An artificial intelligence-based automated government affair examination and approval system and method

Abstract

This application provides an automated government approval system and method based on artificial intelligence. It breaks down the government approval process into multiple independently operable functional modules. Through artificial intelligence and data-driven technologies, it intelligently models key aspects of government processing, such as material handling, rule judgment, process planning, and result generation. By collaborating between modules, it achieves an automated, efficient, and interpretable approval process. This provides a comprehensive solution integrating artificial intelligence information processing capabilities, policy knowledge rule engines, and process automation mechanisms. It offers comprehensive advantages such as flexible adaptation, high processing efficiency, traceable results, and controllable strategies, and can promote the development of government approval systems towards a new stage of efficiency, fairness, and intelligence.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, specifically to an automated government approval system and method based on artificial intelligence. Background Technology

[0002] Today, the traditional approval model, which relies on paper materials and manual processes, is gradually evolving towards electronic systems. With the deployment of e-government platforms, electronic operations have been achieved in areas such as business acceptance, document uploading, approval processes, and result feedback through process modeling and information system integration. Some matters have already achieved cross-departmental collaboration and parallel approval, initially alleviating problems such as slow manual processes and duplicate document submissions.

[0003] However, the inventors of this application have found that most existing government approval systems are based on form-driven and workflow engines, supporting business logic through the configuration of approval nodes and data validation rules. They typically provide standardized electronic forms for users to fill in application information and support the uploading of attachments. After the application is accepted, staff members conduct approval and judgment in the background and process the approval results according to policies and regulations. Although such systems improve the overall efficiency of business operations, their processing still has many technical bottlenecks and operational limitations, specifically the following:

[0004] 1) Insufficient information extraction and structuring capabilities. The materials submitted by users come from various sources, including PDF documents, scanned copies, handwritten application forms, and multi-format spreadsheets. The system lacks the ability to identify and transform unstructured data, which means that approvers still need to manually review the materials and enter key fields into the system, increasing labor costs and the probability of errors.

[0005] 2) The execution of approval rules relies on manual judgment. Although some systems support rule configuration, most current rules are simple field validations or process branch judgments, which are difficult to support complex conditions, multi-level logic, and scenarios with dynamic policy changes. Approving personnel often rely on experience to understand the provisions of laws and regulations and make case-by-case judgments, resulting in frequent inconsistencies in approval standards and results.

[0006] 3) Limited cross-system data linkage capabilities. In matters handled collaboratively by multiple departments, there are issues such as inconsistent data interface standards, information silos, and access barriers between different systems. Approving personnel often need to manually retrieve data from external systems or verify applicant information, company qualifications, historical records, etc., offline, which affects the overall approval timeliness and accuracy.

[0007] 4) Lack of intelligent interaction and process guidance. Currently, most government service systems rely on static pages and fixed field input, making it difficult for users to obtain real-time feedback or personalized guidance during the process. For applicants unfamiliar with the business process, this can easily lead to problems such as misfilling or omissions, increasing the number of communications and corrections, and reducing the user experience.

[0008] In addition, existing systems generally fail to integrate artificial intelligence (AI) technology capabilities, and are still weak in areas such as deep semantic analysis of government data, intelligent recognition of material content, automatic matching of policy provisions, and dynamic understanding of user intent. This makes it impossible for the system to achieve true automation of approval and intelligent assisted judgment, and it is also difficult to adapt to government scenarios with high-frequency processing, high-complexity judgment, and diverse user needs.

[0009] In summary, although the existing government approval technology system has achieved a certain degree of informatization and process automation, it still faces key problems such as insufficient intelligence, low system collaboration efficiency, inconsistent rule execution, and poor user experience. There is an urgent need for a comprehensive solution that integrates artificial intelligence information processing capabilities, policy knowledge rule engines, and process automation mechanisms to promote the development of the government approval system towards a new stage of efficiency, fairness, and intelligence. Summary of the Invention

[0010] This application provides an automated government approval system and method based on artificial intelligence. It breaks down the government approval process into multiple independently operable functional modules. Through artificial intelligence and data-driven technologies, it intelligently models key aspects of government processing, such as material handling, rule judgment, process planning, and result generation. By collaborating between modules, it achieves an automated, efficient, and interpretable approval process. This provides a comprehensive solution integrating artificial intelligence information processing capabilities, policy knowledge rule engines, and process automation mechanisms. It offers comprehensive advantages such as flexible adaptation, high processing efficiency, traceable results, and controllable strategies, and can promote the development of government approval systems towards a new stage of efficiency, fairness, and intelligence.

[0011] In the first aspect, this application provides an automated government approval system based on artificial intelligence. The automated government approval system includes a data access module, an intelligent parsing module, a rule reasoning module, an approval path planning module, a process scheduling module, and a result generation and feedback module. Each module operates independently and collaborates with an event-driven communication mechanism through a standard interface.

[0012] The data access module is used to receive approval materials submitted by users through government service terminals under preset material formats and channels, and convert them into internal standard data formats.

[0013] The intelligent parsing module is used to perform text recognition on approval materials, and based on the text recognition results, to identify the fields required for approval, thus obtaining the recognition results of the fields required for approval.

[0014] The rule reasoning module is used to perform compliance verification on the identification results of the fields required for approval;

[0015] The approval path planning module is used to plan the approval path based on the field identification results that meet the compliance requirements of the approval, and obtain a suitable approval path graph. The approval path graph is configured in a directed graph structure, where nodes represent approval links or responsible entities, and edges represent reachable paths or conditional branches.

