dialog agent. Embodiments involve utilizing an automated virtual agent to receive a natural language request and generate a corresponding response, automatically identifying and resolving a corresponding knowledge gap between the request and response, and refining the automated virtual agent with the resolved knowledge gap. ## Claims ### Claim 1 A computer system comprising: a processor operatively coupled to a memory; and an artificial intelligence (AI) platform in communication with the processor, the AI platform having tools to direct performance of a dialog system, the tools comprising: a dialog manager configured to receive natural language (NL) related to interaction with an automated virtual dialog agent, the NL comprising one or more dialog events, each of the dialog events including input and output, the input comprising one or more input instances and the output comprising one or more corresponding output instances or actions; an AI manager, operatively coupled to the dialog manager, configured to apply the received one or more dialog events to a learning program, the application including the learning program configured to: interpret the one or more input instances and the one or more corresponding output instances or actions, including represent the one or more dialog events in a model; identify a knowledge gap associated with one or more of the interpreted instances or actions; measure distance between subgraphs of the model with a common node label; leverage structure and functionality of one or more corresponding models to compare content and to determine any similarities or differences corresponding to concept relations; provide an ontology representation of any similarities or differences corresponding to concept relations; and dynamically generate one or more alternative knowledge items to bridge the identified knowledge gap; and a director configured to refine the automated virtual dialog agent in relation to the one or more dialog events, the refinement corresponding to the bridged knowledge gap to improve performance of the dialog system. ### Claim 2 (depends on 1) The computer system of claim 1, wherein the learning program is configured to analyze multiple pairs of input and output of the dialogue system, identify one or more features common among the multiple pairs, and leverage the identified one or more common features to solicit pieces of knowledge. ### Claim 3 (depends on 2) The computer system of claim 2, wherein the AI manager is configured to update a corresponding domain knowledge with an explanation and the solicited pieces of knowledge. ### Claim 4 (depends on 3) The computer system of claim 3, wherein the AI manager is configured to receive ground truth data to bootstrap the domain knowledge. ### Claim 5 (depends on 1) The computer system of claim 1, wherein the interpretation of the one or more input instances further comprises the AI manager to leverage the model to compare text of question answer pairs and determine any similarities and differences in terms of concept relations. ### Claim 6 (depends on 2) The computer system of claim 2, wherein the solicitation of pieces of knowledge further comprises the AI manager configured to generate one or more questions from the determined similarities and differences, present the generated one or more questions through the dialog system, and receive at least one answer to the generated one or more questions, the at least one answer functioning as an indicator to the identified knowledge gap. ### Claim 7 (depends on 5) The computer system of claim 5, wherein the measured distance corresponds to complexity of the identified knowledge gap. ### Claim 8 (depends on 1) The computer system of claim 1, wherein the identification of the knowledge gap includes the AI manager configured to determine when the one or more input instances comprises one or more concepts absent from the one or more corresponding output instances or actions. ### Claim 9 (depends on 1) The computer system of claim 1, wherein the AI manager is further configured to solicit selection of at least one of the one or more generated alternative knowledge items, the selection to bridge the identified knowledge gap. ### Claim 10 (depends on 1) The computer system of claim 1, wherein the refinement corresponding to the bridged knowledge gap to improve performance of the dialog system is conducted online with an end-user interacting with an AI solution. ### Claim 11 (depends on 1) The computer system of claim 1, wherein the model is an abstract meaning representation (AMR) tree or a parse tree. ### Claim 12 A computer program product to support active learning, the computer program product comprising: a computer readable storage medium having a program code embodied therewith, the program code executable by a processor to: receive natural language (NL) relating to interaction with an automated virtual dialog agent of a dialog system, the NL comprising one or more dialog events, each of the one or more dialog events including input and output, the input comprising one more input instances and the output comprising one or more corresponding output instances or output actions; apply the received dialogue events to an artificial intelligence (AI) learning program to: interpret the one or more input instances and the one or more corresponding output instances or output actions, including represent the one or more dialog events in a model identify a knowledge gap associated with one or more of the interpreted instances or actions; measure distance between subgraphs of the model with a common node label; leverage structure and functionality of one or more corresponding models to compare content and to determine any similarities or differences corresponding to concept relations; provide an ontology representation of any similarities or differences corresponding to concept relations; and dynamically generate one or more alternative knowledge items to bridge the knowledge gap; and refine the automated virtual dialog agent in relation to the dialog event corresponding to the bridged knowledge gap to improve performance of the dialog system. ### Claim 13 (depends on 12) The computer program product of claim 12, further comprising program code executable by the processor to analyze multiple pairs of input and output of the dialogue system, identify one or more features common among the multiple pairs, and leverage the identified one or more common features to solicit pieces of