Intelligent workflow generation method and device, equipment and medium

Abstract

The invention relates to the technical field of natural languages, can be applied to financial science and technology and medical health old-age service scenes, and discloses an intelligent workflow generation method, device, equipment and medium, and the method comprises the steps: obtaining a natural language inputted by a user, and analyzing the natural language to obtain a standardized service concept matched with the natural language; acquiring a technical component matched with the standardized business concept through a preset mapping relation table; context information matched with the standardized business concept is inquired through a preset knowledge base, and the context information comprises business rules; and generating a standardized workflow according to the standardized business concept, the technical component and the context information. The business intention of the user can be accurately identified, and the requirements of various fields are met.

Description

Technical Field

[0001] This invention relates to the field of natural language technology, and in particular to an intelligent workflow generation method, apparatus, device, and medium. Background Technology

[0002] As enterprises advance their digital transformation, low-code or no-code automated workflow systems have become core tools for business automation, enabling users to quickly create processes using natural language. These systems are widely used in scenarios such as order management, risk control and early warning, and medical record management. While these systems were initially designed with process creation efficiency and convenience in mind, they have revealed significant shortcomings in adapting to enterprise-level standardization and precision requirements, especially in sensitive areas such as fintech and healthcare, where they are incompatible with the logic of generating core business processes.

[0003] Existing systems cannot accurately parse business concepts from users' natural language input, relying solely on keyword or template matching. In the fintech sector, when faced with input triggering audits for transactions exceeding 50,000, the system can only recognize surface-level keywords to generate simple processes, failing to associate them with the business concepts required for high-risk transaction risk control audits. This could lead to compliance risks such as unauthorized access and leakage of sensitive transaction data. In the healthcare sector, when users input compliance verification for post-operative medication use, the system cannot recognize the corresponding business concepts, only generating scattered data queries and notification nodes, which deviates from healthcare compliance requirements.

[0004] The inventors realized that existing technologies rely on keyword and template matching, which cannot identify the core business intent behind natural language, leading to a disconnect between processes and the essence of business. Summary of the Invention

[0005] This invention provides an intelligent workflow generation method, apparatus, computer equipment, and medium to solve the technical problem that existing workflow generation systems cannot accurately identify business intent.

[0006] Firstly, an intelligent workflow generation method is provided, including: Obtain natural language input from the user and parse the natural language to obtain standardized business concepts that match the natural language; The technical components that match the standardized business concept are obtained by using a preset mapping table; By querying a preset knowledge base, contextual information that matches the standardized business concept is obtained, wherein the contextual information includes business rules; A standardized workflow is generated based on the standardized business concept, the technical components, and the context information.

[0007] Secondly, an intelligent workflow generation device is provided, comprising: The first acquisition module is used to acquire natural language input by the user and parse the natural language to obtain standardized business concepts that match the natural language. The second acquisition module is used to acquire technical components that match the standardized business concept through a preset mapping relationship table; The first query module is used to query context information that matches the standardized business concept through a preset knowledge base, wherein the context information includes business rules; The first generation module is used to generate a standardized workflow based on the standardized business concept, the technical components, and the context information.

[0008] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described intelligent workflow generation method.

[0009] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the above-described intelligent workflow generation method.

[0010] In the above-described intelligent workflow generation method, apparatus, computer equipment, and storage medium, the user's input natural language is acquired, and the natural language is parsed to obtain a standardized business concept matching the natural language; a preset mapping table is used to obtain technical components matching the standardized business concept; a preset knowledge base is used to query contextual information matching the standardized business concept, wherein the contextual information includes business rules; and a standardized workflow is generated based on the standardized business concept, the technical components, and the contextual information. In this invention, natural language can be parsed to obtain a standardized business concept, and technical components matching the standardized business concept can be obtained through a preset mapping table. Contextual information matching the standardized business concept can also be queried through a preset knowledge base. Finally, a standardized workflow is generated based on the standardized business concept, technical components, and contextual information, which can accurately identify the user's business intent and meet the requirements of various fields. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1This is a flowchart illustrating an intelligent workflow generation method according to an embodiment of the present invention; Figure 2 yes Figure 1 A schematic diagram of a specific implementation method for step S10; Figure 3 yes Figure 1 A flowchart illustrating another specific implementation of step S10; Figure 4 yes Figure 1 A schematic diagram of a specific implementation method for step S20; Figure 5 yes Figure 1 A schematic diagram of a specific implementation method for step S30; Figure 6 This is a schematic diagram of the first sub-process of the intelligent workflow generation method in one embodiment of the present invention; Figure 7 This is a schematic diagram of the second sub-process of the intelligent workflow generation method in one embodiment of the present invention; Figure 8 This is a block diagram of an intelligent workflow generation system according to an embodiment of the present invention; Figure 9 This is a schematic diagram of the structure of an intelligent workflow generation device in one embodiment of the present invention; Figure 10 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention; Figure 11 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] The intelligent workflow generation method provided in this invention can be applied to various personal computers, laptops, smartphones, and tablets. It involves: acquiring natural language input from the user and parsing the natural language to obtain standardized business concepts matching the natural language; obtaining technical components matching the standardized business concepts through a preset mapping table; querying contextual information matching the standardized business concepts through a preset knowledge base, wherein the contextual information includes business rules; and generating a standardized workflow based on the standardized business concepts, the technical components, and the contextual information. This method can accurately identify the user's business intent and meet the needs of various business domains.

