Financial service inquiry method and device based on artificial intelligence model, equipment and medium

Abstract

This application discloses a method, apparatus, device, and medium for financial business inquiry based on an artificial intelligence model, relating to the field of artificial intelligence planning. The method includes: determining a preset constraint dimension corresponding to an initial artificial intelligence model; obtaining constraint information corresponding to the preset constraint dimension; applying rule constraints to the initial artificial intelligence model based on the constraint information for each preset constraint dimension to obtain a set of candidate financial business planning domains that satisfy a first preset condition; determining a target financial business planning domain that satisfies a second preset condition from the set of candidate financial business planning domains; and loading the target financial business planning domain into the initial artificial intelligence model to obtain a target financial business question-and-answer model, so that the user can use the target financial business question-and-answer model to conduct financial business inquiries. In this way, this application can stably and efficiently determine the target financial business planning domain.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence planning, and in particular to financial business inquiry methods, apparatus, equipment and media based on artificial intelligence models. Background Technology

[0002] With the widespread application of artificial intelligence technology in enterprise-level systems and automated business scenarios, the concept of "financial business planning domain" has been introduced to address the problem of the lack of a unified, reasonable, and constrained business capability modeling method for artificial intelligence in complex business scenarios.

[0003] The financial business planning domain, as the core foundational model for planning and reasoning, is used to define the searchable state space and action set of an artificial intelligence system. In practical systems, it is often necessary to maintain multiple financial business planning domains simultaneously to support different business types, different permission boundaries, or different risk strategies.

[0004] In existing technical solutions, the selection method for concepts similar to financial business planning domains typically treats the question of "whether a financial business planning domain is available" as a Boolean judgment problem, that is, determining whether a financial business planning domain enters the planning process based on rule conditions. This approach ignores the relative matching degree between user intent and the capabilities of the financial business planning domain, cannot quantitatively compare multiple "partially available" financial business planning domains, and is also difficult to dynamically determine the optimal financial business planning domain at runtime.

[0005] In conclusion, how to stably and efficiently determine the target financial business planning domain is an urgent problem to be solved. Summary of the Invention

[0006] In view of this, the purpose of this invention is to provide a method, apparatus, device, and medium for financial business inquiry based on an artificial intelligence model, capable of stably and efficiently determining the target financial business planning domain. The specific solution is as follows: Firstly, this application provides a financial business inquiry method based on an artificial intelligence model, including: Determine the preset constraint dimensions corresponding to the initial artificial intelligence model; Obtain the constraint information corresponding to the preset constraint dimension; wherein, the constraint information includes any one or more of the following: user intent information, financial business planning domain capability information, permission constraint information, and financial business risk cost information; Based on the constraint information, the initial artificial intelligence model is subjected to rule constraints in each of the preset constraint dimensions to obtain a set of candidate financial business planning domains that satisfy the first preset condition. From the set of candidate financial business planning domains, a target financial business planning domain that meets the second preset condition is determined. The target financial business planning domain is loaded into the initial artificial intelligence model to obtain the target financial business question-and-answer model, so that the user can use the target financial business question-and-answer model to inquire about financial business.

[0007] Optionally, the preset constraint dimensions include financial business constraint dimensions, role and permission constraint dimensions, user intent constraint dimensions, and financial business risk and execution cost constraint dimensions. Accordingly, the initial artificial intelligence model is subjected to rule constraints on each of the preset constraint dimensions based on the constraint information to obtain a set of candidate financial business planning domains that satisfy the first preset condition, including: Determine the rule convergence level corresponding to each of the preset constraint dimensions; The constraint order for applying rule constraints to the initial artificial intelligence model is obtained based on the rule convergence level. Obtain financial business inquiry requests sent by the user client; Determine the initial set of financial business planning domains based on the aforementioned financial business inquiry request; Based on the constraint order of each preset constraint dimension, and according to the constraint information and preset scoring function, the initial set of financial business planning domains is subjected to rule constraints to obtain the current set of candidate financial business planning domains that meet the first preset condition.

[0008] Optionally, determining the initial set of financial business planning domains based on the financial business inquiry request includes: The financial business inquiry request is semantically parsed by a preset artificial intelligence semantic encoding model to generate a corresponding target business semantic feature vector. A preset mapping table is determined; the preset mapping table is used to store the mapping relationship between the preset financial business planning domain and the business semantic feature vector. Based on the preset mapping table, the initial financial business planning domain set corresponding to the target business semantic feature vector is determined from the preset financial business planning domain set.

