Interaction method and system combining domain knowledge graph and dynamic intention clarification

Abstract

The invention belongs to the technical field of information interaction, and provides an interaction method and system combining a domain knowledge graph and dynamic intention clarification, and the method comprises the steps: analyzing a table structure in an enterprise management system database, mapping fields into entity attributes of the knowledge graph, extracting rules in a business process document of an enterprise management system, and storing the rules in the business process document of the enterprise management system; converting into relation constraints of the knowledge graph, wherein each relation is configured with a weight value; the input of a user is obtained, semantic recognition is conducted on the input through a pre-trained large model, key elements of the intention of the user are extracted, the priority of missing item content is calculated according to the weight value and the key elements, and the missing item content needing to be clarified is recognized according to the priority; generating a clarification problem according to the missing item content needing to be clarified; and obtaining an answer of the user about the clarified question, and performing integrity, consistency and compliance verification on the answer. The clarification process efficiency can be improved, and the user experience is improved.

Description

Technical Field

[0001] The present invention belongs to the field of information interaction technology, and specifically relates to an interaction method and system combining domain knowledge graph and dynamic intent clarification. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Enterprise Resource Planning (ERP) systems generally use form filling or menu navigation for interaction, which lacks a dynamic recognition mechanism, resulting in low interaction efficiency and a poor user experience.

[0004] Although there are a few systems that support natural language interaction, there are two main problems: When users ask questions in natural language, general dialogue models lack knowledge in the field of enterprise management and cannot respond accurately due to insufficient professional information or ambiguity. They also mechanically follow up on pre-set questions, ignoring business process constraints and lacking intelligent prioritization based on business scenarios, resulting in an inefficient clarification process. Furthermore, they lack the ability to clearly identify and prioritize issues before asking follow-up questions. Figure One consistency and business compliance, which can easily lead to operational errors. Summary of the Invention

[0005] In order to solve the above problems, the present invention proposes an interactive method and system that combines domain knowledge graph and dynamic intent clarification. The present invention extracts entities, attributes and relationships from the ERP business database, constructs a structured knowledge network, generates a prioritized sequence of clarification questions based on the knowledge graph and user context, and optimizes the knowledge graph according to user feedback, which can achieve accurate response, improve the efficiency of the clarification process, and enhance the user experience.

[0006] According to some embodiments, the present invention adopts the following technical solutions: An interactive method combining domain knowledge graph and dynamic intent clarification includes the following steps: Parse the table structure in the enterprise management system database, map the fields to entity attributes of the knowledge graph, extract the rules in the business process documents of the enterprise management system, and convert them into relationship constraints of the knowledge graph. Each relationship is configured with a weight value. Obtain user input, use a pre-trained large model to perform semantic recognition on the input, extract the key elements of the user's intent, calculate the priority of the missing content based on the weight value and key elements, and identify the missing content that needs to be clarified based on the priority; Generate clarification questions based on the missing items that need clarification; Obtaining the user's answer to the clarification question, performing integrity, consistency and compliance checks on the answer.

[0007] As an alternative embodiment, it further comprises the following steps: updating the knowledge graph according to the user's answer to the clarification question, and adjusting the weight value of the corresponding relationship configuration.

[0008] As a further embodiment, if the user modifies several attributes in the clarification question more than a set number of times, a new relationship constraint is added and the knowledge graph is updated. If the user ignores the content of the clarification question more than a set number of times, the relationship constraint related to the content is reduced and the weight value of the corresponding relationship configuration is reduced.

[0009] As an alternative embodiment, the process of mapping fields to entity attributes of the knowledge graph includes: parsing all table structures in the enterprise management system database, extracting fields contained in the table structures, and constructing corresponding attributes.

[0010] As an alternative embodiment, in the process of extracting rules in the business process document of the enterprise management system, natural language processing technology is used to extract rules in the business process document of the enterprise management system, and the rules are converted into relationship constraints of the knowledge graph.

[0011] As an alternative embodiment, the node types of the knowledge graph include business entities, business processes and rule constraints.

[0012] As an alternative embodiment, the process of calculating the priority of the missing item content according to the weight value and the key element includes: the priority is: P =α×W_rule +β×F_usage +γ×R_risk; Where W_rule is the weight of the mandatory attribute defined in the knowledge graph, F_usage is the filling frequency of the attribute in the user's historical operation, and R_risk is the risk coefficient, which is configured according to whether the attribute involves an imminent business risk. The coefficients α, β and γ are dynamically adjusted by rules or large model reasoning capabilities.

