NATURAL LANGUAGE-BASED PROCESS FLOW SYSTEM

Abstract

This invention relates to a system (1) that enables the user to determine the action he wants to perform in the mobile application from a single sentence request given in writing or voice and to create a screen flow by sequencing the application screens to complete the action.

Description

1 TARIFF NATURAL LANGUAGE-BASED PROCESS FLOW SYSTEM Technical Area This invention allows the user to record the action they want to perform in a mobile application in written or spoken form. to be determined from the single-sentence request given and for the process to be completed a system that allows application screens to be arranged in order to create a screen flow It is related to the system. Previous Technique Today, mobile applications allow users to perform certain actions. Menus and tabs, in-app search, static shortcuts enable this. and access to the relevant transaction screens is provided using digital assistants. (Very 15) In applications with multiple screens and processes, users need to find the correct screens. He / She must enter the necessary information and complete the verification and approval steps himself / herself. It is necessary. Current systems can direct the user to a specific screen, However, the processing capabilities of the screens, switching between screens, and verification... By evaluating the requirements and changes in the application version, process 20 It does not automatically generate the steps. Therefore, the processes may be incomplete or faulty execution, users being directed to the wrong screens, and processes Deficiencies arise in the form of being left unfinished. Therefore, considering the studies and shortcomings in the current technology, 25 When this is the case, the written action that the user wants to take in the mobile application or determined by a single-sentence request given verbally, the process The processing characteristics of the application screens required for implementation, based on screen transitions and security / verification requirements evaluation and appropriate screen flow for completing the process 30 It appears that a system is needed to enable its creation. 2 United States Regulation US2024119931A1, which is included in the known state of the art. The patent document states that the electronic device makes different changes according to the user's voice commands. an electronic device and control method that performs operations in applications It is mentioned that in this invention, each application's memory is stored in the device's memory. User interfaces are nodes, and the operations that can be performed on these interfaces are 5 User interface graphics, shown as edges, are stored. The graphics, interface metadata and interface elements retrieved from the operating system It is created using the processes that can be performed. User's voice By processing the natural language understanding, the user command, the parameter of the command, and The target application is determined. Multiple applications capable of processing the command are selected. If found, the target application takes into account the frequency of application usage. It can be selected by taking the target. The target is determined according to the user command and parameter. The target node in the application's user interface graphic where the operation will be performed. is determined. The interface displayed on the screen is the current node. is being defined; if the target application is not open, the application is run and the current node 15 Connections between the current node and the target node are being detected. a sequence of actions to reach the goal by examining it, especially via the shortest path is created. Actions are executed sequentially at intermediate nodes and the target When the node is reached, the action corresponding to the user command is determined. This is done using parameters. If an action cannot be performed, a message of 20 is sent to the target node. A new route to reach the destination is being calculated; new interfaces will be introduced when the application is updated. and the processes are added to the user interface graphic. Brief Description of the Invention The purpose of this invention is to enable users to input text or audio messages in multi-screen mobile applications. The single-sentence request provided included the target transaction type, transaction parameters, and missing information. parameters determine transaction priority and security level; application Screen ID, accepted parameters, required user inputs for each screen, conditions for switching between screens, user authorization, authentication requirements, 30 User consent and application version information will be evaluated and the process will be completed. 3 A system that arranges the necessary screens in order for completion. to accomplish. Detailed Description of the Invention The "Natural Language-Based Process Flow" implemented to achieve the purpose of this invention. The "System" is shown in the attached figure; Figure 1. Schematic view of the system described in the invention. The parts shown in the figure are individually numbered, and the corresponding numbers correspond to these numbers. given below: 1. System 2. Application 15 3. Database 4. Server The user can specify the action they want to perform in the mobile application in written or spoken form. It is determined from the single sentence request given and 20 for the process to be completed. To create a screen flow by arranging application screens in a sequence. The system in question, developed for the purpose of invention (1); - the user's smartphone, tablet computer, desktop computer or portable an electronic device in the form of a computer that automatically processes data. 25 Receiving the user's single-sentence transaction request, either in writing or verbally. configured to display the generated screen stream sequentially. at least one application (2), - target transaction type for user request received via application (2), existing and missing parameters, operation priority and security level information, application 30 Screen ID, processing capacity, and past transaction flows for each screen. 