[0015] The intelligent workflow generation system and method based on multi-layer semantic mapping provided by this invention can be applied to business areas with extremely high requirements for process compliance, traceability, and standardization, such as fintech and healthcare. In the fintech field, business processes (such as anti-fraud monitoring, credit approval, transaction compliance verification, and customer identity verification) must strictly comply with regulatory policies and internal risk control rules, and the processes must be fully auditable and reusable across business lines. The weak semantic understanding and black-box problem of traditional low-code platforms are particularly prominent in this field. With the intelligent workflow generation system of this invention, users only need to input natural language requirements, such as triggering anti-fraud audit and freezing the account when a single transaction exceeds 50,000 yuan and is logged in from a different location, and simultaneously generating a risk control audit report. The system can achieve precise adaptation.

[0016] In the healthcare field, core processes (such as postoperative complication monitoring, medication dispensing verification, medical record review and transfer, and adherence to surgical procedures) must strictly follow clinical pathway guidelines and medical quality management standards. The accuracy and traceability of these processes directly impact medical safety. Traditional systems lack the ability to standardize medical business concepts and dynamically integrate clinical knowledge, easily leading to processes that deviate from clinical practice. The intelligent workflow generation system of this invention allows medical staff or administrators to input natural language requirements, such as a postoperative patient's temperature exceeding 38.5°C for two consecutive hours. This automatically sends an alert to the attending physician and nursing station, simultaneously recording the patient's progress log and linking it to infection control procedures. The system achieves precise adaptation across the entire process.

[0017] Please see Figure 1 As shown, Figure 1 A flowchart illustrating the intelligent workflow generation method provided in this embodiment of the invention includes the following steps: S10: Obtain the natural language input by the user and parse the natural language to obtain a standardized business concept that matches the natural language.

[0018] Standardized business concepts are standardized semantic identifiers pre-defined based on specific business scenarios, industry standards, and enterprise needs. They possess unified naming rules, defined boundaries, and relationships, such as in the fintech and healthcare business domains. In the fintech domain, standardized business concepts may include, but are not limited to, credit-risk-evaluation, insurance-claim-audit, anti-fraud-monitoring, and customer-credit-query. In the healthcare domain, standardized business concepts may include, but are not limited to, patient-vital-signs-abnormal, chronic-disease-follow-up, medical-insurance-reimbursement-audit, clinical-pathway-execution, and drug-allergy-alert.

[0019] In some embodiments, such as Figure 2 As shown, in step S10, that is, parsing the natural language to obtain standardized business concepts that match the natural language, the specific steps include the following: S11: Preprocess the natural language to obtain structured statements; S12: Extract the core semantic elements of the structured statement and perform semantic reasoning on the core semantic elements to abstract the business intent corresponding to the natural language; S13: Confirm the business concept corresponding to the business intent through a preset business concept table to obtain the standardized business concept.

[0020] The preprocessing process can incorporate domain-specific language processing rules, specifically including: first, performing word segmentation, part-of-speech tagging, and stop word removal on the natural language; second, identifying core components of the sentence through syntactic analysis, such as subject, indicator, condition, action, and object; and finally, removing redundant information and completing omitted components to form a structured sentence. For example, in the fintech field, if a user inputs that a loan applicant's debt ratio exceeds 60% to trigger manual review, the preprocessed result would be: Subject: Loan applicant; Key indicator: Debt ratio; Judgment criteria: greater than 60%; Action executed: Trigger manual review; Related object: Auditing department.

[0021] In the healthcare field, when a user inputs the message "When a diabetic patient's fasting blood glucose level exceeds 7.0 mmol / L for three consecutive days, a follow-up reminder will be sent to the attending physician," the preprocessed result is: Subject: Diabetic patients; Key indicator: Fasting blood glucose; Judgment criteria: >7.0 mmol / L for 3 consecutive days; Action performed: Send a follow-up reminder; Recipient: Attending physician; Related requirements: Follow-up records must be archived.

[0022] Then, the core semantic elements of the structured statements are extracted, and deep semantic reasoning is performed in conjunction with domain knowledge to abstract the business intent corresponding to the natural language. The extraction of core semantic elements focuses on key dimensions such as subject type, core indicators, constraints, execution actions, and associated objects. The semantic reasoning process can employ a hybrid mechanism of large-model semantic vector matching and domain ontology support. The domain ontology contains knowledge such as hierarchical relationships of business concepts, attribute definitions, and scenario association rules in the fintech / healthcare domains. For example, in the fintech domain, core semantic elements such as loan application, excessive debt ratio, and manual review are extracted from structured statements. By matching these with the concept vector of credit risk assessment in the fintech ontology and combining the business logic of excessive debt → risk warning → manual intervention, the business intent of credit risk assessment triggering manual intervention is ultimately abstracted. In the healthcare domain, core semantic elements such as diabetic patients, persistently elevated blood sugar, and doctor follow-up reminders are extracted. By matching these with the concept vector of chronic disease follow-up management in the healthcare ontology and combining the clinical logic of abnormal chronic disease indicators → clinical intervention → follow-up reminders, the business intent of follow-up intervention for abnormal chronic disease indicators is abstracted.

[0023] Finally, a pre-defined business concept table and a semantic similarity matching algorithm are used to confirm the business concepts corresponding to the business intentions, thus obtaining standardized business concepts. The pre-defined business concept table is a structured data table containing information such as business concept ID, name, definition, core attributes, applicable scenarios, and semantic vectors. Semantic similarity matching employs cosine similarity calculation and a threshold filtering mechanism (domain-adaptive similarity thresholds, such as 0.85 for the financial domain and 0.88 for the medical domain) to ensure matching accuracy. For example, in the fintech domain, the business intention of triggering manual intervention for credit risk assessment is semantically matched with "credit-risk-evaluation" in the pre-defined business concept table, achieving a similarity of 0.92, confirming a match for the standardized business concept. In the healthcare domain, the business intention of intervening in the follow-up of abnormal indicators for patients with chronic diseases is matched with "chronic-disease-follow-up," achieving a similarity of 0.91, confirming the standardized business concept.