[0013] As an alternative embodiment, the process of identifying the missing item content that needs to be clarified according to the priority includes: if the value of the priority is greater than a set threshold, a clarification question is generated, and the content of the question is arranged according to the size of the priority of the missing item content that needs to be clarified.

[0014] An interactive system combining a domain knowledge graph and dynamic intent clarification, comprising: The knowledge graph construction module is configured to parse a table structure in an enterprise management system database, map fields to entity attributes of a knowledge graph, extract rules in a business process document of the enterprise management system, and convert the rules into relationship constraints of the knowledge graph, and each relationship is configured with a weight value; The missing item priority calculation module is configured to obtain input of a user, perform semantic recognition on the input by using a pre-trained large model, extract key elements of an intention of the user, calculate a priority of content of the missing item according to the weight value and the key elements, and identify content of the missing item that needs to be clarified according to the priority; The clarification question generation module is configured to generate a clarification question according to the content of the missing item that needs to be clarified. The interaction verification module is configured to obtain an answer of the user to the clarification question, and perform completeness, consistency and compliance verification on the answer.

[0015] An electronic device includes a memory and a processor, and computer instructions stored on the memory and running on the processor, when the computer instructions are run by the processor, the steps in the above method are completed.

[0016] Compared with the prior art, the beneficial effects of the present application are: The present application extracts entities, attributes and relationships from a business database of an enterprise management system, constructs a structured knowledge network, and incorporates knowledge in the field of enterprise management, so that the large model can understand the association relationship of business entities.

[0017] The present application generates a priority-ordered clarification question sequence based on a knowledge graph and user context, and in the generation process, the priority is calculated according to the weight value of the relationship constraint in the knowledge graph, and the business process constraints are considered, which improves the efficiency of clarification.

[0018] The present application verifies business rules in real time during clarification, controls the intention Figure One consistency and business compliance, ensures the completeness, consistency and compliance of the clarification content, and avoids unnecessary risks.

[0019] The present application optimizes the knowledge graph and clarification strategy according to user feedback, counts the modification and ignored information of the user to the clarification question, updates the knowledge graph and adjusts the weight value, further optimizes the entire process, can realize accurate response, improves the efficiency of the clarification process, and improves the user experience.

[0020] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the following preferred embodiments are described in detail below, and the accompanying drawings are described as follows. BRIEF DESCRIPTION OF DRAWINGS

[0021] The accompanying drawings, which form a part of this specification, are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification. The embodiments of the application, and their

[0022] Figure 1 is a schematic flow diagram of an embodiment. DETAILED DESCRIPTION

[0023] The application will be further described with reference to the drawings and embodiments.

[0024] It should be noted that the following detailed description is exemplary in nature and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0025] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments in accordance with the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, steps, operations, elements, components, and / or groups thereof, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof.

[0026] The embodiments in the application and the features in the embodiments can be combined with each other without conflict.

[0027] Embodiment one An interactive method combining a domain knowledge graph and dynamic intent clarification, as shown in Figure 1 includes the following steps: The table structure in the enterprise management system database is parsed, the fields are mapped to the entity attributes of the knowledge graph, the rules in the business process document of the enterprise management system are extracted, and the rules are converted into the relationship constraints of the knowledge graph, and each relationship is configured with a weight value; The input of the user is obtained, the pre-trained large model is used for semantic recognition of the input, the key elements of the user's intent are extracted, the priority of the missing item content is calculated according to the weight value and the key elements, and the missing item content that needs to be clarified is identified according to the priority; According to the missing item content that needs to be clarified, a clarification question is generated; The answer of the user about the clarification question is obtained, and the completeness, consistency and compliance of the answer are checked.

[0028] In some embodiments, the following steps are further included: updating the knowledge graph according to the answer of the user to the clarification question, and adjusting the weight value of the corresponding relationship configuration.

[0029] If the user modifies one or some attributes in the clarification question more than a set number of times, a new relationship constraint is added, and the knowledge graph is updated. If the user ignores the content of the clarification question more than a set number of times, the relationship constraint related to the content is reduced, and the weight value of the corresponding relationship configuration is reduced.

[0030] In this embodiment, the process of mapping the fields to the entity attributes of the knowledge graph includes: parsing all table structures (such as master data tables and supplier contract tables) in the enterprise management system database, extracting the fields contained in the table structures, and constructing corresponding attributes (for example, the material entity contains attributes: material code, safety stock quantity, and procurement time).