4 to store completion, abandonment, error, and verification result data at least one database structured (3) and - Target operation from the single sentence user request received via application (2) type, existing and missing parameters, transaction priority, security level, To determine the uncertainty score; this information is recorded in the database (3) screen capacity 5 Creating candidate screen sequences by matching profiles and screen transition information; Candidates are analyzed based on transaction scope, transit cost, verification, and user authorization. by selecting the appropriate screen sequence and adding the selected sequence to the executable process flow. by converting it to enable it to be executed through the application (2) It contains at least one server (4). 10 The application (2) in the system (1) that is the subject of the invention, the user makes a payment, package purchase, change tariff, create support request, view information, subscription The only thing related to the action you want to perform, such as cancellation or address update 15 sentences of written input or audio input obtained by converting speech-to-text. obtaining the generated text expression, the screens specified in the created workflow. to open them sequentially, to transfer the specified parameters between screens, Showing information entries regarding missing parameters, verification and user approval. execute the steps and process flow completion, error, correction and abandonment at least one interface that enables receiving user interactions regarding their status 20 It is structured to present. The database (3) in the system (1) which is the subject of the invention, is obtained through the application (2). Target transaction type, current and missing for the user's one-sentence transaction request. Parameters, transaction priority, security level, and uncertainty score information; screen 25 ID, transaction step performed, accepted parameters, mandatory user inputs, validation and user consent requirements, deep linking or navigation addresses, screen transition conditions, transaction completion functions, security sensitivity levels, and application version information for each Screen 30 defines the processing capacity of the screen and the conditions for switching to other screens. Capacity profiles with screen nodes, and directional edges for screen transitions. It is configured to keep a record of the screen graph data being stored. Database (3), transition cost of transitions in the screen graph, user authorization level, validation requirement, deep linking, or in-app navigation Storing information and version validity; completion or abandonment of transaction flows Status, abandoned screen, error type, verification result, user correction, missing 5 parameter request count, stream execution time, screen transition delay, and history to record success rate data and to have this data sent to the server (4) candidate In creating, evaluating, and determining the workflow of screen sequences It is structured to enable its use. The server (4) in the system (1) which is the subject of the invention, any remote communication to communicate with the application (2) and the database (3) using the protocol and In order to exchange data with the application (3) through this established communication It is being configured. Server (4), bill payment, purchase of additional packages, tariff Change, commitment renewal, line operation, installment viewing, digital service 15 activation, opening a service request, creating a support request, adjusting card limits, money transfer, loan application, investment product viewing, return initiation, delivery changing address, using coupons, tracking applications, requesting documents, or scheduling appointments. interpreting user requests in the form of creation and parameters During the identification process, domain-specific trained transducer-based natural language understanding 20 model, intention-domain collaborative learning model, semantic similarity model, or rule model using a supported hybrid parser; converter-based for intent inference. Classifiers, Bidirectional Converter Encoder Representations Encoder Representations from Transformers (BERT) derivatives, multilingual language models, lightweight mobile natural language processing models, or intent-domain collaborative learning 25 named entity recognition for leveraging models and parameter extraction, array labeling, conditional random field-supported labeling, or the Big Language Model (Large Language Model – LLM) based structured output generation methods by using it, the operation and parameter information in the user statement can be processed. It is configured to enable the bringing in. The server (4) has 30 in the database (3). Registered screen ID, screen action step, accepted 6 parameters, required user inputs, validation, and user approval. requirements, transaction completion conditions, deep link or navigation information and by using the data from other screens that can be switched between, each screen screen capacity profiles that define the processing steps it can perform to evaluate; nodes with processing capacity and (2) screens in the application and 5 directional models of allowed transitions between screens as directional edges Using the screen graphic and defining transition conditions, transition costs, and user authorization for edges. level, validation requirement, deep link validation, and version compatibility to match the information with the required steps in the process diagram The server (4) is configured with each mandatory step in the process diagram. To identify the screens that can perform this action, use the screens located on the screen graph. nodes, semantic compatibility with the processing capacity and processing steps of the screens Evaluating with a semantic similarity score, which indicates screen-step matching. learned sorting model, gradient-increasing trees, graph neural network-based node 15 proximity search methods based on suitability models or embedding vectors. Using this, the candidate screen that shows the highest compatibility with the relevant process step. Identifying nodes, valid directional edges for the identified candidate screen nodes Creating multiple candidate screen sequences by sorting them and the candidate in question The required steps in the sequence processing diagram from start to finish It is configured to determine the capabilities to meet these requirements. Server 20 (4), each candidate screen sequence created, is required as defined in the process diagram. Transaction coverage score, screens showing the extent to which the transaction steps have been met. the technical and user interaction load that occurs during the transition between them The toll fee shown is a fee not requested by the user, but is included in the transaction. 