# US 2021/0240503 A1 · System and method for interactively guiding users through a procedure > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2021/0240503 A1 - Document ID: US20210240503A1 - Status: Published application - Kind: A1 - Application number: 16/777,430 - Filed: 2020-01-30 - Published: 2021-08-05 - Earliest claimed: 2020-01-30 - Assignees: Accenture Global Solutions Limited - Applicants: Accenture Global Solutions Limited - Inventors: Anutosh Maitra; Shubhashis Sengupta; Ajay Chatterjee; Abhisek Mukhopadhyay; Shivam Garg - Web page: https://patentgrep.com/patent/US20210240503A1 ## CPC - G06F 16/90332: {Natural language query formulation or dialogue systems} - G06F 40/35: Discourse or dialogue representation - G06F 9/453: {Help systems} - G06F 40/30: Semantic analysis - G06N 3/044: Recurrent networks, e.g. Hopfield networks - G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] - G06N 3/09: Supervised learning - G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound - G06N 5/041: {Abduction} - G06N 20/00: Machine learning - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract An intelligent question and answer (Q&A) system and method for interactively guiding users through a procedure is disclosed. The intelligent Q&A system can dynamically generate process trees (or procedural trees) from the content or procedures presented in a raw document, such as a reference manual. The intelligent Q&A system can include a virtual agent that uses the dynamically generated process trees for interactive conversation with a user. Using the system, the virtual agent can interactively guide users through completing tasks such as updating software or connecting an IoT device to an existing system. ## Claims ### Claim 1 A method of dynamically generating process trees and using the dynamically generated process trees to converse with a user, wherein conversing with the user is accomplished by an artificially intelligent virtual agent, the method comprising: retrieving a document including text; automatically extracting, using a procedure classification model, a section of text in the document corresponding to a procedure; automatically identifying one or more decision points associated with the section of text; automatically identifying, using a clause identification model, a set of clauses associated with the one or more decision points, wherein each clause in the set of clauses comprises sequences of words from the section of text; automatically generating, using a question generation model, at least one question from the set of clauses; automatically generating a process tree from the at least one question and the set of clauses, wherein the process tree includes: one or more decision nodes corresponding to the at least one question; one or more response nodes; and one or more paths connecting the one or more decision nodes and the one or more response nodes; retrieving, using the virtual agent, the process tree; and conversing, using the virtual agent, with the user, wherein conversing with the user includes traversing through at least one path in the process tree. ### Claim 2 (depends on 1) The method according to claim 1, wherein the procedure classification model includes a long-short term memory recurrent neural network and wherein the method further includes training the long-short term memory recurrent neural network using a set of reference manuals. ### Claim 3 (depends on 1) The method according to claim 1, wherein identifying the one or more decision points includes applying a natural language processing parser to the section of text in the document corresponding to the procedure. ### Claim 4 (depends on 1) The method according to claim 1, wherein the clause identification model further includes a long-short term memory recurrent neural network, a first text parser and a second text parser that is different from the first text parser. ### Claim 5 (depends on 4) The method according to claim 4, wherein the long-short term memory recurrent neural network is a bi-directional network and includes a conditional random field layer. ### Claim 6 (depends on 4) The method according to claim 4, wherein the first text parser is used to identify parts-of-speech in sentences, wherein the long-short term memory recurrent neural network is used to generate sequence tags denoting clauses and starting points for the clauses, and wherein the second text parser is used to differentiate between independent clauses and dependent clauses in the clauses identified by the long-short term memory recurrent neural network. ### Claim 7 (depends on 1) The method according to claim 1, wherein the question generation model includes a recurrent neural network and wherein the method further includes training the recurrent neural network with a set of questions and answers. ### Claim 8 (depends on 7) The method according to claim 7, wherein each question in the set of questions and answers is associated with a binary set of answers. ### Claim 9 (depends on 1) The method according to claim 1, wherein at least one of the response nodes in the process tree indicates that the virtual agent should transfer the user to a human agent for further assistance. ### Claim 10 A system for dynamically generating process trees and using the dynamically generated process trees to converse with a user, the system comprising: a device processor; and a non-transitory computer readable medium storing instructions that are executable by the device processor to implement: a process tree generator that generates process trees, wherein the process tree generator further comprises: a procedure classification model that classifies sections of text as procedural text or non-procedural text; a decision points identifier that identifies decision points in procedural text; a clause identification model that uses information about decision points in procedural text to identify independent and dependent clauses in the procedural text; a question generation model that generates questions corresponding to independent clauses identified by the clause identification model; an answer path analyzer that constructs paths between questions generated by the question generation model and dependent clauses identified by the clause identification model; a virtual agent that retrieves process trees generated by the process tree generator and uses the process trees to converse with users. ### Claim 11 (depends on 10) The system according to claim 10, wherein the procedure classification model includes a long-short term memory recurrent neural network. ### Claim 12 (depends on 10) The system according to claim 10, wherein the decision points identifier includes a natural language processing parser. ### Claim 13 (depends on 10) The system according to claim 10, wherein the clause identification model further includes a long-short term memory recurrent neural network, a first text parser and a second text parser that is different from the first text parser. ### Claim 14 (depends on 13) The system according to claim 13, wherein the long-short term memory recurrent neural network is a bi-directional network and includes a conditional random field layer. ### Claim 15 (depends on 10) The system according to claim 10, wherein the question generation model includes a recurrent neural network. ### Claim 16 (depends on 10) The system according to claim 10, wherein the virtual agent is a chat-bot. ### Claim 17 A non-transitory computer-readable medium storing software comprising instructions executable by one or more computers which, upon such execution, cause the one or more computers to: retrieve a document including text; automatically extract a section of text in the document corresponding to a procedure; automatically identify one or more decision points associated with the section of text; automatically identify a set of clauses associated with the one or more decision points, wherein each clause in the set of clauses comprise sequences of words from the section of text; automatically generate at least one question from the set of clauses; automatically generate a process tree from the at least one question and the set of clauses. ### Claim 18 (depends on 17) The non-transitory computer-readable medium according to claim 17, wherein the process tree further includes: one or more decision nodes corresponding to the at least one question; one or more response nodes; and one or more paths connecting the one or more decision nodes and the one or more response nodes. ### Claim 19 (depends on 17) The non-transitory computer-readable medium according to claim 17, wherein the document is a reference manual. ### Claim 20 (depends on 17) The non-transitory computer-readable medium according to claim 17, wherein causing the one or more computers to automatically identify a set of clauses includes executing a pre-trained neural network.