[0016] The process scheduling module is used to drive the execution of each node in the approval path based on the approval path diagram;

[0017] The results generation and feedback module is used to compile approval conclusions based on the execution status of the approval path diagram and push feedback to multiple terminals.

[0018] Secondly, this application provides an automated government approval method based on artificial intelligence. The method is applied to an automated government approval system based on artificial intelligence. The automated government approval system includes a data access module, an intelligent parsing module, a rule reasoning module, an approval path planning module, a process scheduling module, and a result generation and feedback module. Each module operates independently and collaborates through a standard interface and an event-driven communication mechanism. The method includes:

[0019] The data access module receives approval materials submitted by users through government service terminals under preset material formats and channels, and converts them into internal standard data formats.

[0020] The intelligent parsing module performs text recognition on the approval materials and, based on the text recognition results, identifies the fields required for approval, thus obtaining the recognition results for the fields required for approval.

[0021] The rule-based reasoning module performs compliance verification on the recognition results of the fields required for approval;

[0022] The approval path planning module performs approval path planning on the field identification results that meet the compliance requirements of the approval, and obtains a suitable approval path graph. The approval path graph is configured in a directed graph structure, where nodes represent approval links or responsible entities, and edges represent reachable paths or conditional branches.

[0023] Based on the approval path diagram, the process scheduling module drives the execution of each node in the approval path;

[0024] The results generation and feedback module compiles approval conclusions based on the execution status of the approval path diagram and pushes feedback to multiple terminals.

[0025] Thirdly, this application provides a computer-readable storage medium storing a plurality of instructions adapted for loading by a processor to execute the method provided in the second aspect of this application.

[0026] From the above, it can be concluded that this application has the following beneficial effects:

[0027] This application breaks down the government approval process into multiple independently operable functional modules. Through artificial intelligence and data-driven technologies, it intelligently models key aspects of government processing, such as material handling, rule judgment, process planning, and result generation. By collaborating between modules, it achieves an automated, efficient, and interpretable approval process. This provides a comprehensive solution that integrates artificial intelligence information processing capabilities, policy knowledge rule engines, and process automation mechanisms. It has comprehensive advantages such as flexible adaptation, high processing efficiency, traceable results, and controllable strategies, and can promote the development of government approval systems towards a new stage of efficiency, fairness, and intelligence. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a schematic diagram of a system architecture for an automated government approval system based on artificial intelligence, as described in this application.

[0030] Figure 2 This is a flowchart illustrating an automated government approval method based on artificial intelligence, as described in this application. Detailed Implementation

[0031] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules is not necessarily limited to those explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices. The naming or numbering of steps appearing in this application does not imply that the steps in the method flow must be performed in the chronological / logical order indicated by the naming or numbering. The execution order of named or numbered process steps can be changed according to the desired technical purpose, as long as the same or similar technical effect is achieved.

[0033] The module division described in this application is a logical division. In practical applications, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the coupling or direct coupling or communication connection between modules shown or discussed may be through some interfaces, and the indirect coupling or communication connection between modules may be electrical or other similar forms, none of which are limited in this application. Furthermore, the modules or sub-modules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed in multiple circuit modules. Some or all of the modules may be selected to achieve the purpose of the solution in this application according to actual needs.

[0034] First, refer to Figure 1 The diagram shown represents a system architecture of the AI-based automated government approval system of this application. The AI-based automated government approval system provided in this application may specifically include several modules: a data access module, an intelligent parsing module, a rule reasoning module, an approval path planning module, a process scheduling module, and a result generation and feedback module. Each module operates independently and collaborates through a standard interface and an event-driven communication mechanism.

[0035] It is now understood that this application breaks down the government approval process into multiple independently operable functional modules. Through artificial intelligence and data-driven technologies, it intelligently models key aspects of government processing, such as material handling, rule judgment, process planning, and result generation. Through inter-module collaboration, it achieves an automated, efficient, and interpretable approval process.

[0036] Specifically, the modules can communicate with each other through an asynchronous message bus that uses a lightweight message queue protocol, such as Message Queuing Telemetry Transport (MQTT) or Kafka. Each module triggers the corresponding calculation logic after receiving a specific event and publishes the next stage event after processing, thus forming a loosely coupled process link. This mechanism can further improve the system's concurrency capabilities and avoid the overall process from being delayed due to the blocking of approval nodes.

[0037] Each module is deployed in a distributed architecture and has clear boundaries formed by modular service division. In actual engineering implementation, it is recommended to adopt a front-end and back-end separation architecture design, with the front-end interface interacting with the back-end approval engine via HTTP or gRPC.

[0038] Thus, the overall architecture adopts a modular and distributed deployment mode, with each functional unit of the system operating independently and collaborating with each other through standard interfaces and event-driven communication mechanisms to form a flexible and scalable approval service network. After receiving the processing results from the previous module, each module automatically selects the subsequent processing path according to the configured service orchestration strategy or model recommendation, realizing the dynamic and non-linear flow of the approval process.

[0039] Meanwhile, in terms of specific implementation, the backend of the system mainly consists of several microservices, including material identification service, field extraction service, rule engine service, path planning service, scheduling and control service, result generation service, and log archiving service. Each service is deployed as a containerized unit, running on cloud servers or local government dedicated server nodes. Automatic scaling and high availability deployment are achieved through a service orchestration platform (such as a Kubernetes management platform). The modules communicate with each other through RESTful APIs or message queues, and have good horizontal scalability.