[0009] Optionally, the step of applying rule constraints to the initial set of financial business planning domains based on the constraint information and a preset scoring function according to the constraint order of each of the preset constraint dimensions to obtain the current set of candidate financial business planning domains that satisfy the first preset condition includes: In the financial business constraint dimension, the initial financial business planning domain set is scored according to a preset scoring function to obtain the first score corresponding to each financial business planning domain in the initial financial business planning domain set; Determine whether the first score corresponding to each financial business planning domain in the initial set of financial business planning domains is greater than a preset threshold; If so, the corresponding financial business planning domain will be integrated into the current set of candidate financial business planning domains.

[0010] Optionally, the step of applying rule constraints to the initial set of financial business planning domains based on the constraint information and a preset scoring function according to the constraint order of each of the preset constraint dimensions to obtain the current set of candidate financial business planning domains that satisfy the first preset condition includes: Obtain permission constraint information from the user's client; In the role and permission constraint dimension, the corresponding role attributes are determined based on the permission constraint information; Determine the matching degree between the role attribute and each financial business planning domain in the current candidate financial business planning domain set, and generate a corresponding second score based on the matching degree using the preset scoring function; The second score of each financial business planning domain is weighted and fused with the corresponding first score to obtain the comprehensive score of each financial business planning domain. Determine whether the comprehensive score corresponding to each financial business planning domain in the current candidate financial business planning domain set is greater than the preset threshold. If not, the corresponding financial business planning domain is determined to not meet the preset role constraint conditions, and the corresponding financial business planning domain is removed from the current candidate financial business planning domain set to generate a new current candidate financial business planning domain set.

[0011] Optionally, the step of applying rule constraints to the initial set of financial business planning domains based on the constraint information and a preset scoring function according to the constraint order of each of the preset constraint dimensions to obtain the current set of candidate financial business planning domains that satisfy the first preset condition includes: In the user intent constraint dimension, the intent decomposition algorithm of the preset artificial intelligence semantic encoding model is used to decompose the current user intent information into several atomic sub-intents; Establish the logical relationships between the atomic sub-intents, and generate a structured sequence of atomic sub-intents based on the logical relationships; Based on the atomic sub-intention sequence, the decomposed atomic sub-intention is parsed into an intention semantic feature vector; The action set, state transition rules, and inherent constraints of the current candidate financial business planning domain set are obtained, and the action set, state transition rules, and inherent constraints are semantically encoded through the preset artificial intelligence semantic encoding model to generate the financial business planning domain capability feature vector of each financial business planning domain in the current candidate financial business planning domain set. Determine the similarity between the intent semantic feature vector and the financial business planning domain capability feature vector; The third score of each financial business planning domain in the current candidate financial business planning domain set is determined based on the similarity using the preset scoring function. The third score of each financial business planning domain is weighted and merged with the corresponding comprehensive score to obtain a new comprehensive score; Determine whether the new comprehensive score of each financial business planning domain in the current candidate financial business planning domain set is greater than the preset threshold; If not, the corresponding financial business planning domain is determined to not meet the preset user intent constraints, and the corresponding financial business planning domain is removed from the current candidate financial business planning domain set to generate a new current candidate financial business planning domain set.

[0012] Optionally, the step of applying rule constraints to the initial set of financial business planning domains based on the constraint information and a preset scoring function according to the constraint order of each of the preset constraint dimensions to obtain the current set of candidate financial business planning domains that satisfy the first preset condition includes: In the dimension of financial business risk and execution cost constraints, obtain the financial business risk cost information of each financial business planning domain in the current set of candidate financial business planning domains; wherein the financial business risk cost information includes preset risk level, compliance attribute and execution cost indicator; The financial business risk cost information is quantified into a penalty factor and introduced into the preset scoring function to obtain the target penalty scoring function; The comprehensive score of each financial business planning domain in the current candidate financial business planning domain set is reduced using the target punitive scoring function to obtain the reduced comprehensive score; Determine whether the reduced comprehensive score corresponding to each financial business planning domain in the current candidate financial business planning domain set is greater than the preset threshold. If not, the corresponding financial business planning domain is determined to not meet the preset risk and cost constraints, and the corresponding financial business planning domain is removed from the current candidate financial business planning domain set to generate a new current candidate financial business planning domain set.