25 missing pieces of information that need to be obtained later to complete it parameter cost, number of screens, repeated data entry, user confirmation steps and user feedback showing the interaction load resulting from waiting times. Friction score, parameters required for the process that the screens accept Compliance with parameters, status of fulfillment of verification requirements, similar 30 successful completion of the same or similar screen sequences in user requests past success rate showing frequency and screens in the current application version 7 Evaluating its validity based on the information provided; screen transition cost, user friction cost, missing parameter completion cost, verification delay, version incompatibility risk, and security risk all contribute to the total execution cost. within it, calculation and creation of candidate screen sequences and their ranking Constrained shortest path, A-star (A*) search, bundle search, multi-target optimization, 5 policy updating supported by reinforcement learning or contextual multi-faceted approach to leverage rogue-based stream sequencing algorithms is configured. The server (4) is used during the evaluation of candidate screen sequences. User authorization level, transaction sensitivity class, authentication requirement, open User consent rules, allowed deep links, and application version 10. Implementing validity information as mandatory flow constraints; authentication or a process step that does not include user approval, user authorization level switching to screens outside of authorized channels, using unauthorized deep links, or Candidate containing screens or transitions that are not valid in the current application version (2). Screening sequences before presenting them to the user; security and verification 15 deterministic policies and decisions independent of probabilistic model output to do this with restriction control and only for security, verification, user approval, meeting all authorization and minimum transaction coverage requirements among the candidates, those with low total execution cost and high transaction coverage score. The server (4) is configured to select the screen sequence that the user can choose from. The candidate selected to complete the target transaction type determined from the sentence request. the sequence of screens that will open sequentially during the execution of the target operation application (2) screens, process parameters to be transferred between screens, not included in the user's initial request, but necessary for the completion of the target operation. Required time, quantity, service name, product name, account or line information, and transaction 25 Missing parameters in the form of conditions, authentication and explicit user approval. screens that require navigation or deep link commands, errors may occur the screen to return to in case of failure, the condition indicating that the process is complete, and the process An executable containing the screen or process step to be continued when interrupted. It is configured to convert into a compiled workflow. Server (4), 30 the uncertainty score obtained from the user's request is above the specified threshold value 8 If so, the target transaction type can be clarified via application (2) the explanatory question or data entry field to be presented to the user, the target process any of the parameters required for implementation are requested by the user If it is not found, the missing parameter must be entered by the user. To provide a question or data entry field, add it to the compiled workflow and user 5 authorization, authentication, explicit user consent, valid screen or deep link, and The inability to create a screen sequence that meets the minimum transaction coverage requirements. valid application in which the target operation can be safely initiated (2) A question that will redirect to the screen or allow the user to provide missing information. or is configured to create a data entry field. The server (4) creates 10 to transmit the compiled workflow to be executed on the application (2); application Adding or removing screens in this version, deep link schemes, screens If the parameters or navigation rules change, the screen graph and Updating screen capacity profiles, outdated screens and transitions. deactivate and re-create the workflows based on the previous version 15 compile; completion or abandonment of the stream obtained through the application (2) status, abandoned screen, error type, validation result, user correction, missing parameter request count, stream execution time, screen transition delay, and history Using success rate data, intent matching scores, transitions on the screen graph Updating costs and flow selection policy, screen 20 successfully completed. to increase the preference value of arrays and to avoid constantly abandoned, faulty, or by increasing the cost of subsequent passes that generate validation failure Transition costs and historical success rates have been updated in user requests. It is structured to ensure that it is evaluated accordingly. Industrial Application of the Invention Thanks to the system (1) which is the subject of the invention, telecommunications, banking, e-commerce, bill payment in insurance, public services and similar multi-screen digital applications, Subscription transactions, package or tariff changes, payments and money transfers, applications, 30 Text or audio guides for multi-step user processes such as orders, returns, and support requests. 9 Identifying voice-based requests, obtaining missing user information, and taking necessary actions. by including confirmation or verification screens in the process flow of transactions completion is ensured. Around these basic concepts, the invention is “Natural Language Based Process Flow System (1)” 5 It is possible to develop a wide variety of applications related to this, and the invention is presented here. It cannot be limited to the examples given; it is essentially as stated in the claims.