[0040] The system's storage layer includes a relational database (such as PostgreSQL) for storing user materials, field structures, approval process information, etc.; an object storage system for saving uploaded image data and intermediate recognition results; and a log system (such as ELK or ClickHouse) for collecting and analyzing model behavior and operation trajectories.

[0041] The hardware platforms on which each functional module depends for operation include, but are not limited to: multi-core processors (Central Processing Unit, CPU), graphics accelerators (Graphics Processing Unit, GPU or Neural Processing Unit, NPU) for running relevant models, storage units for program and data storage, and network interfaces for connecting to external systems. While ensuring functional accuracy, the system also supports operation in private deployment environments, meeting the high requirements for information security and data isolation in government scenarios.

[0042] Next, we will elaborate on the working process of the modules shown above and the collaborative relationships between them.

[0043] (1) Data access module

[0044] The data access module is used to receive approval materials submitted by users through government service terminals under preset material formats and channels, and convert them into internal standard data formats.

[0045] Understandably, the data access module corresponds to the input part of this application system. It allows users to initiate approval tasks and submit relevant approval materials independently through the relevant user interface. Of course, in practice, there are also situations where the system initiates approval tasks independently or other individuals initiate them. That is, the current approval task may not necessarily be initiated by the user currently submitting the approval materials, which corresponds to diverse application needs in real-world scenarios.

[0046] As an example, users can submit approval materials through the corresponding government service terminals.

[0047] Furthermore, it can be seen that this application has pre-determined different material formats and different material channels in terms of details, forming a preset material format and preset material channel, so as to meet the diverse and standardized requirements for submitting approval materials.

[0048] As an exemplary embodiment here, for the data access module, the preset material format may specifically involve structured electronic forms, scanned copies, and PDF documents, etc., and the preset material channel may specifically involve web forms, WeChat mini programs, and government apps, etc. Furthermore, after the original materials submitted by the user are encrypted and desensitized for sensitive fields including ID card numbers and corporate tax numbers, they are converted into an internal standard data format (which can be denoted as D_input) and tracked during the auditing process through a unique identifier (ID).

[0049] Understandably, using methods such as hashing or salting for encryption and desensitization helps to further improve the information security of this application system and better protect user privacy. Under this setting, subsequent data processing or approval processes can be easily understood to proceed under encrypted and desensitized conditions.

[0050] Furthermore, for the anonymized data, federated learning technology can be used to complete collaborative model training in subsequent model training stages without exposing the original data. This mechanism ensures cross-regional knowledge sharing while protecting data security.

[0051] In addition, the data access module can preprocess the data it receives, which can enhance data quality, filter out abnormal data, or perform adaptive processing for subsequent data processing.

[0052] Taking image materials such as business licenses, ID card scans, and tax registration certificates as examples, specific preprocessing operations such as noise reduction, correction, and edge enhancement can be performed.

[0053] For the structured electronic forms, the field content can be cleaned and standardized to generate a unified "field-value" mapping dictionary structure.

[0054] Of course, preprocessing can be handled not only by the data access module, but also by the subsequent intelligent parsing module or even other modules. This is possible in practice, and can be flexibly configured according to actual needs.

[0055] (2) Intelligent parsing module

[0056] The intelligent parsing module is used to perform text recognition on approval materials, and based on the text recognition results, it identifies the fields required for approval to obtain the recognition results of the fields required for approval.

[0057] Understandably, the intelligent parsing module is used to process the approval materials accessed by the preceding data access module into specific fields that can be processed by the subsequent rule reasoning module, which are the fields required for approval as mentioned here. This process realizes the transformation of raw materials into business semantic structures, laying a good foundation for intelligent approval.

[0058] This involves two stages of processing: the first stage is the text recognition stage, and the second stage is the structured extraction stage.

[0059] Specifically, in the first stage, for unstructured materials such as images and PDF documents, technologies such as Optical Character Recognition (OCR) can be used to identify the text / text involved.

[0060] As an example, the text recognition result can be configured as an output format of a two-dimensional character matrix M_text.

[0061] In the second stage, a pre-trained named entity recognition model and pattern matching mechanism can be used to employ a deep neural network model based on multi-channel feature fusion, combined with multi-dimensional features such as lexical, syntactic, and entity context, to achieve accurate field positioning and semantic classification, and extract the fields required for subsequent approval, such as subject name (e.g., company name), subject type, item category, certificate type, processing area, validity period, application time, contact information, unified social credit code, registered capital, legal representative, and other specific fields.

[0062] The extraction function can be represented in the following form:

[0063] ,

[0064] Here, F represents the set of structured fields, which is the core input for subsequent rule-based reasoning and process planning. This indicates the input text. This represents the extraction function.

[0065] In addition, the intelligent parsing module here can also support noise reduction and word segmentation of redundant fields, and ambiguity elimination of polysemous expressions in long texts, thereby ensuring that the field extraction accuracy meets the definition of the minimum information unit required for approval.

[0066] All extracted fields are uniformly converted into an approval field set, which serves as the subsequent input data structure. Each field in the field set is accompanied by attributes such as field type, extraction source, confidence level, and whether it requires manual confirmation. The system can set prompts for low-confidence fields for manual verification or to trigger supplementary tasks.

[0067] (3) Rule-based reasoning module

[0068] The rule reasoning module is used to perform compliance verification on the recognition results of the fields required for approval.

[0069] Understandably, the compliance check here is to verify whether the content of the materials meets the requirements. For example, if the applicant is an individual, the system determines whether a valid identity document has been provided; if the application is for a change of qualification, the system checks whether the original certificates and the basis for the change have been uploaded.