[0013] Optionally, determining the target financial business planning domain that meets the second preset condition from the set of candidate financial business planning domains includes: The candidate financial business planning domain set is sorted according to a preset order to generate a candidate list of target financial business planning domains; Determine whether the financial business planning domains in the candidate list of target financial business planning domains meet the preset rule convergence conditions; the preset rule convergence conditions include any one or more of the preset conditions for large score differences, preset conditions for stable change amplitude, and preset conditions for consistent multi-dimensional scores. If so, the financial business planning domain that meets the preset highest score condition in the candidate list of target financial business planning domains will be directly determined as the target financial business planning domain. If not, then the step of applying rule constraints to the initial set of financial business planning domains according to the constraint order of each of the preset constraint dimensions, based on the constraint information and the preset scoring function, to obtain the current set of candidate financial business planning domains that satisfies the first preset condition is executed again.

[0014] Secondly, this application provides a financial business inquiry device based on an artificial intelligence model, comprising: The constraint dimension determination module is used to determine the preset constraint dimensions corresponding to the initial artificial intelligence model; The constraint information acquisition module is used to acquire constraint information corresponding to the preset constraint dimension; wherein, the constraint information includes any one or more of the following: user intent information, financial business planning domain capability information, permission constraint information, and financial business risk cost information. The set acquisition module is used to apply rule constraints to the initial artificial intelligence model based on the constraint information for each of the preset constraint dimensions, so as to obtain a set of candidate financial business planning domains that satisfy the first preset condition. The planning domain determination module is used to determine the target financial business planning domain that meets the second preset condition from the set of candidate financial business planning domains. The model acquisition module is used to load the target financial business planning domain into the initial artificial intelligence model to obtain the target financial business question-and-answer model, so that the user terminal can use the target financial business question-and-answer model to conduct financial business inquiries.

[0015] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is used to execute the computer program to implement the aforementioned financial business inquiry method based on an artificial intelligence model.

[0016] Fourthly, this application provides a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned financial business inquiry method based on an artificial intelligence model.

[0017] In summary, this application first determines the preset constraint dimensions corresponding to the initial artificial intelligence model; then obtains the constraint information corresponding to the preset constraint dimensions; wherein, the constraint information includes any one or more of user intent information, financial business planning domain capability information, permission constraint information, and financial business risk cost information; based on the constraint information, the initial artificial intelligence model is subjected to rule constraints on each of the preset constraint dimensions to obtain a set of candidate financial business planning domains that meet the first preset condition; a target financial business planning domain that meets the second preset condition is determined from the set of candidate financial business planning domains; the target financial business planning domain is loaded into the initial artificial intelligence model to obtain a target financial business question-and-answer model, so that the user terminal can use the target financial business question-and-answer model to inquire about financial business. As can be seen from the above, this application first obtains an initial AI (Artificial Intelligence) model and determines the preset constraint dimensions, while collecting constraint information related to these dimensions, such as user intent, financial business planning domain capabilities, permission constraints, and risk costs. Next, the initial model is adjusted at the rule level using this constraint information to form a set of candidate financial business planning domains, and a target financial business planning domain that meets the requirements is selected through a preset scoring function. Finally, the selected target financial business planning domain is integrated into the initial model to generate the final target financial business question-and-answer model for user-side business consultation. In this way, by constructing an AI-based multi-dimensional scoring model for the financial business planning domain, the selection process is modeled as a computable, sortable, and adjudicable decision-making process. The target financial business planning domain is determined when the scoring results meet the convergence conditions, thus providing a stable and controllable technical constraint foundation for subsequent planning and reasoning processes. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 This application discloses a flowchart of a financial business inquiry method based on an artificial intelligence model. Figure 2 This is a schematic diagram of the structure of a financial business inquiry device based on an artificial intelligence model disclosed in this application; Figure 3 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation

[0020] 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 embodiments of the present invention, and not all embodiments. 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.