Claims

REQUESTS 1. The user's desired action in the mobile application, in written or spoken form. the determination and process based on the single-sentence request given. To complete the process, the application screens must be sequenced to create a screen flow of 5. enabling its creation; - the user's smartphone, tablet computer, desktop computer or run on an electronic device in the form of a portable computer, an application that automatically processes data and produces meaningful results and / or execute software, provided the user gives written or verbal instructions. Receiving a single-sentence transaction request, and then sequentially following the generated screen flow. at least one application configured to enable display (2), - Target transaction type for user request received via application (2), Information on existing and missing parameters, transaction priority, and security level. screen ID, processing capacity, and past operation information for application screens. 15 Completion, abandonment, error, and verification result data related to the streams. containing at least one database (3) configured to store and - target from the single sentence user request received via application (2) transaction type, existing and missing parameters, transaction priority, security to determine the level, uncertainty score; this information in the database (3) 20 by matching registered screen capacity profiles and screen transition information, the candidate Creating screen sequences; evaluating candidates based on process coverage, transition cost, Verification analyzes user authorization to select the appropriate screen sequence. select and convert the selected sequence into an executable process flow for implementation. (2) At least one server configured to enable execution via 25 a payment system characterized by (4) (1).

2. The user can make payments, purchase packages, change tariffs, and request support. creating, viewing information, canceling subscriptions, or updating addresses. The single-sentence written input describing the operation you wish to perform is as follows: 30 or text obtained by converting audio input to speech-to-text 11 obtaining the expression, the screens specified in the created workflow sequentially. to open, to move between screens, to transfer specified parameters, missing to display data inputs related to parameters, validation and user execute the approval steps and complete the process flow, identify errors, and correct them. and at least one of the following 5 that enables us to collect user interactions regarding abandonment situations. characterized by the application (2) which is configured to provide an interface A system like the one in claim 1 (1).

3. The user's one-sentence transaction request received via Application (2) Target transaction type, existing and missing parameters, transaction priority, security 10 information on the level and uncertainty score; screen ID, transaction performed. steps, accepted parameters, required user inputs, validation and User approval requirements, deep linking or navigation addresses, screen transition conditions, transaction completion functions, security sensitivity levels and application version information; each screen's operation 15 Screen capacity, which defines the capacity and transition conditions to other screens. Profiles with screen nodes, directional edges for screen transitions Data structured to keep a record of the screen graph data being stored. any of the above claims characterized by base (3) such a system (1). 20 4. The transition cost and user authorization level for each transition shown in the on-screen graph. verification requirement, deep linking, or in-app navigation information and maintaining version validity; completion or abandonment of transaction flows Status, abandoned screen, error type, verification result, user 25 Correction, number of missing parameter requests, stream execution time, screen transitions. to record delay and past success rate data and this data Generation of candidate screen sequences by the server (4), to be used in evaluation and determination of process flow 30 characterized by the database structured to provide (3) a system like any of the above requests (1). 12 5. Using any remote communication protocol, the application (2) and data to communicate with the base (3) and through this communication to the application (3) server (4) configured to exchange data with a system like any of the above characterized claims (1). 5 6. Bill payment, purchasing additional packages, changing tariffs, renewing contracts, line the process, installment payment viewing, digital service activation, opening a fault report, creating a support request, adjusting card limits, transferring money, credit application, investment product viewing, return initiation, delivery address 10 Change, use coupon, application tracking, document request or appointment interpreting the user request in the form of creation and Domain-specific trained transducer used during parameter determination. natural language comprehension model based on intent-domain collaborative learning model, semantic Using a similarity model or rule-based hybrid discriminator; intention 15 Converter-based classifiers for inference, Bidirectional Converter Encoder Representations derivatives, multilingual language models, lightweight mobile natural language from language processing models or intent-domain collaborative learning models named entity recognition, array for utilizing parameter extraction labeling, conditional random field-supported labeling, or Big Language Model 20 using structured output generation methods based on user making the operation and parameter information in the sentence processable. The above is characterized by the server (4) configured to provide a system like any of the requests (1).

7. Screen ID registered in the database (3), the action performed by the screen steps, accepted parameters, required user inputs, validation, and User consent requirements, transaction completion conditions, deep linking or navigation information and data relating to other accessible screens. 30 that define the steps each screen can perform. to evaluate screen capacity profiles; screens (2) in the application 13 allowed range between nodes and screens with processing capacity using directional screen graphics that model transitions as directional edges and edge-specific transition conditions, transition cost, user authorization level, validation requirements, deep link validation, and version compatibility. 5 to match the information with the required steps in the process diagram from the above requests characterized by the configured server (4) a system like any other (1).