[0024] In some embodiments, such as Figure 3 As shown, in step S10, which involves parsing the natural language to obtain standardized business concepts that match the natural language, the steps further include the following: S14: Determine whether the standardized business concept is compatible with the business rules in the preset knowledge base; S15: If the standardized business concept is compatible with the business rules in the preset knowledge base, then proceed to the step of obtaining the technical components that match the standardized business concept through the preset mapping relationship table; S16: If the standardized business concept does not match the business rules in the preset knowledge base, a rule conflict will be prompted and a rematch will be performed after obtaining user confirmation.

[0025] The pre-built knowledge base is a multi-source integrated knowledge system built for the fields of fintech and healthcare. It includes a structured business rule base (such as the "Credit Risk Control Management Measures" and "Payment and Settlement Business Specifications" of financial institutions, and the "Clinical Diagnosis and Treatment Guidelines" and "Detailed Rules for Medical Insurance Reimbursement Policies" of hospitals), an unstructured document base (such as policy interpretation documents and operation manuals), a historical process case base, and a threshold standard base. The adaptation judgment dimensions include: the applicable scope of business concepts matches the rules, the core thresholds are consistent with the rule requirements, the related processes are compatible with the rule constraints, and the timeliness and validity of the rules (such as whether it is the latest version of the rules). For example, in the fintech sector, determining whether a debt ratio >60% triggers manual review for credit-risk-evaluation complies with the differentiated rule in the "Credit Risk Control Management Measures V3.0" that allows for a relaxed debt ratio threshold of up to 70% for high-quality customers, and whether it aligns with the risk control requirements following the current LPR interest rate adjustment; in the healthcare sector, determining whether a fasting blood glucose level >7.0 mmol / L for chronic disease follow-up complies with the age-appropriate rule in the "Guidelines for the Diagnosis and Treatment of Diabetes (2023 Edition)" that allows for an adjustment of the threshold to 7.5 mmol / L for patients aged 65 and above, and whether it is consistent with the chronic disease management guidelines of the patient's department.

[0026] If the standardized business concept matches the business rules in the preset knowledge base, proceed to step S20 to continue acquiring the matching technical components.

[0027] If the standardized business concept is incompatible with the business rules in the preset knowledge base, such as threshold conflicts, mismatched scope of application, or expired rules, the system will automatically generate a rule conflict warning message, clarifying the conflict type, source of the conflicting rule, and suggested adjustment plan, and display it to the user through a visual interface. After the user confirms the adjustment direction, such as adopting the system suggestion, customizing the adjustment threshold, and selecting the applicable rule version, steps S11-S13 will be executed again to complete the re-matching of the standardized business concept. For example, in the fintech field, the system may indicate a conflict between the current debt ratio threshold of 60% and the 70% debt ratio threshold for high-quality customers in the "Credit Risk Control Management Measures V3.0," suggesting that the threshold be adjusted according to the customer rating. The system may also indicate a conflict between the current blood glucose follow-up threshold of 7.0 mmol / L and the threshold of 7.5 mmol / L for elderly patients over 65 years old in the "Guidelines for the Diagnosis and Treatment of Diabetes (2023 Edition)," suggesting that the threshold be automatically adjusted to 7.5 mmol / L.

[0028] S20: Obtain the technical components that match the standardized business concept through a preset mapping relationship table.

[0029] The pre-defined mapping table serves as the core bridge connecting standardized business concepts and technical implementations. Its structure has been optimized for domain adaptation and can include fields such as business concept ID, business concept name, technical component type, technical component ID, interface parameter requirements, adaptation scenarios, version number, list of associated components, availability status, and update time. Technical components are customized based on the node capabilities of low-code or no-code platforms, taking into account the system integration needs of the fintech and healthcare sectors. They cover core functionalities such as data querying, conditional judgment, notification push, approval workflow, log recording, and system integration.

[0030] In some embodiments, such as Figure 4 As shown, in step S20, that is, obtaining the technical components that match the standardized business concept through a preset mapping relationship table, the specific steps include the following: S21: Based on the standardized business concept, query the preset mapping relationship table to obtain the technical components to be verified corresponding to the standardized business concept; S22: Verify the technical component to be verified to confirm whether the technical component to be verified is usable; S23: If the technical component to be verified is available, then the technical component to be verified shall be used as the technical component; S24: If the technical component to be verified is unavailable, the technical component to be verified is replaced with a candidate technical component that is functionally equivalent to the technical component to be verified, and the candidate technical component is used as the technical component after the candidate technical component passes the verification.

[0031] Based on the ID and name of the standardized business concept, a preset mapping relationship table is queried to filter out all technical components directly related to the business concept, forming a set of technical components to be verified. For example, in the fintech field, the set of technical components to be verified for credit-risk-evaluation includes n8n-node-conditional, n8n-node-credit-check, n8n-node-approval, n8n-node-notify, n8n-node-logging, and n8n-node-db-insert. Similarly, in the healthcare field, the set of technical components to be verified for chronic-disease-follow-up includes n8n-node-conditional, n8n-node-emr-query, n8n-node-doctor-notify, n8n-node-medical-record, n8n-node-follow-up-form, and n8n-node-logging.