[0021] Currently, with the widespread application of artificial intelligence (AI) technology in enterprise-level systems and automated business scenarios, the concept of a "financial business planning domain" has been introduced to address the lack of a unified, reasonable, and constrained modeling method for AI's business capabilities in complex business scenarios. The financial business planning domain, as the core foundational model for planning and reasoning, is used to define the searchable state space and action set of the AI ​​system. In practical systems, multiple financial business planning domains often need to be maintained simultaneously to support different business types, different permission boundaries, or different risk strategies. Existing technical solutions typically treat the "availability of the financial business planning domain" as a Boolean judgment problem, determining whether a financial business planning domain enters the planning process based on rule conditions. This approach ignores the relative matching degree between user intent and the capabilities of the financial business planning domain, making it impossible to quantitatively compare multiple "partially available" financial business planning domains and dynamically determine the optimal financial business planning domain at runtime. To address these technical problems, this application discloses a financial business inquiry method, apparatus, device, and medium based on an AI model, capable of stably and efficiently determining the target financial business planning domain.

[0022] See Figure 1 As shown in the figure, this invention discloses a financial business inquiry method based on an artificial intelligence model, including: Step S11: Determine the preset constraint dimensions corresponding to the initial artificial intelligence model.

[0023] In this embodiment, several core constraint dimensions are preset during the initial construction of the artificial intelligence model. These dimensions aim to define clear boundaries for the model's design and behavior. These preset constraint dimensions include financial business constraints, role and permission constraints, user intent constraints, and financial business risk and execution cost constraints. The financial business constraints focus on the industry norms, business rules, and professional logic that the model must follow when processing various financial activities. The role and permission constraints limit the model's functional positioning in different application scenarios and the scope of data and functions it can access and operate. The user intent constraints focus on accurately understanding and identifying the deep goals and context of user input commands. Furthermore, the financial business risk and execution cost constraints emphasize guiding the model to simultaneously assess and control potential compliance risks, market risks, and the economic and operational costs required to implement the recommendations when outputting decisions or suggestions.

[0024] Step S12: Obtain the constraint information corresponding to the preset constraint dimension; wherein, the constraint information includes any one or more of the following: user intent information, financial business planning domain capability information, permission constraint information, and financial business risk cost information.

[0025] In this embodiment, constraint information corresponding to the preset constraint dimensions is obtained. This constraint information includes one or more combinations of user intent information, financial business planning domain capability information, permission constraint information, and financial business risk cost information. It should be noted that user intent information is mainly extracted through semantic analysis and pattern mining of text data such as historical customer service dialogues and business consultation records; financial business planning domain capability information comes from the financial institution's internal business manuals, product specifications, compliance process documents, and knowledge summaries based on expert experience; permission constraint information is mapped and organized based on the enterprise's organizational chart, job descriptions, and access control lists in the information system; and financial business risk cost information is obtained by integrating historical risk case libraries, compliance audit reports, business cost accounting data, and external regulatory policy texts.

[0026] Step S13: Based on the constraint information, apply rule constraints to the initial artificial intelligence model for each of the preset constraint dimensions to obtain a set of candidate financial business planning domains that satisfy the first preset condition.

[0027] In this embodiment, after determining the preset constraint dimensions and their corresponding constraint information, it is necessary to determine the rule convergence level corresponding to each preset constraint dimension; obtain the constraint order for applying rule constraints to the initial artificial intelligence model based on the rule convergence level; obtain the financial business inquiry request sent by the user terminal; determine the initial financial business planning domain set based on the financial business inquiry request; and apply rule constraints to the initial financial business planning domain set according to the constraint order of each preset constraint dimension, based on the constraint information and the preset scoring function, to obtain the current candidate financial business planning domain set that satisfies the first preset condition. Specifically, after determining the preset constraint dimensions and their corresponding constraint information, in order to establish an orderly and efficient rule application process, the rule convergence level corresponding to each preset constraint dimension is first defined. The rule convergence level aims to characterize the priority and order of action of rules of different dimensions in constraint logic. Based on this hierarchical division, the constraint order for applying rule constraints to the initial artificial intelligence model can be obtained. Upon receiving a financial business inquiry request from a user, the system first parses the request and generates an initial, broad set of financial business planning domains. This set includes various potential business solutions or action paths that the model initially deems likely to meet the user's needs. Subsequently, strictly following the previously determined constraint order, the system sequentially calls the constraint information and preset scoring functions under each preset constraint dimension to progressively filter and evaluate the initial set. Each round of constraints aims to eliminate options that do not meet the requirements of that dimension and retain or optimize the options that do meet the conditions. Through this sequential, multi-round constraint process, a current set of candidate financial business planning domains that satisfies the first preset condition is ultimately obtained. For example, a current set of candidate financial business planning domains with a comprehensive score higher than a threshold and passing all key dimension checks is obtained.