[0031] Of course, in other embodiments, the table structure can contain other types.

[0032] In this embodiment, in the process of extracting the rules in the business process document of the enterprise management system, a natural language processing technology is used to extract the rules in the business process document of the enterprise management system, and the rules are converted into relationship constraints of the knowledge graph (for example: creating a purchase order needs to meet the contract state of the supplier is valid, etc.).

[0033] The natural language processing technology can use a pre-trained deep learning model.

[0034] In this embodiment, the node types of the knowledge graph include business entities (materials / orders), business processes (approval / warehousing), rule constraints (permissions / thresholds), and the like.

[0035] Each relationship is accompanied by a weight value.

[0036] The weight value can be pre-configured / set. In some embodiments, a pre-trained deep learning model can also be used to adaptively set the weight value.

[0037] However, in each embodiment, the core setting principle of the weight value is: according to the attributes or entities corresponding to each relationship, the importance of the completeness and correctness of the business process to which it belongs. If the importance is higher, the weight value is larger, and correspondingly, if the importance is lower, the weight value is smaller.

[0038] According to the business mandatory and data constraints of the attributes, the weight is preliminarily divided into three levels, and the value range of each level is different. Specifically, the assignment method shown in Table 1 can be used for setting: Table 1: Weight value assignment example

[0039] In this embodiment, the input of the user is obtained, the pre-trained large model is used to perform semantic recognition on the input, the key elements of the user's intention are extracted, the priority of the missing item content is calculated according to the weight value and the key elements, and in the process of identifying the missing item content that needs to be clarified, the semantic understanding can be automatically performed by the large model according to the user input, the key elements of the user's intention are extracted, and the missing item (item that needs to be clarified) content is identified based on the relationship graph.

[0040] The large model can be a large language model (LLM). The training process of the large language model can adopt the prior art, which will not be described here.

[0041] For example, the user input "please place a purchase order" uses the intent parser to identify the action verb "place an order" and the business object "purchase"; Load the mandatory entity chain associated with "purchase order" from the knowledge graph: Purchase order -> must be associated -> material; Purchase order -> must be associated -> supplier; Material -> attribute constraint -> minimum purchase quantity; The -> indicates the correlation.

[0042] In this embodiment, the process of calculating the priority of the missing item content according to the weight value and the key elements includes: P = α × W_rule + β × F_usage + γ × R_risk; Wherein, W_rule is the weight of the mandatory attribute defined in the knowledge graph, F_usage is the filling frequency of the attribute in the user's historical operation, and R_risk is the risk coefficient, which is configured according to whether the attribute involves the business risk that will be triggered soon, for example, the supplier contract expires in 3 days, it is considered that there is a risk, R_risk = 1.0; The coefficients α, β and γ are dynamically adjusted by rules or large model reasoning ability, and in this embodiment, the initial value α is set to 0.6, the initial value β is set to 0.3, and the initial value γ is set to 0.1.

[0043] The large model can use a pre-trained large language model.

[0044] Of course, in other embodiments, the above initial values can be adjusted, as long as the sum of the initial values of the three coefficients is equal to one.

[0045] Similarly, in other embodiments, the initial values of the coefficients α, β and γ also need to be set according to the importance of the mandatory attributes defined in the knowledge graph, the filling frequency of the attribute in the user's historical operation and the risk coefficient, in order to ensure accuracy and adjustment efficiency.

[0046] In some embodiments, the system can continuously optimize the missing item content to be clarified through a closed-loop learning module, and the adjustment basis can include: User Ignored Rate: If the clarification question of a high-weight attribute is frequently skipped or the default value is selected by the user, the system will gradually reduce its weight (e.g., W_rule = W_rule * 0.9, i.e., the updated weight value is reduced to 0.9 times the original weight value), indicating that the attribute is not as critical as the preset for the current user or scenario.

[0047] User Correction Rate: If the value of a low-weight attribute (such as an automatically selected supplier) filled by the system is frequently modified by the user, the system will appropriately increase the weight of the attribute (e.g., W_rule = min(W_rule + 0.1, 1.0), i.e., the weight value is added by 0.1, and the upper threshold value 1.0 is compared, if it is less than the upper threshold value, the updated weight value can be selected, otherwise, the upper threshold value is taken as the updated weight value), and record the user's new preference in the knowledge graph.