knowledge. ### Claim 14 (depends on 13) The computer program product of claim 13, further comprising program code executable by the processor to update a corresponding domain knowledge with an explanation and the solicited pieces of knowledge, and enrich a corresponding domain knowledge with the explanation and the solicited pieces of knowledge. ### Claim 15 (depends on 12) The computer program product of claim 12, wherein the program code to interpret the one or more input instances further comprises program code executable by the processor to leverage the model to compare text of question answer pairs and determine any similarities and differences in terms of concept relations. ### Claim 16 (depends on 13) The computer program product of claim 13, wherein the program code executable by the processor to solicit pieces of knowledge further comprises program code executable by the processor to generate one or more questions from the determined similarities and differences, present the generated one or more questions through the dialog system, and receive at least one answer to the generated one or more questions, the at least one answer functioning as an indicator to the identified knowledge gap. ### Claim 17 (depends on 15) The computer program product of claim 15, wherein the measured distance corresponds to complexity of the identified knowledge gap. ### Claim 18 A computer-implemented method directed at performance of a dialog system, the method comprising: receiving, by a processor of a computing device, natural language (NL) relating to interaction with an automated virtual dialog agent of a dialog system, the NL comprising one or more dialog events, each of the one or more dialog events including input and output, the input comprising one more input instances and the output comprising one or more corresponding output instances or output actions; applying, by the processor, the received one or more dialog events to an artificial intelligence (AI) platform comprising a learning program, the applying comprising: interpreting the one or more input instances and the one or more corresponding output instances or output actions, including representing the one or more dialog events in a model; identifying a knowledge gap associated with the interpreted one or more input instances and the one or more corresponding output instances or output actions; measuring distance between subgraphs of the model with a common node label; leveraging structure and functionality of one or more corresponding models to compare content and to determine any similarities or differences corresponding to concept relations; providing an ontology representation of any similarities or differences corresponding to concept relations; and dynamically generating one or more alternative knowledge items to bridge the knowledge gap; and refining the automated virtual dialog agent in relation to the one or more dialog events corresponding to the bridged knowledge gap to improve performance of the dialog system. ### Claim 19 (depends on 18) The computer-implemented method of claim 18, further comprising the learning program analyzing multiple pairs of input and output of the dialog system, identifying one or more features common among the multiple pairs, and leveraging the identified one or more common features to solicit pieces of knowledge. ### Claim 20 (depends on 19) The computer-implemented method of claim 19, further comprising updating a corresponding domain knowledge with an explanation and the solicited pieces of knowledge, and enriching a corresponding domain knowledge with the explanation and the solicited pieces of knowledge. ### Claim 21 (depends on 19) The computer-implemented method of claim 19, further comprising receiving ground truth data to bootstrap the domain knowledge. ### Claim 22 (depends on 18) The computer-implemented method of claim 18, wherein interpreting the one or more input instances further comprises leveraging the model to compare text of question answer pairs and determine any similarities and differences in terms of concept relations. ### Claim 23 (depends on 19) The computer-implemented method of claim 19, wherein soliciting pieces of knowledge further comprises generating one or more questions from the determined similarities and differences, presenting the generated one or more questions through the dialog system, and receiving at least one answer to the generated one or more questions, the at least one answer functioning as an indicator to the identified knowledge gap. ### Claim 24 (depends on 22) The computer-implemented method of claim 22, wherein the measured distance corresponds to complexity of the identified knowledge gap. ### Claim 25 A computer system comprising: an artificial intelligence (AI) platform in communication with a processor, the AI platform having tools to direct performance of a dialog system, the tools comprising: a dialog manager configured to receive natural language (NL) related to interaction with an automated virtual dialog agent of the dialog system, the NL comprising one or more dialog events, each of the dialog events including input and output, the input comprising one or more input instances and the output comprising one or more corresponding output instances; an AI manager, operatively coupled to the dialog manager, configured to apply the received one or more dialog events to a learning program, the learning program configured to: represent the one or more dialog events in a model, leverage the model to compare text of question answer pairs, and determine one or more similarities and differences in terms of concept relations, wherein the model represents the one or more dialog events; identify a knowledge gap between the one or more input instances and the one or more corresponding output instances; measure distance between one or more subgraphs of the model with a common node label, wherein the measured distance corresponds to complexity of the identified knowledge gap; leverage structure and functionality of one or more corresponding models to compare content and to determine any similarities or differences corresponding to concept relations; provide an ontology representation of any similarities or differences corresponding to concept relations; and dynamically generate one or more alternative knowledge items to bridge the knowledge gap; and a director configured to refine the automated virtual dialog agent, the refinement corresponding to the bridged knowledge gap.