[0070] The results of all the verifications or judgments can be summarized into a compliance status checklist, which includes: whether it meets the approval requirements, whether there are any omissions, a list of missing fields, the types and descriptions of materials to be supplemented, and suggestions for subsequent processing.

[0071] Understandably, providing clear and highly visual reminders for non-compliant areas can create real-time feedback, allowing users to take appropriate action to pass compliance checks.

[0072] In terms of details, the rule reasoning module, during the specific compliance verification process, needs to follow the legality judgment rules required by the relevant approval procedures. These rules can be stored in a rule base, which can be stored in a Domain Specific Language (DSL) format, supporting combinational logic expressions. Rule matching is achieved by constructing Boolean expressions and matching them with field values, outputting a judgment flag J to indicate whether a field is compliant, missing, or conflicting. This module has the ability to dynamically load rules and manage versions, ensuring flexible adaptation to different policies in various regions. Its judgment logic is expressed as follows:

[0073] ,

[0074] Where J is the inference result vector, R represents the rule set of the current business type, supports structured tag output, and F represents the set of structured fields input.

[0075] Furthermore, as an exemplary embodiment here, for the rule reasoning module, each rule is composed of elements such as logical expression, weight condition, mandatory field validation, cross-field logical validation and dependency relationship. Compliance judgment is made through field comparison, and specifically it may involve qualification verification rules, regional applicability judgment rules, applicant authority judgment rules and material completeness judgment rules.

[0076] The rule expression uses the following structure:

[0077] ,

[0078] Where R represents the rule set; This represents the i-th rule; Indicates the total number of rules; Represents structured fields; This represents the condition for the rule function. A result of 1 indicates that the rule is satisfied, and a result of 0 indicates that the rule is not satisfied.

[0079] (4) Approval Path Planning Module

[0080] The approval path planning module is used to plan the approval path based on the field identification results required for approval to meet compliance requirements, and obtain a suitable approval path graph. The approval path graph is configured in a directed graph structure, where nodes represent approval links or responsible entities, and edges represent reachable paths or conditional branches (or sequential dependencies or parallel relationships).

[0081] In simple terms, the approval path planning module here is to construct the approval process involved in the current approval task, and specifically involves the form of a directed graph. That is to say, it uses graph search and strategy prediction models to infer the optimal path by combining historical data, business configuration and AI model output.

[0082] Specifically, the approval path planning module can automatically generate corresponding approval execution paths based on the application type, compliance status, and preset process models in the business database. In terms of details, the path consists of a set of nodes, each node representing an approval sub-task. Node types include automated approval nodes, manual review nodes, and hybrid processing nodes. The system configures trigger conditions, processing content, and output structure for each node. The approval path supports a hybrid serial and parallel structure and uses a graph structure to represent the path flow. The dependencies between each node can be defined by logical expressions, such as: "If the materials are complete, skip the manual verification node and directly enter the system verification node."

[0083] The approval path planning module can also statistically analyze the number of nodes, node type combinations, and approval time of each generated approval path graph to evaluate whether the approval path is optimal. It also continuously updates the path generation strategy through reinforcement learning algorithms, thereby selecting a more efficient path graph structure in the next round of approval of similar matters.

[0084] In addition, this feature can also support path visualization and decision interpretation, facilitating the monitoring of the approval process.

[0085] The path function is expressed as follows:

[0086] ,

[0087] Where G represents the generated approval path diagram, F represents the set of input structured fields, and J represents the inference result vector. Represents a path function.

[0088] Specifically, as an exemplary embodiment, the path planning strategy model is based on constructing a state space using a graph neural network, and determines a suitable path structure by combining path cost and strategy preference.

[0089] The path cost function can be expressed by the following formula:

[0090] ,

[0091] in, This represents the total cost of path P; This represents the average processing time of the i-th stage; Indicates the weight of each stage; Indicates the total number of stages; This represents the path stability adjustment coefficient; This represents the total impact value of uncertain factors in the path.

[0092] Understandably, this section presents a reference factor for path cost in directed graph-based approval path planning, and provides a practical implementation scheme based on specific quantitative formulas.

[0093] (5) Process scheduling module

[0094] The process scheduling module is used to drive the execution of each node in the approval path based on the approval path diagram.

[0095] It is understandable that the process scheduling module is used to advance the approval process of different planned approval nodes. Its scheduling logic supports an event-driven mechanism, that is, when a node completes its processing, the system automatically publishes the processing result to the downstream node, triggering the next approval step. The processing methods of the nodes include manual confirmation, AI approval suggestions, and automatic decision-making. The process status is managed using a finite state machine model, and the transition function can be expressed as:

[0096] ,

[0097] in, This represents the current state at time t. This indicates the triggering event at time t. This represents the transition function, and the system dynamically updates the state based on the policy rule set.

[0098] Thus, the previously constructed approval path diagram is sent in, and the process scheduling module begins to process the approval tasks node by node. Whenever an approval node is activated, the system determines the node type and selects the corresponding execution logic.

[0099] For automated approval nodes, the system calls the program to perform logical tasks such as field validation, material consistency comparison, and external interface validation, and records the processing results; for manual approval nodes, the system pushes the task to the government affairs back-end personnel for processing through the interface, and the manual approval node returns the result to the system through the interface after processing.

[0100] The execution result, processing time, status code, and exception information of each approval node will be recorded to ensure the traceability and closed-loop control of the process.

[0101] (5) Result generation and feedback module

[0102] The results generation and feedback module is used to compile approval conclusions based on the execution status of the approval path diagram and push feedback to multiple terminals.