[0028] Furthermore, to determine the initial set of financial business planning domains, a preset artificial intelligence semantic encoding model is used to semantically parse the financial business inquiry request, generating a corresponding target business semantic feature vector; a preset mapping table is determined; the preset mapping table is used to store the mapping relationship between the preset financial business planning domains and the business semantic feature vectors; based on the preset mapping table, the initial set of financial business planning domains corresponding to the target business semantic feature vector is determined from the preset set of financial business planning domains. Specifically, the received financial business inquiry request is deeply parsed using a preset artificial intelligence semantic encoding model, which can understand the natural language meaning of the user's request and convert it into the corresponding target business semantic feature vector. A preset mapping table is pre-established and maintained, storing the known mapping relationship between the preset financial business planning domains and their corresponding business semantic feature vectors. Each preset financial business planning domain has been processed by the same semantic encoding model, generating its feature vector and recording it in the table. Finally, by querying the preset mapping table, the target business semantic feature vector generated by the user request is compared with the feature vectors stored in the preset mapping table for similarity calculation or matching. Based on the matching results, from the preset, complete set of financial business planning domains, those planning domains whose semantic features are highly relevant to the user's request are retrieved. These retrieved planning domains together constitute the initial set of financial business planning domains for the current specific query.

[0029] It's important to understand that after determining the initial set of financial business planning domains, rules are applied to each dimension of the initial set of financial business planning domains according to the constraint order. First, in the financial business constraint dimension, the initial set of financial business planning domains is scored according to a preset scoring function to obtain a first score for each financial business planning domain in the initial set. It is then determined whether the first score for each financial business planning domain in the initial set is greater than a preset threshold; if so, the corresponding financial business planning domain is integrated into the current candidate set of financial business planning domains. Specifically, in the financial business constraint dimension, the preset scoring function is invoked, and each financial business planning domain in the initial set is quantitatively evaluated based on constraint information such as financial business-related professional knowledge, compliance requirements, and business process logic. For example, whether a planning domain meets the application conditions for a specific product, whether its business logic is self-consistent, and whether it meets basic regulatory requirements are all converted into specific scores. Through this scoring process, a first score is calculated for each financial business planning domain in the initial set of financial business planning domains. Next, it is determined whether the first score obtained by each financial business planning domain is greater than a preset threshold. This threshold represents the minimum standard required for a solution to be considered "basically feasible" or "preliminarily qualified" in this business dimension. Only financial business planning domains with a first score greater than this preset threshold are considered to have passed the initial screening for business compliance and feasibility in this stage. These qualified planning domains are then selected and integrated into a temporary set for storing the qualified results of the current stage, i.e., the current candidate financial business planning domain set. Planning domains that fail to meet the threshold are eliminated in this dimension.

[0030] Secondly, the user's permission constraint information is obtained; based on the permission constraint information, the corresponding role attributes are determined in the role and permission constraint dimension; the matching degree between the role attributes and each financial business planning domain in the current candidate financial business planning domain set is determined, and a corresponding second score is generated based on the matching degree using the preset scoring function; the second score of each financial business planning domain is weighted and fused with the corresponding first score to obtain a comprehensive score for each financial business planning domain; it is determined whether the comprehensive score corresponding to each financial business planning domain in the current candidate financial business planning domain set is greater than the preset threshold; if not, it is determined that the corresponding financial business planning domain does not meet the preset role constraint conditions, and the corresponding financial business planning domain is removed from the current candidate financial business planning domain set to generate a new current candidate financial business planning domain set. Specifically, user role and permission information is introduced to perform accessibility scoring on the current candidate financial business planning domain set. Each planning domain is associated with at least one set of accessible roles. Based on the role attributes of the current user or agent, the matching degree between them and the current candidate financial business planning domain set is calculated, and a role accessibility score, i.e., the second score, is generated. Role accessibility score, as a dimension factor in the overall score of the planning domain, is integrated with the first score to calculate a comprehensive score. The comprehensive score of each financial business planning domain is compared with a preset threshold. If the comprehensive score of a planning domain is not greater than the preset threshold, it is determined that the planning domain does not meet the preset role constraints and is thus removed from the current set of candidate financial business planning domains. Through this screening process, a current set of candidate financial business planning domains is generated that not only meets the requirements in the financial business dimension but also passes the user role and permission dimensions.