[0032] Each component in the set of technical components to be verified undergoes multi-dimensional verification to confirm its availability. Verification dimensions may include: component availability status (whether it is running normally and not disabled), interface connectivity (whether the interface with related systems is smooth, such as credit reporting systems and electronic medical record systems), permission compatibility (whether the component has the permission to access the target system), performance threshold (whether the component's concurrent processing capability meets business requirements), and version compatibility (whether the component version is compatible with the workflow system version). The verification process is implemented through automated interface testing, permission verification scripts, and performance probing tools, generating a verification report for each component.

[0033] If all verification dimensions of the technical component to be verified pass, the component will be included in the final set of technical components as the basis for subsequent workflow generation.

[0034] If any verification dimension of the technical component to be verified fails, the alternative component replacement mechanism is activated. First, based on the principles of functional equivalence, interface compatibility, and similar performance, candidate components with the same function as the unavailable component are selected from the alternative component library of the preset mapping relationship table. Among them, candidate components can be sorted by functional similarity, historical success rate, and maintenance cost. Second, the multi-dimensional verification of S22 is repeated on the candidate components. Finally, the candidate components that pass the verification are added to the technical component set. If no usable candidate components are found, a component unavailability alarm is pushed to the user, and suggestions for manual intervention are provided. For example, in the fintech field, when n8n-node-credit-check (credit inquiry node) is unavailable due to interface maintenance, n8n-node-third-party-credit (third-party credit inquiry node) is selected from the alternative component library. After verifying interface connectivity and permission adaptation, it is used as a replacement component. In the healthcare field, when n8n-node-emr-query (electronic medical record query node) is unavailable due to insufficient permissions, it is replaced by n8n-node-his-query (hospital information system query node). After passing the verification, it is included in the technical component set.

[0035] S30: Query context information that matches the standardized business concept through a preset knowledge base, wherein the context information includes business rules.

[0036] The pre-built knowledge base can adopt a hybrid structured and unstructured storage architecture, supporting dynamic updates and intelligent retrieval. Its core components include: ① A structured business rule base: storing explicit rule clauses, threshold standards, and execution processes, stored in structured formats such as tables and databases for easy automated retrieval and matching; ② An unstructured document base: storing unstructured content such as policy documents, operation manuals, and clinical guidelines, which are transformed into searchable knowledge units through text parsing and semantic annotation; ③ A historical process case base: storing past workflow cases implemented by the enterprise, including information such as business scenarios, technical components used, parameter configurations, and execution results; ④ A threshold standard base: storing core indicator thresholds in the field, such as debt ratio thresholds and wealth management redemption amount thresholds in the financial field, and blood glucose / blood pressure thresholds and follow-up timeliness thresholds in the medical field.

[0037] In some embodiments, such as Figure 5 As shown, in step S30, which involves querying contextual information matching the standardized business concept through a preset knowledge base, the specific steps include: S31: Extract business rules and rule thresholds that match the standardized business concept from the preset knowledge base to obtain the context information to be verified; S32: Perform a consistency check on the context information to be verified to obtain the context information.

[0038] Based on the ID, name, and core attributes of standardized business concepts, precise searches are performed from multiple sources within a pre-defined knowledge base to extract information directly related to the business concept, such as business rules, rule thresholds, association requirements, and execution specifications, forming a set of contextual information to be verified. The search process employs a fusion mechanism of vector retrieval and keyword retrieval to ensure the comprehensiveness and accuracy of the search. For example, in the fintech field, credit-risk-evaluation extracts rules from a structured business rule base, requiring manual review if the debt ratio exceeds 60%, with a review period of no more than 2 business days. The review results must be synchronized to the customer's credit report. Thresholds of 60% (ordinary customers) and 70% (high-quality customers) are extracted from a threshold standard library. Related requirements include checking the customer's credit records for the past 6 months and verifying the authenticity of income certificates. In the healthcare field, chronic-disease-follow-up extracts rules from a structured business rule base, requiring follow-up within 48 hours if diabetic patients' fasting blood glucose exceeds 7.0 mmol / L for 3 consecutive days. Follow-up content includes medication adjustment suggestions, dietary guidance, and exercise suggestions. Thresholds of 7.0 mmol / L (ordinary patients) and 7.5 mmol / L (elderly patients over 65 years old) are extracted from a threshold standard library. Related requirements include that follow-up records must be entered into electronic medical records and synchronized to the chronic disease management system.

[0039] The set of context information to be verified undergoes multi-dimensional consistency verification, eliminating conflicting information and supplementing missing information to ultimately obtain accurate and complete context information. The core dimensions of consistency verification include: ① Rule timeliness verification: confirming whether the extracted rules are the latest valid versions, eliminating expired or obsolete rules; ② Rule conflict detection: checking for logical conflicts between rules from different sources, resolving conflicts according to rule priority (e.g., regulatory policies > internal company rules > departmental rules); ③ Threshold rationality verification: verifying the rationality of thresholds in conjunction with business scenarios and applicable objects (e.g., customer ratings, patient ages), making dynamic adjustments as necessary; ④ Information completeness supplementation: if the extracted context information is missing, supplementing relevant information from supplementary documents in the knowledge base or historical cases. For example, in the fintech field, it was found that the debt ratio threshold of 60% for ordinary customers and 70% for high-quality customers did not conflict. The missing information that manual review requires signature confirmation from two or more reviewers in the risk control department was supplemented. In the healthcare field, it was found that the 7.0 mmol / L blood glucose threshold is applicable to ordinary patients. However, this patient is an elderly patient. The threshold was automatically adjusted to 7.5 mmol / L. The missing requirement that follow-up visits should inquire about the patient's recent medication adherence was supplemented.