[0048] Operation Success Rate: The final successful operation verifies the accuracy of the clarification process in reverse. If an attribute is 100% specified in a successful operation, its weight may be fine-tuned and strengthened.

[0049] In this embodiment, the process of identifying the missing item content that needs to be clarified according to the priority includes: if the value of the priority is greater than a set threshold (e.g., the set threshold is 0.7), a clarification question is generated, and the content in the question is arranged according to the priority of the missing item content that needs to be clarified, and the generation of the clarification question is performed in order of priority.

[0050] Of course, in some embodiments, a threshold value of the priority score can also be set, and if the priority is higher than the set threshold, the generation of the clarification question is performed, and if it is lower than the set threshold, the question can be ignored and skipped (e.g., "material selection" with high priority is clarified first, and "delivery address" with low priority is skipped).

[0051] Next, according to the user's answer to the generated clarification question, three-level verification is performed: The completeness verification process is to check whether all mandatory attributes in the knowledge graph have been filled, and if there are mandatory attributes that have not been filled, a dialogue is generated to guide the user to fill in; For example, if the procurement order lacks material name and quantity, the material name and quantity are mandatory attributes, and a dialogue needs to be generated to ask the user to fill in the name and quantity of the material.

[0052] Of course, if all mandatory attributes have been filled, it is considered that the completeness verification is passed.

[0053] Enter the consistency verification process.

[0054] The consistency verification process is to compare the historical answer logic (for example, the user first says "budget is sufficient" and then inputs "purchase amount exceeds the limit"), determine whether the context information is consistent, and if not, generate a dialogue for further confirmation. If the logic of the latest answer and the historical answer does not match, a new dialogue is generated to further confirm the point of logical contradiction.

[0055] For example, if the purchase order is missing the material name and quantity, and the user fills in the material name and quantity do not match, the material name is desktop computer and the quantity is 10 tons, it is considered that the two do not match, a new dialogue is generated, please confirm whether you want to purchase desktop computers, the quantity is 10 or 10 tons? Of course, when generating a new dialogue, it can be limited according to the relationship or restrictions stored in the knowledge graph, such as the material name is desktop computer and the quantity is 10 tons, it is considered that the two do not match, and the current knowledge graph stores the minimum purchase quantity of desktop computer as 20, a new dialogue can be generated, please confirm whether you want to purchase desktop computers, the quantity is 10 or 10 tons, please confirm? The user's answer is 10, which can be further questioned, the minimum purchase quantity of desktop computer is 20, please confirm whether it is 10? If the logic of the latest answer and the historical answer matches, it is considered that the consistency verification is passed.

[0056] Next, the compliance verification is performed.

[0057] The compliance verification process is to query the ERP business rule library in real time, and if there is information that does not comply with the ERP business rules (such as IF purchase amount > user permission), the corresponding information is intercepted.

[0058] Of course, in some embodiments, the order of integrity, consistency and compliance can also be changed when actually performed.

[0059] In some embodiments, an evaluation mechanism can also be provided to guide the user to operate "accept / modify / ignore" on the clarification question; according to the user's operation, the knowledge graph is updated.

[0060] For example, for a purchase order / purchase contract, for a certain material A, when the user repeatedly modifies a certain attribute (such as changing "supplier X" to "supplier Y"), a new relationship is automatically added in the knowledge graph: material A → optional supplier → supplier Y; For example, if the user frequently ignores a clarification question (such as "delivery address") for the generated clarification dialogue, the weight of the associated attribute is reduced (for example, its weight W_rule is reduced by 0.1).

[0061] Embodiment two An interactive system combining a domain knowledge graph and dynamic intent clarification, comprising: A knowledge graph construction module configured to parse table structures in an enterprise management system database, map fields to entity attributes of a knowledge graph, extract rules in business process documents of the enterprise management system, and convert the rules to relationship constraints of the knowledge graph, each relationship being configured with a weight value; A missing item priority calculation module configured to obtain user input, perform semantic recognition on the input using a pre-trained large model, extract key elements of the user's intent, calculate the priority of missing item content according to the weight value and the key elements, and identify missing item content that needs to be clarified according to the priority; A clarification question generation module configured to generate clarification questions according to the missing item content that needs to be clarified; An interactive verification module configured to obtain user answers to the clarification questions and perform completeness, consistency, and compliance verification on the answers.

[0062] In some embodiments, the following modules are further included: A knowledge graph update module configured to update the knowledge graph and adjust the weight value of the corresponding relationship configuration according to the user's answers to the clarification questions.