[0103] As is easy to understand, the result generation and feedback module is responsible for the final processing of the approval task. Based on the execution status of each approval node, the output data, and the final summary judgment, it organizes the approval conclusion. In this process, it may involve specific approval conclusion information such as approval status (approved, rejected, returned for modification, etc.), approval basis chain (including key judgment links), rule set used, approval time record, and material completeness description (suggestion to supplement materials). The final result is composed of structured data and natural language description, and is pushed to users and business systems. It supports the generation of legally valid approval documents through electronic signatures.

[0104] All approval results will be pushed to the business system database for registration and filing via the synchronous interface. Users can also obtain approval status, progress nodes, and feedback content in real time through the approval query interface. Simultaneously, based on the actual needs of government services, results can also be output to SMS services, government app push channels, or SMS reminder interfaces to achieve multi-terminal notification functionality.

[0105] In addition, as an example, the result generation feedback module can generate supplementary explanatory text while outputting the approval result to help users understand the logical basis of the approval conclusion.

[0106] For example, if the approval fails, the system will automatically list the rejected rule entries, corresponding fields, and suggested supplementary materials based on the rule matching path.

[0107] This explanatory text is generated based on the large language model completion method and business language expression templates. It can be understood as a processing instruction. Without involving core algorithm details, it is presented in the "because...therefore..." structure, which can effectively improve approval transparency and user satisfaction.

[0108] Furthermore, as an exemplary embodiment, the automated government approval system based on artificial intelligence in this application may also include a training management module;

[0109] This training management module is used to uniformly manage the text recognition model of the intelligent parsing module for text recognition, the named entity recognition model of the intelligent parsing module for recognizing fields required for approval, the rule adaptation model of the rule reasoning module for compliance verification, and the path planning strategy model of the approval path planning module for approval path planning. It manages the training, evaluation, and updating of each model to maintain the system's adaptability in different government affairs scenarios.

[0110] In practice, the training management module can also build training samples by collecting information such as historical approval data, user interaction behavior, and rule adaptation feedback, and continuously optimize model performance through methods such as supervised learning, transfer learning, and reinforcement learning.

[0111] Furthermore, for the named entity recognition model involved, the following loss function can be used during training:

[0112] ,

[0113] in, This represents the total loss of the model. Used to measure the difference between the predicted result and the true label; n represents the length of the input sequence, expressed as the total number of characters or words in the text; C represents the number of categories, including names, places, organizations, and others; This indicates whether the true label of the i-th character or word belongs to the c-th class. If it does, the value is 1; otherwise, it is 0. This represents the probability that the model predicts the i-th character or word belongs to the c-th class, which is obtained through the softmax operation.

[0114] Understandably, this application provides a highly adaptable and practical application scheme for the loss function involved in the model training process of the named entity recognition model.

[0115] Similarly, on the other hand, for path planning strategy models, the following loss function can be used during training, with the expected path deviation as the optimization objective:

[0116] ,

[0117] in, Indicates the expected deviation of the path; Represents the actual path sequence vector; This represents the model's predicted path; N represents the total number of training samples. This strategy can effectively improve the prediction accuracy and execution efficiency of the approval path.

[0118] In addition to the training management module, as an exemplary embodiment, the AI-based automated government approval system of this application may also include a log audit module.

[0119] This log auditing module is used to record the entire process of each approval request. That is, it records the entire process of each approval request in the form of audit logs. Specifically, it can include data access time, processing steps, rule hit status, approver suggestions, system suggestions, final results, original material content, field extraction results, rule judgment path, approval node sequence, node result status, records of manual intervention behavior, and user feedback content. It uses an immutable chain structure for storage, which supports the regulatory platform to read, replay behavior, and trace the source.

[0120] Based on the aforementioned log data, the system can periodically trigger the model fine-tuning training process. Taking the named entity recognition model as an example, the system can compare the differences between the fields extracted by the model and the fields modified by the final approver, and generate training sample pairs for fine-tuning the model; in the subsequent rule matching part, i.e., the rule adaptation model, the rule expressions can also be adjusted and optimized based on the fields manually added by the approver and the rule feedback.

[0121] The log auditing module can also periodically link with the preceding training management module to use historical processing behavior as a training sample for the training management module's related model training work, thereby achieving closed-loop optimization system capabilities.

[0122] Meanwhile, during the system's operation or approval process mentioned above, the system can also determine whether there is abnormal data or rule conflicts. If there are uncertainties or high-risk approval behaviors, the system will trigger a manual review mechanism. This manual review mechanism is easy to understand; it will push the current approval context information to the manual operation console and suspend the subsequent automatic processing flow until the manual operation provides a clear opinion. While ensuring the efficient operation of the system, it retains the ability of human-machine collaboration to deal with complex situations, which can be regarded as a backup measure.

[0123] At the same time, the system also has self-learning capabilities. Its core is the approval feedback learning module of the linkage training management module. Through the result generation feedback module, in the further feedback collection and processing, it can record feedback collection information such as backend data of approval results, user appeals, and manual review opinions, and transform them into a kind of training sample. This sample is provided to the relevant model training work of the training management module, which also enables the system to periodically retrain the model and fine-tune the parameters, thereby improving the accuracy, rationality and user satisfaction of the approval process.

[0124] Thus, taking the closed-loop learning mechanism here as an example, the entire system of this application has a high data closed-loop capability, and can realize a continuous iterative closed loop of "user material submission → intelligent judgment and processing → result output → result feedback → data archiving and learning → model optimization and update".