[0040] S40: Generate a standardized workflow based on the standardized business concept, the technical components, and the context information.

[0041] The workflow generation process is a fully automated process based on component concatenation, parameter population, logic definition, and exception handling. Specifically, it includes: ① Component concatenation: Defining the execution order of technical components based on the execution logic of business concepts and contextual requirements (e.g., condition judgment → data query → approval node → notification push → log recording) to ensure that component concatenation conforms to the natural logic of the business process; ② Parameter population: Automatically populating the threshold standards, interface parameters, and associated objects from the contextual information into the corresponding technical components (e.g., populating the debt ratio > 60% into the condition judgment node, and the attending physician's mobile phone number into the notification push node); ③ Logic definition: Clarifying the branching logic between components (e.g., the satisfying and non-satisfied branches of the condition judgment node), parallel / serial execution relationships, and waiting time thresholds; ④ Exception handling: Adding exception handling branches to key technical components, such as a retry mechanism when a component call fails (setting the number of retries and retry interval), a degradation processing flow (e.g., automatically switching to email notification when SMS sending fails in the notification push node), and an exception alarm mechanism (e.g., sending alarm information to the administrator when approval times out).

[0042] The generated standardized workflow has clear execution logic, complete parameter configuration, and standardized node definitions. For example, the final form of the credit risk assessment workflow in the fintech field is: condition judgment node (debt ratio > 60%) → credit inquiry node (inquire about the customer's credit records for the past 6 months) → manual approval node (signature by 2 reviewers in the risk control department, with a maximum waiting time of 2 working days) → notification push node (send the approval results to the customer and business manager) → data entry node (synchronize the approval results to the credit management system) → log recording node (record the entire process execution).

[0043] The final form of the chronic disease follow-up reminder workflow in the medical and health field is as follows: Condition judgment node (fasting blood glucose > 7.5 mmol / L for 3 consecutive days) → Electronic medical record query node (query patient's medical history and medication records) → Doctor notification node (send follow-up reminder to the attending physician, including the patient's basic information and abnormal indicators) → Follow-up form generation node (generate standardized chronic disease follow-up form) → Medical record update node (enter follow-up records into electronic medical records) → Log recording node (record process execution time, participants, and execution results).

[0044] In some embodiments, such as Figure 6 As shown, the intelligent workflow generation method further includes the following steps: S50: Embed audit information in the standardized workflow, wherein the audit information includes the source information of the natural language and the context information.

[0045] Audit information is the core support for ensuring the interpretability and traceability of workflows. It can include basic fields and extended fields: Basic fields can include the user's original natural language input, the workflow generator (generator version), the generation time, and the referenced policy / guideline sources (including document name and chapter number); Extended fields can include workflow modification records (such as the time of subsequent modifications, the modifier, and the content of the modifications), approval node records (such as the reviewer, review comments, and review time), compliance verification results (such as whether it complies with regulatory policies and the compliance verification time), and process execution trajectory related fields (such as process execution ID and associated business order number). Audit information is embedded in the workflow's metadata in a structured form for easy subsequent querying, retrieval, and export.

[0046] In some embodiments, to reduce redundant development and lower enterprise costs, this invention designs a complete workflow reuse mechanism. Through business scenario similarity matching and personalized configuration, it enables the rapid reuse of existing workflows, such as... Figure 7 As shown, the specific steps are as follows: S60: If a workflow to be generated is detected, the business scenario of the workflow to be generated is obtained to obtain a first business scenario; S70: Confirm the business scenario for which a workflow has been generated to obtain multiple second business scenarios; S80: If the first business scenario is similar to any one of the multiple second business scenarios, then the second business scenario similar to the first business scenario shall be taken as the target business scenario. S90: Obtain the workflow of the target business scenario to obtain the workflow to be configured; S100: Configure the workflow to be configured according to the business requirements of the first business scenario to obtain the workflow to be generated.

[0047] When the system detects a new workflow generation requirement from a user (i.e., a workflow to be generated), it first obtains the core business scenario information of the workflow to be generated through natural language parsing and business intent recognition, forming the first business scenario. The descriptive dimensions of the first business scenario include business intent, core indicators, constraints, execution actions, applicable objects, and related systems. For example, in the fintech field, the first business scenario is that a car loan applicant's debt ratio exceeds 60%, triggering manual review, and the review result is synchronized to the car loan management system; in the healthcare field, the first business scenario is that a hypertensive patient's systolic blood pressure exceeds 160 mmHg for two consecutive days, sending a follow-up reminder, and the follow-up content must include suggestions for adjusting antihypertensive medication.

[0048] The system queries the historical database of generated workflows, extracts business scenario information for all implemented workflows, and forms multiple second business scenarios. Each second business scenario includes descriptive dimensions consistent with the first business scenario, facilitating subsequent similarity comparisons. For example, in the fintech field, second business scenarios include triggering manual review when a mortgage applicant's debt ratio exceeds 60%, with the review results synchronized to the mortgage management system; and triggering manual review when a business loan applicant's debt ratio exceeds 60%, with the review results synchronized to the business loan management system. In the healthcare field, second business scenarios include sending follow-up reminders when a diabetic patient's fasting blood glucose exceeds 7.5 mmol / L for three consecutive days, with follow-up content including suggestions for adjusting hypoglycemic medication; and sending follow-up reminders when a hyperlipidemic patient's triglyceride level exceeds 2.3 mmol / L for two consecutive days, with follow-up content including suggestions for adjusting lipid-lowering medication.