[0063] The knowledge graph update module is specifically configured to add new relationship constraints and update the knowledge graph if the user modifies a number of attributes in the clarification questions more than a set number of times. If the user ignores the content of the clarification questions more than a set number of times, the relationship constraints related to the content are reduced and the weight value of the corresponding relationship configuration is reduced.

[0064] The process in which the knowledge graph construction module maps fields to entity attributes of the knowledge graph includes parsing all table structures in the enterprise management system database, extracting fields contained in the table structures, and constructing corresponding attributes.

[0065] In the process in which the knowledge graph construction module extracts rules in business process documents of the enterprise management system, natural language processing technology is used to extract rules in business process documents of the enterprise management system and convert the rules to relationship constraints of the knowledge graph.

[0066] The node types of the knowledge graph include business entities, business processes, and rule constraints.

[0067] The process in which the missing item priority calculation module calculates the priority of missing item content according to the weight value and the key elements includes that the priority is: P = a x W rule + b x F usage + g x R risk; Wherein, W rule is the mandatory attribute weight defined in the knowledge graph, F usage is the filling frequency of the attribute in the user's historical operation, and R risk is the risk coefficient, which is configured according to whether the attribute involves the business risk to be triggered soon; The coefficients a, b and g are dynamically adjusted by rules or large model reasoning capabilities.

[0068] The process of identifying the missing item content that needs to be clarified by the priority calculation module includes: if the value of the priority is greater than a set threshold, a clarification question is generated, and the content in the question is arranged according to the size of the priority of the missing item content that needs to be clarified.

[0069] It can be understood that the above-mentioned units / modules can be combined into one or several other units / modules respectively or all, or some of them can be further split into a plurality of units with smaller functions to constitute, which can realize the same operation without affecting the realization of the technical effects of the embodiments of the present application.

[0070] The above-mentioned modules of the system are divided based on logical functions. In actual application, the function of one module can also be realized by multiple modules, or the functions of multiple modules can be realized by one module. For example, the missing item priority calculation module in the embodiment can include: The user answer acquisition module is configured to acquire the input of the user; The semantic analysis module is configured to use a pre-trained large model to perform semantic recognition on the input and extract key elements of the user's intention; The priority calculation module is configured to calculate the priority of the missing item content according to the weight value and the key elements; The missing item identification module is configured to identify the missing item content that needs to be clarified according to the priority.

[0071] Similarly, in other embodiments of the present application, the system can also include other units / modules. In actual application, these functions can also be realized by other units, and can be realized by cooperation of multiple units.

[0072] Embodiment three An electronic device includes a memory and a processor, and computer instructions stored on the memory and running on the processor, when the computer instructions are run by the processor, the steps in the method provided by embodiment one are completed.

[0073] Alternatively, when the computer instructions are run by the processor, the following steps are completed: The table structure in the enterprise management system database is parsed, fields are mapped to entity attributes of the knowledge graph, rules in the business process document of the enterprise management system are extracted, and are converted into relationship constraints of the knowledge graph, and each relationship is configured with a weight value; An input of a user is acquired, a pre-trained large model is used to perform semantic recognition on the input, key elements of the user's intention are extracted, a priority of a missing item content is calculated according to the weight value and the key elements, and the missing item content that needs to be clarified is identified according to the priority; A clarification question is generated according to the missing item content that needs to be clarified; An answer of the user to the clarification question is acquired, and integrity, consistency and compliance verification is performed on the answer.

[0074] Some embodiments further include the following steps: updating the knowledge graph according to the answer of the user to the clarification question, and adjusting the weight value of the corresponding relationship configuration.

[0075] For example, if the user modifies a plurality of attributes in the clarification question more than a set number of times, a new relationship constraint is added, and the knowledge graph is updated; If the user ignores the content of the clarification question more than a set number of times, the relationship constraint related to the content is reduced, and the weight value of the corresponding relationship configuration is reduced.

[0076] Those skilled in the art will understand that embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) containing computer usable program code. CD - ROM , optical storage, etc.) containing computer usable program code.

[0077] The present application is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure One The functions specified in one or more flows and / or blocks in the flowcharts and / or block diagrams. Figure One The functions specified in one or more flows and / or blocks in the flowcharts and / or block diagrams.