[0125] In conclusion, regarding the above system settings, this application breaks down the government approval process into multiple independently operable functional modules. Through artificial intelligence and data-driven technologies, it intelligently models key aspects of government processing, such as material handling, rule judgment, process planning, and result generation. By collaborating between modules, it achieves an automated, efficient, and interpretable approval process. This provides a comprehensive solution integrating artificial intelligence information processing capabilities, policy knowledge rule engines, and process automation mechanisms. It offers advantages such as flexible adaptation, high processing efficiency, traceable results, and controllable strategies, and can drive the government approval system towards a new stage of efficiency, fairness, and intelligence.

[0126] In terms of details, there are:

[0127] 1) By introducing artificial intelligence technology, the level of automation in government approval processes has been significantly improved.

[0128] In traditional government approval processes, material identification, information comparison, rule judgment, and process flow mostly rely on manual operation, which not only has a long processing cycle and a high error rate, but also makes it difficult to adapt to scenarios with high concurrency and multiple types of business applications.

[0129] In contrast, this application integrates intelligent algorithms such as image recognition, natural language processing, rule reasoning, and path optimization, which can automatically complete the entire process from material receipt to approval conclusion output without human intervention, greatly improving approval efficiency and reducing labor costs.

[0130] 2) The modular architecture proposed in this application has good system scalability and flexibility.

[0131] The various modules work together through standard interfaces and asynchronous communication mechanisms, enabling the system to quickly adapt or replace some logical components when dealing with different government business scenarios without the need for overall reconstruction.

[0132] Meanwhile, the architecture supports microservice deployment, which facilitates later system maintenance, function iteration and cross-regional deployment, and achieves unified platform support while meeting the differences in local policies.

[0133] 3) Through the synergy of the rule engine and the approval path planning model, dynamic adaptation to complex approval rules and personalized customization of approval processes are achieved.

[0134] When faced with different types of approval items, the system can automatically determine the applicable rules based on the materials submitted by the user, combine approval nodes, and predict the optimal approval path, thereby ensuring that the process is both legal and compliant, as well as efficient and controllable. This mechanism can effectively reduce problems such as duplicate approvals and rollbacks caused by inconsistent understanding of rules or unreasonable node settings, and improve the transparency and continuity of approvals.

[0135] 4) The feedback learning mechanism introduced in this application constructs an adaptive closed-loop optimization system.

[0136] By continuously collecting information such as historical approval data, user interaction records, and manual review opinions, the system can periodically optimize relevant models, thereby continuously improving their accuracy and adaptability in actual operation.

[0137] This feature enables the system to evolve by becoming more accurate with use, making it particularly suitable for real-world application environments where government rules change frequently and business scenarios are highly diverse.

[0138] It also strengthens the compliance of data processing and the interpretability of results. In the approval process, the system generates processing records and explanations for each decision node, which facilitates review by regulatory authorities and user appeals.

[0139] Meanwhile, the log auditing module provides a fully traceable approval process. Combined with the model's interpretable algorithm, it enables manual confirmation and accountability for intelligent decision-making results, thus addressing the risks and concerns arising from the "black box" nature of current intelligent systems.

[0140] 4) It demonstrates a clear advantage in improving the standardization and consistency of government services.

[0141] By modeling the approval rules in a structured and executable logical expression, the system can maintain consistent rule application standards across different service windows and different operators, thereby effectively eliminating the inconsistency in results caused by misunderstandings and operational differences in traditional manual approval.

[0142] This standardization capability not only enhances the standardization of government services, but also provides important technical support for cross-regional government collaboration and policies such as "one-stop online services".

[0143] At the same time, it has achieved the ability to integrate and process multimodal government data, which has a significant advantage in dealing with complex application materials.

[0144] Traditional government systems rely heavily on structured forms, while real-world government documents often exist in the form of scanned copies, paper certificates, and unstructured text. By integrating text recognition, natural language processing, entity recognition, and semantic analysis algorithms, this system can deeply analyze mixed text and image input materials in multiple formats, extract key elements, and construct a unified data representation format. This effectively bridges the "last mile" of government data processing, significantly improving the system's intelligent understanding capabilities and the breadth of its document processing capabilities.

[0145] 6) This application also has a positive effect on improving user experience.

[0146] By building a front-end intelligent guidance mechanism and a back-end personalized recommendation mechanism, the system can provide real-time prompts for missing fields, offer sample content, and predict potentially non-compliant items during the user's submission of materials or filling out of forms, reducing the difficulty for users to fill out forms and reducing the number of rejections due to incomplete information.

[0147] Meanwhile, the approval feedback information is not limited to a conclusion of approval or disapproval, but also includes the basis for approval, explanations of the rules, and supplementary suggestions, enabling users to understand the process more clearly and improving their trust and satisfaction with government services.

[0148] 7) This application supports access to government approval processes through multiple channels and terminals, and has good service coverage.

[0149] The system can be deployed on various government service terminals, including government websites, government apps, WeChat official accounts, mini-programs, and offline self-service terminals, achieving integrated online and offline access. In different terminal environments, the system ensures consistent approval processes and synchronized data processing through unified data standards and service interfaces, contributing to the comprehensive digital transformation and intelligent upgrading of government services.

[0150] 8) The modular decoupling feature of this application also gives it good engineering maintainability and technical sustainability.

[0151] In actual deployment, the system can flexibly adjust the operating logic of each module according to business load, actual regional needs or policy evolution. This not only facilitates fault location and performance optimization, but also reduces the development and testing costs when new policies are launched, avoiding common problems in traditional centralized systems such as update lag and complex adaptation.