8. Able to perform each mandatory step in the process diagram. To identify the screens, locate the screen nodes on the screen graph, 10 showing the semantic compatibility of the screens with the processing capacity and processing steps. Evaluating with semantic similarity score; in screen-step matching learned ordering model, gradient-increasing trees, graph neural network based node fitness model or embedding vector-based proximity search by utilizing these methods, the highest compliance with the relevant process step is 15. Identifying candidate screen nodes showing the identified candidate screens Multiple candidates can be sorted by ranking nodes along valid oriented edges. creating a screen array and the processing scheme of the candidate arrays meeting the mandatory steps from start to finish 20 characterized by the server (4) configured to determine their status a system like any of the above-mentioned requests (1).

9. Each generated candidate screen sequence must be defined in the process diagram as mandatory. Transaction coverage score, which indicates the extent to which transaction steps have been met. Technical and user issues that occur when switching between screens 25 Transition cost, which indicates the interaction load, is not requested by the user. However, information that needs to be obtained later in order to complete the process cost of missing parameters, number of screens, repetitive information resulting from logins, user approval steps and waiting times The user friction score, which indicates the interaction load, requires 30 for the process. Compatibility of parameters with parameters accepted by the screens, verification. 14 The status of meeting requirements is the same as in similar user requests. or showing the frequency of successful completion of similar screen sequences past success rate and screens are valid in the current application version. Evaluating based on the information that is not available; screen transition cost, user Friction cost, missing parameter completion cost, verification 5 total execution delay, version incompatibility risk and security risk calculating the cost and creating candidate screen sequences. shortest path in the ranking, A-star search, bundle search, multi Targeted optimization, policy updating supported by reinforcement learning. or contextual multi-arm rogue-based stream sequencing algorithms 10 characterized by the server (4) configured to benefit from a system like any of the above requests (1).

10. User authorization level during the evaluation of candidate screen sequences, Transaction sensitivity class, authentication requirement, explicit user consent 15 rules, allowed deep links, and application version validation. implementing this information as mandatory flow constraints; authentication or the process step does not include user approval, user authorization Deep linking that goes beyond the level of access to screens that are not allowed. screen that is not valid in the current application version (2) or 20 candidate screen sequences, including transitions, are presented to the user before... to eliminate; security and validation decisions from probabilistic model output to do it independently through deterministic policy and constraint control and only security, verification, user consent, authorization and Candidates who meet all of the minimum transaction coverage requirements are 25 among them, those with low total execution cost and high transaction coverage score. The server (4) is characterized by being configured to select the screen sequence which is available. a system like any of the above-mentioned requests (1).

11. The target transaction type determined from the user's single-sentence request is 30. the selected candidate screen sequence for completion, target operation The application (2) screens that will open sequentially during the execution, The process parameters to be transferred between screens will be determined by the user's initial settings. not requested but necessary for the completion of the target transaction Time, quantity, service name, product name, account or line information, and transaction condition. Missing parameters such as authentication and explicit user consent 5 screens requiring navigation or deep link commands, errors the screen to return to if the issue occurs, and the screen indicating that the process is complete. condition and screen or process to continue when the process is interrupted to convert it into an executable compiled process flow that includes the step 10 of the above requests characterized by the configured server (4) a system like any other (1).

12. The uncertainty score obtained from the user's request exceeding the specified threshold value. If this is the case, the target transaction type needs to be clarified. Explanatory question or data entry to be presented to the user via application (2) 15 the area, parameters necessary for performing the target operation It is incomplete if no one makes a user request. a question or data entry that will allow the user to enter the parameter. Adding the field to the compiled workflow and user authorization, identity verification, explicit user consent, valid screen or deep link, and 20 a screen array that meets minimum transaction coverage requirements If it cannot be created securely, the target operation redirection to the current application screen (2) that can be started or missing A question or data entry field that will allow information to be obtained from the user. The above 25 is characterized by the server (4) configured to create it. a system like any of the requests (1).

13. Execute the generated compiled workflow on application (2) to transmit; to add or remove screens in the application version, deep connection diagrams, screen parameters or navigation 30 Screen graphics and screen capacity profiles in case the rules change. 16 to update, deactivate outdated screens and transitions, and Recompiling the process flows created based on the previous version; Completion or abandonment status of the stream obtained via application (2), abandoned screen, error type, validation result, user correction, missing Number of parameter requests, stream execution time, screen transition delay, and 5 Using historical success rate data, intent matching scores are displayed on the screen. Updating the transition costs and flow selection policy in the graph, to increase the preferred value of successfully completed screen series and continuously abandoned, error-prone, or validation-failed transitions by increasing the cost, transition costs in subsequent user requests and 10 To ensure that past success rates are evaluated in an updated manner. the above requests are characterized by the server (4) configured to be so. a system like any other (1). 20 30

Images