[0049] A multi-dimensional similarity calculation model is employed to determine the similarity between a first business scenario and multiple second business scenarios. The similarity calculation dimensions include business intent similarity (e.g., the intent similarity between auto loan approval and mortgage loan approval), core indicator similarity (e.g., the similarity between debt ratio indicators), action type similarity (e.g., the action similarity between manual approvals), applicable object similarity (e.g., the object similarity between auto loan customers and mortgage customers), and associated system similarity (e.g., the system similarity between auto loan management systems and mortgage loan management systems). Each dimension is assigned a corresponding weight (e.g., business intent weight 0.4, core indicator weight 0.3, action type weight 0.15, applicable object weight 0.1, associated system weight 0.05), and a weighted summation is used to calculate the overall similarity. If the overall similarity exceeds a preset threshold (e.g., 0.8), the first business scenario is determined to be similar to the second business scenario, and the second business scenario is selected as the target business scenario. For example, in the fintech field, the overall similarity between the debt ratio review of auto loan applicants and the debt ratio review of mortgage loan applicants is 0.89, which is considered similar, and the debt ratio review of mortgage loan applicants is taken as the target business scenario; in the healthcare field, the overall similarity between the follow-up of abnormal systolic blood pressure in hypertensive patients and the follow-up of abnormal blood glucose in diabetic patients is 0.83, which is considered similar, and the follow-up of abnormal blood glucose in diabetic patients is taken as the target business scenario.

[0050] Based on the identification information of the target business scenario, the corresponding workflow is retrieved from the historical database as the workflow to be configured. The workflow to be configured contains complete component connection logic, parameter configuration, exception handling mechanism and audit information, and can be used directly as a basic template.

[0051] Based on the personalized business requirements of the first business scenario, the workflow to be configured is specifically configured to generate the final form of the workflow. Configuration operations include: ① Parameter replacement: Replacing parameters that do not match the first business scenario (e.g., replacing the mortgage management system interface with the auto loan management system interface, and replacing blood glucose indicators with systolic blood pressure indicators); ② Threshold adjustment: Adjusting core thresholds according to the requirements of the first business scenario (e.g., adjusting fasting blood glucose 7.5 mmol / L to systolic blood pressure 160 mmHg); ③ Node addition / deletion: Adding or deleting some technical components according to business needs (e.g., adding a vehicle information query node for auto loan approval, and adding a blood pressure monitoring frequency suggestion node for hypertension follow-up); ④ Related system switching: Switching related system interfaces to adapt to the first business scenario (e.g., switching the mortgage credit query interface to the auto loan credit query interface); ⑤ Detail optimization: Adjusting details such as the execution order, waiting time, and notification content of components. For example, when configuring a mortgage loan approval workflow in the fintech field, the mortgage loan credit inquiry node is replaced with the auto loan credit inquiry node, and the associated system is switched from the mortgage loan management system to the auto loan management system, while the rest of the logic remains unchanged; when configuring a diabetes follow-up workflow in the healthcare field, the blood glucose test indicator is replaced with the systolic blood pressure test indicator, the threshold is adjusted to 160 mmHg, and an antihypertensive medication adjustment suggestion field is added to the follow-up form to complete the personalized configuration.

[0052] like Figure 8 As shown, the intelligent workflow generation method provided by this invention can be applied to an intelligent workflow generation system. This system can be flexibly configured on any computer device such as a server or cloud computing platform, and may include the following modules: Intent recognition module The intent recognition module serves as the semantic entry point of the system, responsible for receiving natural language input from users (supporting multiple methods such as text input and speech-to-text input). Through domain-adapted natural language processing technology and semantic reasoning mechanisms, it completes the transformation from natural language to standardized business concepts, while also possessing enhanced functions such as fuzzy input processing, multi-turn dialogue interaction, and semantic error correction.

[0053] In the fintech field, the intent recognition module can be equipped with a built-in financial-specific word segmentation dictionary, terminology library, and ontology library. It supports natural language parsing for subdivided scenarios such as credit, payment, wealth management, and insurance. It can accurately identify professional terms in financial business (such as debt ratio, LPR interest rate, and anti-fraud monitoring) and complex business intents (such as freezing funds and sending SMS notifications when the redemption amount of wealth management products exceeds 50,000 yuan, and updating the customer's asset account simultaneously), ensuring the accuracy of business intent recognition.

[0054] Knowledge base module The knowledge base module serves as the system's knowledge support center, responsible for storing, managing, and retrieving multi-source knowledge in the fintech / healthcare fields. It provides services such as dynamic rule matching, consistency verification, and knowledge updates to ensure that the workflow generation process has sufficient knowledge support.

[0055] For the fintech sector, the knowledge base module can integrate structured business rule bases (such as the "Credit Risk Control Management Measures," "Payment and Settlement Business Specifications," and "Wealth Management Product Sales Management Measures"), unstructured document bases (such as regulatory policy interpretations and operation manuals), historical process case bases (such as past credit review processes and anti-fraud monitoring processes), and threshold standard bases (such as debt ratio thresholds for various types of customers and wealth management redemption amount thresholds). It supports dynamic updates of knowledge, including automatic synchronization of the latest policies issued by regulatory authorities (through interface integration with the regulatory policy release platform), updates of internal rules entered through manual review, and threshold optimization suggestions based on historical process execution data. It has knowledge version management and rollback functions, allowing users to query knowledge content from any historical version and supporting version rollback after accidental operations.

[0056] Business Concept Module The Business Concept module serves as the "central management hub" for domain business concepts. It is responsible for the creation, maintenance, querying, and association management of standardized business concepts, and maintains the mapping relationship between "business concepts and technical components," providing support for the precise bridging of semantics and technology.