[0152] The above is an introduction to the AI-based automated government approval system provided in this application. Based on this system, this application also provides an AI-based automated government approval method. Clearly, this method is applied to the AI-based automated government approval system, which, in brief, includes several modules: a data access module, an intelligent parsing module, a rule reasoning module, an approval path planning module, a process scheduling module, and a result generation and feedback module. Each module operates independently and collaborates through a standard interface and an event-driven communication mechanism.

[0153] Based on the above system design, the automated government approval method based on artificial intelligence provided in this application, with reference to... Figure 2 The diagram shown represents a flowchart of an automated government approval method based on artificial intelligence, which may specifically include the following steps S201 to S206:

[0154] Step S201: The data access module receives the approval materials submitted by the user through the government service terminal under the preset material format and preset material channel, and converts them into the internal standard data format.

[0155] In step S202, the intelligent parsing module performs text recognition on the approval materials and, based on the text recognition results, identifies the fields required for approval to obtain the recognition results of the fields required for approval.

[0156] Step S203: The rule reasoning module performs compliance verification on the recognition results of the fields required for approval;

[0157] Step S204: The approval path planning module performs approval path planning on the field identification results that meet the compliance requirements of the approval, and obtains a suitable approval path graph. The approval path graph is configured with a directed graph structure, where nodes represent approval links or responsible entities, and edges represent reachable paths or conditional branches.

[0158] Step S205: Based on the approval path diagram, the process scheduling module drives the execution of each node in the approval path.

[0159] In step S206, the result generation feedback module compiles the approval conclusions based on the execution status of the approval path diagram and pushes feedback to multiple terminals.

[0160] In one exemplary embodiment, for the data access module, the preset material formats include structured electronic forms, scanned documents, and PDF documents, and the preset material channels include web forms, WeChat mini programs, and government apps. The original materials submitted by users are converted into an internal standard data format after being encrypted and de-identified for sensitive fields including ID card numbers and corporate tax numbers, and are tracked in the audit process through a unique identifier.

[0161] In yet another exemplary embodiment, the automated government approval system further includes a training management module;

[0162] The method also includes:

[0163] The training management module manages the text recognition model used by the intelligent parsing module for text recognition, the named entity recognition model used by the intelligent parsing module for recognizing fields required for approval, the rule adaptation model used by the rule reasoning module for compliance verification, and the path planning strategy model used by the approval path planning module for approval path planning, including training, evaluation, and updating.

[0164] In yet another exemplary embodiment, for the named entity recognition model, the following loss function is used during training:

[0165] ,

[0166] in, This represents the total loss of the model. Used to measure the difference between the predicted result and the true label; n represents the length of the input sequence, expressed as the total number of characters or words in the text; C represents the number of categories, including names, places, organizations, and others; This indicates whether the true label of the i-th character or word belongs to the c-th class. If it does, the value is 1; otherwise, it is 0. This represents the probability that the model predicts the i-th character or word belongs to the c-th class, which is obtained through the softmax operation.

[0167] In yet another exemplary embodiment, for the path planning strategy model, a state space is constructed based on a graph neural network, and a suitable path structure is determined by combining path cost and strategy preference.

[0168] The path cost function is expressed by the following formula:

[0169] ,

[0170] in, This represents the total cost of path P; This represents the average processing time of the i-th stage; Indicates the weight of each stage; Indicates the total number of stages; This represents the path stability adjustment coefficient; This represents the total impact value of uncertain factors in the path.

[0171] In yet another exemplary embodiment, for the path planning strategy model, the following loss function is used during training, with the following expected path deviation as the optimization objective:

[0172] ,

[0173] in, Indicates the expected deviation of the path; Represents the actual path sequence vector; This represents the model's predicted path; N represents the total number of training samples.

[0174] In another exemplary embodiment, for the rule reasoning module, each rule consists of a logical expression, weight conditions, mandatory field validation, cross-field logical validation, and dependency binding. Compliance is judged through field comparison, and specifically involves qualification verification rules, regional applicability judgment rules, applicant authority judgment rules, and material completeness judgment rules.

[0175] The rule expression uses the following structure:

[0176] ,

[0177] in, Represents a set of rules; This represents the i-th rule; Indicates the total number of rules; Represents structured fields; This represents the condition for the rule function. A result of 1 indicates that the rule is satisfied, and a result of 0 indicates that the rule is not satisfied.

[0178] In yet another exemplary embodiment, the automated government approval system further includes a log auditing module;

[0179] The method also includes:

[0180] The log auditing module records a full-process audit log for each approval request, including data access time, processing steps, rule hit status, approver suggestions, system suggestions, final results, original material content, field extraction results, rule judgment path, approval node sequence, node result status, records of manual intervention behavior, and user feedback. It uses an immutable chain structure for storage, supporting the regulatory platform to read, replay, and trace the behavior.

[0181] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the AI-based automated government approval method described above can be found in, for example... Figure 1 The description of the AI-based automated government approval system in the corresponding embodiment will not be repeated here.

[0182] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0183] Therefore, this application provides a computer-readable storage medium storing a plurality of instructions that can be loaded by a processor to execute the present application. Figure 2 The steps of the AI-based automated government approval method in the corresponding embodiment can be found in the following example. Figure 2 The description of the AI-based automated government approval method in the corresponding embodiments will not be repeated here.

[0184] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0185] Because of the instructions stored in the computer-readable storage medium, the present application can be executed as described above. Figure 2 The steps of the AI-based automated government approval method in the corresponding embodiment can therefore achieve the results of this application. Figure 2 The beneficial effects that the AI-based automated government approval method can achieve in the corresponding embodiments are detailed in the preceding description and will not be repeated here.