[0078] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flow Figure One The flow or flows and / or blocks Figure One The flow or flows and / or blocks

[0079] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow Figure One The flow or flows and / or blocks Figure One The flow or flows and / or blocks

[0080] The above description is only preferred embodiments of the present application and is not intended to limit the present application. The present application can be variously changed and modified by those skilled in the art without departing from the spirit and scope of the present application. Any modification, equivalent replacement, improvement, etc. made by those skilled in the art without departing from the spirit and scope of the present application should be included in the scope of the present application.

Claims

1. An interactive method that combines a domain knowledge graph with dynamic intent clarification, characterized in that, The method comprises the following steps: parsing the table structure in the enterprise management system database, mapping the fields to entity attributes of the knowledge graph, extracting rules in the business process document of the enterprise management system, and converting the rules to relationship constraints of the knowledge graph, each relationship being configured with a weight value; obtaining the input of the user, performing semantic recognition on the input by using a pre-trained large model, extracting key elements of the user's intention, calculating the priority of the missing item content according to the weight value and the key elements, and identifying the missing item content that needs to be clarified according to the priority; generating a clarification question according to the missing item content that needs to be clarified; obtaining the answer of the user to the clarification question, and performing completeness, consistency and compliance verification on the answer.

2. The method of claim 1, wherein the method further comprises: The method further comprises the following steps: updating the knowledge graph according to the answer of the user to the clarification question, and adjusting the weight value of the corresponding relationship configuration.

3. The interactive method combining domain knowledge graph and dynamic intent clarification as claimed in claim 2, characterized in that: If the user modifies a number of attributes in the clarification question more than a set number of times, a new relationship constraint is added, and the knowledge graph is updated; if the user ignores the content of the clarification question more than a set number of times, the relationship constraint related to the content is reduced, and the weight value of the corresponding relationship configuration is reduced.

4. The interactive method combining domain knowledge graph and dynamic intent clarification according to claim 1, characterized in that: The process of mapping the fields to entity attributes of the knowledge graph comprises: parsing all table structures in the enterprise management system database, extracting the fields contained in the table structures, and constructing corresponding attributes.

5. The method of claim 1, wherein the method further comprises: In the process of extracting rules in the business process document of the enterprise management system, a natural language processing technology is used to extract rules in the business process document of the enterprise management system, and the rules are converted to relationship constraints of the knowledge graph.

6. The interactive method combining domain knowledge graph and dynamic intent clarification according to claim 1, characterized in that: The node types of the knowledge graph include business entities, business processes and rule constraints.

7. The method of claim 1, wherein the method further comprises: The process of calculating the priority of the missing item content according to the weight value and the key elements comprises: the priority is: P = α × W_rule + β × F_usage + γ × R_risk; wherein W_rule is a mandatory attribute weight defined in the knowledge graph, F_usage is the filling frequency of the attribute in the user's historical operation, and R_risk is a risk coefficient configured according to whether the attribute involves an imminent business risk; coefficients α, β and γ are dynamically adjusted by rules or large model reasoning capabilities.

8. The interactive method combining domain knowledge graph and dynamic intent clarification according to claim 1, characterized in that: The process of identifying the missing item content that needs to be clarified according to the priority comprises: if the value of the priority is greater than a set threshold, a clarification question is generated, and the content in the question is arranged according to the size of the priority of the missing item content that needs to be clarified.

9. An interactive system that combines a domain knowledge graph with dynamic intent clarification, characterized in that, The method comprises: a knowledge graph construction module configured to parse the table structure in the enterprise management system database, map the fields to entity attributes of the knowledge graph, extract rules in the business process document of the enterprise management system, and convert the rules to relationship constraints of the knowledge graph, each relationship being configured with a weight value; a missing item priority calculation module configured to obtain the input of the user, perform semantic recognition on the input by using a pre-trained large model, extract key elements of the user's intention, calculate the priority of the missing item content according to the weight value and the key elements, and identify the missing item content that needs to be clarified according to the priority; a clarification question generation module configured to generate a clarification question according to the missing item content that needs to be clarified; and a verification module configured to obtain the answer of the user to the clarification question, and perform completeness, consistency and compliance verification on the answer. An interaction verification module configured to obtain a user's response to a clarification question, and perform integrity, consistency, and compliance checks on the response.

10. An electronic device, characterized by comprising: A computer program product comprising a memory and a processor, and computer instructions stored on the memory and running on the processor, the computer instructions, when run by the processor, completing the steps in the method of any one of claims 1-8.

Images