[0031] Next, in the user intent constraint dimension, the current user intent information is decomposed into several atomic sub-intents using the intent decomposition algorithm of the preset artificial intelligence semantic encoding model; logical relationships are established between the atomic sub-intents, and a structured sequence of atomic sub-intents is generated based on the logical relationships; the decomposed atomic sub-intents are parsed into intent semantic feature vectors according to the atomic sub-intent sequence; the action set, state transition rules, and inherent constraints of the current candidate financial business planning domain set are obtained, and the action set, state transition rules, and inherent constraints are semantically encoded using the preset artificial intelligence semantic encoding model to generate the financial business planning domain of each financial business planning domain in the current candidate financial business planning domain set. Capability feature vector; determine the similarity between the intent semantic feature vector and the financial business planning domain capability feature vector; use the preset scoring function to determine the third score of each financial business planning domain in the current candidate financial business planning domain set based on the similarity; weight and fuse the third score of each financial business planning domain with the corresponding comprehensive score to obtain a new comprehensive score; determine whether the new comprehensive score of each financial business planning domain in the current candidate financial business planning domain set is greater than the preset threshold; if not, determine that the corresponding financial business planning domain does not meet the preset user intent constraint condition, and remove the corresponding financial business planning domain from the current candidate financial business planning domain set to generate a new current candidate financial business planning domain set. Specifically, in terms of user intent constraints, an intent decomposition algorithm is used to break down user intent into several atomic sub-intents. Each atomic sub-intent corresponds to a business operation of the smallest granularity, such as: acquiring data A + performing calculation B + sending notification C. First, the natural language input by the user is semantically parsed to identify core verbs, nouns, and relational words. Based on a pre-defined business operation library, the identified semantic elements are mapped to standardized atomic operations. Logical relationships between atomic sub-intents are established, such as sequential execution, conditional execution, and parallel execution. A structured sequence of atomic sub-intents is generated as the basis for subsequent semantic encoding. Through this atomic decomposition method, the system can effectively reduce the "illusion" problem in intent parsing and improve parsing accuracy. Simultaneously, the action set, state transition rules, and inherent constraints in the planning domain are semantically encoded to construct a planning domain capability feature vector. By calculating the semantic similarity between the user intent feature vector and the planning domain capability feature vector, and combining at least one of the following indicators, a planning domain intent matching score is generated: the coverage ratio of the user intent in the planning domain action set; the reachability probability of key actions; and the predicted value of the degree of constraint satisfaction. The generated score is weighted and merged with the comprehensive score obtained in the previous stage to calculate a new comprehensive score for each planning domain. Each new comprehensive score is then checked to see if it exceeds a preset threshold.If the conditions are not met, the planning domain is deemed to have failed the user intent constraint and is removed from the current candidate set, thereby generating a new set of current candidate financial business planning domains after further convergence. Furthermore, during intent decomposition, an intent gatekeeping and dynamic correction mechanism is introduced. Before parsing the user intent, an intent clarity assessment model is used to determine whether the current intent is clear. For example, is the length of the user's input text too short, does it contain explicit business keywords, or are there any ambiguous expressions? For vague or unclear intents, a user intent clarification process is triggered, guiding the user to supplement their explanation through interactive question-and-answer methods, such as asking specific questions like, "Do you mean you need to query sales data or inventory data?" Intent judgment can also be based on information from multiple sources, including the user's current natural language description; system context information such as the currently logged-in user role, the current operation page, and already filled form data; relevant business documents and policy descriptions retrieved from the knowledge base; and related tasks and confirmation results previously initiated by the user. This multi-source information is weighted and fused to form the final user intent representation. Understandably, user intents may become invalid or require adjustment as the environment changes. For example, if an atomic sub-intent cannot obtain corresponding action support after selecting a planning domain, it is automatically marked as a "sub-intent that cannot be fully satisfied." When the context changes, the validity of the parsed intents is re-evaluated, and invalid intents are marked or cleared. When encountering anomalies that the planning domain cannot handle during task execution, the original intent is corrected based on execution feedback to make it closer to the executable task objective. It's important to know that a confidence score can also be assigned to each parsed atomic sub-intent. Risk control measures are implemented based on the score. For example, when the confidence score is below a preset threshold (0.7), the intent is marked as a "low-confidence intent," and the following risk control measures are taken: for intents with medium confidence (0.5-0.7), logs are recorded and they are entered into a manual review queue; for intents with low confidence (0.5), execution is suspended and user confirmation is requested.