[0057] In the fintech field, the business concept module can manage standardized financial business concepts such as credit-risk-evaluation and payment-reconciliation-abnormal, maintaining the core attributes (e.g., definition, applicable scenarios, core indicators), relationships (e.g., parent-child concepts, related concepts, such as credit risk assessment being a sub-concept of anti-fraud monitoring), and lifecycle status (creation, activation, deactivation, archiving) of each concept; it maintains a many-to-many mapping relationship between business concepts and technical components, where one business concept can correspond to multiple technical components (e.g., credit risk assessment corresponds to condition judgment, credit inquiry, manual approval, etc.), and one technical component can adapt to multiple business concepts (e.g., the log recording component adapts to all financial business concepts); it supports dynamic adjustment of mapping relationships, allowing the addition, modification, or deletion of mapping relationships according to business needs, with adjustments synchronized to all modules of the system in real time.

[0058] Generate module As the core engine of the system, the generation module is responsible for receiving standardized business concepts output by the intent recognition module, contextual information provided by the knowledge base module, and technical component mapping relationships provided by the business concept module. It automatically generates standardized workflows, while embedding audit information and supporting workflow reuse configurations and process optimization suggestions.

[0059] For the fintech sector, the generation module can generate executable workflows that meet financial compliance requirements, ensuring that the process includes complete compliance nodes (such as approvals, filings, and log records) and audit information (including regulatory policy references); it supports cross-product line workflow reuse (such as the reuse of risk control processes for mortgage loans, auto loans, and business loans), and can quickly generate new workflows through scenario similarity matching and personalized configuration; it also has a process optimization suggestion function, which recommends process optimization solutions to users based on historical execution data (such as process execution time, approval pass rate, and anomaly rate) (such as simplifying redundant approval nodes and adjusting thresholds to improve review efficiency).

[0060] Technical support module The technical support module is responsible for providing support services such as management, verification, interface adaptation, and monitoring of technical components to ensure the stable execution and flexible expansion of the workflow.

[0061] For the fintech sector, the technical support module can support interface adaptation with the core systems of financial institutions, including credit reporting systems, payment and settlement systems, credit management systems, customer relationship management systems (CRM), and compliance and regulatory reporting systems. It provides standardized interface call schemes and permission management mechanisms; it has technical component monitoring functions to monitor the operational status of components in real time (such as interface connectivity, response time, and call success rate). When a component malfunctions (such as interface timeout or call failure), it automatically triggers alarms and initiates a backup component replacement mechanism; it supports the expansion and upgrading of technical components, and can quickly integrate new technical components according to new financial business needs (such as adding digital currency payment reconciliation functions) without modifying the core system architecture.

[0062] As can be seen, in the above solution, natural language can be parsed to obtain standardized business concepts, and technical components that match the standardized business concepts can be obtained through a preset mapping table. Contextual information that matches the standardized business concepts can also be queried through a preset knowledge base. Finally, a standardized workflow is generated based on the standardized business concepts, technical components, and contextual information. This approach can accurately identify the user's business intent and meet the requirements of various fields.

[0063] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0064] In one embodiment, an intelligent workflow generation device is provided, which corresponds one-to-one with the intelligent workflow generation method in the above embodiments. For example... Figure 9As shown, the intelligent workflow generation device includes a first acquisition module 101, a second acquisition module 102, a first query module 103, and a first generation module 104. Detailed descriptions of each functional module are as follows: The first acquisition module 101 is used to acquire natural language input by the user and parse the natural language to obtain a standardized business concept that matches the natural language. The second acquisition module 102 is used to acquire technical components that match the standardized business concept through a preset mapping relationship table; The first query module 103 is used to query context information that matches the standardized business concept through a preset knowledge base, wherein the context information includes business rules; The first generation module 104 is used to generate a standardized workflow based on the standardized business concept, the technical components, and the context information.

[0065] In one embodiment, the first acquisition module 101 is specifically used for: The natural language is preprocessed to obtain structured statements; Extract the core semantic elements of the structured statement and perform semantic reasoning on the core semantic elements to abstract the business intent corresponding to the natural language; The standardized business concept is obtained by confirming the business concept corresponding to the business intent through a preset business concept table.

[0066] In one embodiment, the first acquisition module 101 is further configured to: Determine whether the standardized business concept is compatible with the business rules in the preset knowledge base; If the standardized business concept matches the business rules in the preset knowledge base, then proceed to the step of obtaining the technical components that match the standardized business concept through the preset mapping relationship table; If the standardized business concept does not match the business rules in the preset knowledge base, a rule conflict will be indicated, and a rematch will be performed after obtaining user confirmation.

[0067] In one embodiment, the second acquisition module 102 is specifically used for: Based on the standardized business concept, query the preset mapping relationship table to obtain the technical components to be verified corresponding to the standardized business concept; The technical component to be verified is verified to confirm whether it is usable; If the technical component to be verified is available, then the technical component to be verified shall be used as the technical component. If the technical component to be verified is unavailable, it is replaced with an alternative technical component that is functionally equivalent to the technical component to be verified, and the alternative technical component is used as the technical component after the alternative technical component passes the verification.

[0068] In one embodiment, the first query module 103 is specifically used for: Based on the standardized business concept, extract business rules and rule thresholds that match the standardized business concept from the preset knowledge base to obtain the context information to be verified; The context information to be verified is obtained by performing a consistency check on the context information to be verified.

[0069] In one embodiment, the intelligent workflow generation device further includes: An insertion module is used to embed audit information into the standardized workflow, wherein the audit information includes source information of the natural language and the context information.