[0186] The above provides a detailed description of the AI-based automated government approval system, AI-based automated government approval method, and computer-readable storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. An automated government approval system based on artificial intelligence, characterized in that, The automated government approval system includes a data access module, an intelligent parsing module, a rule reasoning module, an approval path planning module, a process scheduling module, and a result generation and feedback module. Each module operates independently and collaborates with an event-driven communication mechanism through a standard interface. The data access module is used to receive approval materials submitted by users through government service terminals under preset material formats and channels, and convert them into internal standard data formats. The intelligent parsing module is used to perform text recognition on the approval materials, and based on the text recognition results, to recognize the fields required for approval, and obtain the recognition results of the fields required for approval. The rule reasoning module is used to perform compliance verification on the identification results of the fields required for approval; The approval path planning module is used to plan the approval path based on the identification results of the fields required for the approval that meet the compliance requirements, and to obtain a suitable approval path graph. The approval path graph is configured in a directed graph structure, where nodes represent approval links or responsible entities, and edges represent reachable paths or conditional branches. The process scheduling module is used to drive the execution of each node in the approval path based on the approval path diagram. The result generation and feedback module is used to compile approval conclusions based on the execution status of the approval path diagram and push feedback to multiple terminals. For the path planning strategy model, a state space is constructed based on a graph neural network, and a suitable path structure is determined by combining path cost and strategy preference. The path cost function is expressed by the following formula: ， in, This represents the total cost of path P; This represents the average processing time of the i-th stage; Indicates the weight of each stage; Indicates the total number of stages; This represents the path stability adjustment coefficient; This represents the total impact value of uncertain factors in the path; For the path planning strategy model, the following path expectation deviation is used as the optimization objective during training: ， in, This indicates the expected deviation of the path; Represents the actual path sequence vector; This represents the model's predicted path; N represents the total number of training samples.

2. The automated government approval system according to claim 1, characterized in that, For the data access module, the preset material formats include structured electronic forms, scanned documents, and PDF documents, and the preset material channels include web forms, WeChat mini programs, and government apps. After the original materials submitted by the user are encrypted and desensitized for sensitive fields including ID card numbers and enterprise tax numbers, they are converted into the internal standard data format and tracked in the audit process through a unique identifier.

3. The automated government approval system according to claim 1, characterized in that, The automated government approval system also includes a training management module; The training management module is used to manage the text recognition model for text recognition by the intelligent parsing module, the named entity recognition model for recognition of fields required for approval by the intelligent parsing module, the rule adaptation model for compliance verification by the rule reasoning module, and the path planning strategy model for approval path planning by the approval path planning module, including training, evaluation, and updating.

4. The automated government approval system according to claim 3, characterized in that, For the named entity recognition model, the following loss function is used during training: ， in, This represents the total loss of the model. Used to measure the difference between the predicted result and the true label; n represents the length of the input sequence, expressed as the total number of characters or words in the text; C represents the category, including names, locations, and organizations; This indicates whether the true label of the i-th character or word belongs to the c-th category. If it does, the value is 1; otherwise, it is 0. This represents the probability that the model predicts the i-th character or word belongs to the c-th class, and is derived by the softmax operation.

5. The automated government approval system according to claim 1, characterized in that, For the rule reasoning module, each rule consists of a logical expression, weight conditions, mandatory field validation, cross-field logical validation, and dependency binding. It makes compliance judgments through field comparisons and specifically involves qualification verification rules, regional applicability judgment rules, applicant authority judgment rules, and material completeness judgment rules. The rule expression uses the following structure: ， in, Represents a set of rules; This represents the i-th rule; Indicates the total number of rules; Represents structured fields; This represents the condition for the rule function. A result of 1 indicates that the rule is satisfied, and a result of 0 indicates that the rule is not satisfied.

6. The automated government approval system according to claim 1, characterized in that, The automated government approval system also includes a log auditing module; The log auditing module is used to record the entire process of each approval request, including data access time, processing steps, rule hit status, approver suggestions, system suggestions, final results, original material content, field extraction results, rule judgment path, approval node order, node result status, records of manual intervention behavior, and user feedback. It is stored in an immutable chain structure, which supports the regulatory platform to read, replay, and trace the behavior.

7. An automated government approval method based on artificial intelligence, characterized in that, The method is applied to the AI-based automated government approval system according to any one of claims 1 to 6. The automated government approval system includes a data access module, an intelligent parsing module, a rule reasoning module, an approval path planning module, a process scheduling module, and a result generation and feedback module. Each module operates independently and collaborates with an event-driven communication mechanism through a standard interface. The method includes: The data access module receives approval materials submitted by users through government service terminals under preset material formats and channels, and converts them into internal standard data formats. The intelligent parsing module performs text recognition on the approval materials and, based on the text recognition results, identifies the fields required for approval, thus obtaining the recognition results of the fields required for approval. The rule-based reasoning module performs compliance verification on the recognition results of the fields required for approval; The approval path planning module performs approval path planning on the field identification results that meet the compliance requirements of the approval, and obtains a suitable approval path graph. The approval path graph is configured in a directed graph structure, where nodes represent approval links or responsible entities, and edges represent reachable paths or conditional branches. Based on the approval path diagram, the process scheduling module drives the execution of each node in the approval path; The results generation and feedback module compiles approval conclusions based on the execution status of the approval path diagram and pushes feedback to multiple terminals.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of instructions adapted for loading by a processor to perform the method of claim 7.

Images