[0070] In one embodiment, the intelligent workflow generation device further includes: The detection module is used to obtain the business scenario of the workflow to be generated if a workflow to be generated is detected, thereby obtaining a first business scenario. The confirmation module is used to confirm the business scenario for which a workflow has been generated in order to obtain multiple second business scenarios. The setting module is used to select the second business scenario that is similar to the first business scenario as the target business scenario if the first business scenario is similar to any one of the multiple second business scenarios. The third acquisition module is used to acquire the workflow of the target business scenario to obtain the workflow to be configured. The configuration module is used to configure the workflow to be configured according to the business requirements of the first business scenario to obtain the workflow to be generated.

[0071] This invention provides an intelligent workflow generation device that can parse natural language to obtain standardized business concepts, acquire technical components that match the standardized business concepts through a preset mapping table, and query contextual information that matches the standardized business concepts through a preset knowledge base. Finally, it generates a standardized workflow based on the standardized business concepts, technical components, and contextual information, which can accurately identify the user's business intent and meet the requirements of various fields.

[0072] Specific limitations regarding the intelligent workflow generation device can be found in the limitations of the intelligent workflow generation method described above, and will not be repeated here. Each module in the aforementioned intelligent workflow generation device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0073] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 10 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a server-side intelligent workflow generation method.

[0074] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 11 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a client-side intelligent workflow generation method. In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Obtain natural language input from the user and parse the natural language to obtain standardized business concepts that match the natural language; The technical components that match the standardized business concept are obtained by using a preset mapping table; By querying a preset knowledge base, contextual information that matches the standardized business concept is obtained, wherein the contextual information includes business rules; A standardized workflow is generated based on the standardized business concept, the technical components, and the context information.

[0075] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Obtain natural language input from the user and parse the natural language to obtain standardized business concepts that match the natural language; The technical components that match the standardized business concept are obtained by using a preset mapping table; By querying a preset knowledge base, contextual information that matches the standardized business concept is obtained, wherein the contextual information includes business rules; A standardized workflow is generated based on the standardized business concept, the technical components, and the context information.

[0076] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0077] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0078] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0079] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for generating intelligent workflows, characterized in that, include: Obtain natural language input from the user and parse the natural language to obtain standardized business concepts that match the natural language; The technical components that match the standardized business concept are obtained by using a preset mapping table; By querying a preset knowledge base, contextual information that matches the standardized business concept is obtained, wherein the contextual information includes business rules; A standardized workflow is generated based on the standardized business concept, the technical components, and the context information.

2. The method according to claim 1, characterized in that, The process of parsing the natural language to obtain standardized business concepts that match the natural language includes: The natural language is preprocessed to obtain structured statements; Extract the core semantic elements of the structured statement and perform semantic reasoning on the core semantic elements to abstract the business intent corresponding to the natural language; The standardized business concept is obtained by confirming the business concept corresponding to the business intent through a preset business concept table.

3. The method according to claim 2, characterized in that, After obtaining the standardized business concept by confirming the business concept corresponding to the business intent through a preset business concept table, the process further includes: Determine whether the standardized business concept is compatible with the business rules in the preset knowledge base; If the standardized business concept matches the business rules in the preset knowledge base, then proceed to the step of obtaining the technical components that match the standardized business concept through the preset mapping relationship table; If the standardized business concept does not match the business rules in the preset knowledge base, a rule conflict will be indicated, and a rematch will be performed after obtaining user confirmation.

4. The method according to claim 1, characterized in that, The step of obtaining the technical components that match the standardized business concept through a preset mapping table includes: Based on the standardized business concept, query the preset mapping relationship table to obtain the technical components to be verified corresponding to the standardized business concept; The technical component to be verified is verified to confirm whether it is usable; If the technical component to be verified is available, then the technical component to be verified shall be used as the technical component. If the technical component to be verified is unavailable, it is replaced with an alternative technical component that is functionally equivalent to the technical component to be verified, and the alternative technical component is used as the technical component after the alternative technical component passes the verification.

5. The method according to claim 1, characterized in that, The step of querying contextual information that matches the standardized business concept through a preset knowledge base includes: Based on the standardized business concept, extract business rules and rule thresholds that match the standardized business concept from the preset knowledge base to obtain the context information to be verified; The context information to be verified is obtained by performing a consistency check on the context information to be verified.

6. The method according to claim 1, characterized in that, The method further includes: Audit information is embedded in the standardized workflow, wherein the audit information includes the source information of the natural language and the context information.

7. The method according to claim 1, characterized in that, The method further includes: If a workflow to be generated is detected, the business scenario of the workflow to be generated is obtained to obtain the first business scenario; Identify the business scenarios for which workflows have already been generated to obtain multiple second business scenarios; If the first business scenario is similar to any one of the multiple second business scenarios, then the second business scenario that is similar to the first business scenario shall be taken as the target business scenario. Obtain the workflow of the target business scenario to obtain the workflow to be configured; Configure the workflow to be configured according to the business requirements of the first business scenario to obtain the workflow to be generated.

8. An intelligent workflow generation device, characterized in that, include: The first acquisition module is used to acquire natural language input by the user and parse the natural language to obtain standardized business concepts that match the natural language. The second acquisition module is used to acquire technical components that match the standardized business concept through a preset mapping relationship table; The first query module is used to query context information that matches the standardized business concept through a preset knowledge base, wherein the context information includes business rules; The first generation module is used to generate a standardized workflow based on the standardized business concept, the technical components, and the context information.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent workflow generation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the intelligent workflow generation method as described in any one of claims 1 to 7.

Images