[0032] In the dimension of financial business risk and execution cost constraints, financial business risk cost information for each financial business planning domain in the current candidate financial business planning domain set is obtained. This financial business risk cost information includes a preset risk level, compliance attributes, and execution cost indicators. The financial business risk cost information is quantified into a penalty factor and introduced into the preset scoring function to obtain a target penalty scoring function. The comprehensive score of each financial business planning domain in the current candidate financial business planning domain set is reduced using the target penalty scoring function to obtain a reduced comprehensive score. It is determined whether the reduced comprehensive score corresponding to each financial business planning domain in the current candidate financial business planning domain set is greater than the preset threshold. If not, the corresponding financial business planning domain is determined not to meet the preset risk and cost constraints, and the corresponding financial business planning domain is removed from the current candidate financial business planning domain set to generate a new current candidate financial business planning domain set. Specifically, when multiple planning domains are close in intent matching scores, financial business risk cost information corresponding to each planning domain in the current candidate financial business planning domain set is obtained, and the planning domain score is adjusted punitively. Financial business risk cost information includes three core quantitative indicators: preset risk level, compliance attributes, and execution cost indicators. These multi-dimensional risk cost information are then quantified into a unified penalty factor. The design logic of this penalty factor is that the higher the risk level, the more sensitive the compliance attributes, and the greater the execution cost, the higher the penalty factor value. This penalty factor is introduced into a preset scoring function and modified to generate a target penalty scoring function specifically for this dimension. Using this target penalty scoring function, the accumulated comprehensive score of each planning domain in the current candidate financial business planning domain set is reduced. The degree of reduction is positively correlated with the magnitude of the penalty factor; that is, the comprehensive score of high-risk or high-cost planning domains will experience a greater reduction, thus obtaining the reduced comprehensive score for each planning domain. After score reduction, the reduced comprehensive score of each planning domain is compared with a preset threshold. If the reduced comprehensive score of a planning domain is not greater than the preset threshold, the planning domain is determined to have failed the compliance test of risk and execution cost, and it is removed from the current candidate set. After this dimension-based screening, a new, converged set of current candidate financial business planning domains is generated. The planning domains in this set not only meet user intent, business compliance, and permission requirements, but also prioritize the retention of solutions with controllable risks and reasonable costs under the same conditions.

[0033] Step S14: Determine the target financial business planning domain that meets the second preset condition from the set of candidate financial business planning domains.

[0034] In this embodiment, after determining a set of candidate financial business planning domains constrained by rules across multiple dimensions, the set of candidate financial business planning domains is sorted according to a preset order to generate a candidate list of target financial business planning domains. It is then determined whether the financial business planning domains in the candidate list of target financial business planning domains meet preset rule convergence conditions. These preset rule convergence conditions include any one or more of the following: a preset condition of large score differences, a preset condition of stable change magnitude, and a preset condition of consistent scores across multiple dimensions. If yes, the financial business planning domain in the candidate list of target financial business planning domains that meets the preset condition of highest score is directly determined as the target financial business planning domain. If not, the step of re-executing the rule constraint process based on the constraint information and a preset scoring function according to the constraint order of each preset constraint dimension to obtain the current set of candidate financial business planning domains that meets the first preset condition is executed. Specifically, after the score calculation is completed at each stage, the score ranking results of the candidate planning domains are analyzed. The candidate financial business planning domains in the candidate set are arranged according to a preset sorting order, such as descending order of comprehensive score, ascending order of risk level, or ascending order of execution cost, to generate a candidate list of target financial business planning domains. The planning domain selection process is considered convergent when the candidate list of target financial business planning domains meets at least one of the following conditions: The score difference between the highest-scoring planning domain and the second-best planning domain exceeds a preset threshold. For example, planning domain A scores 0.92, and planning domain B scores 0.76, with a difference of 0.16, exceeding the threshold of 0.15, thus indicating convergence; The change in planning domain scores in consecutive stages is lower than a stability threshold. For example, planning domain A scores 0.85 in the first stage and 0.88 in the second stage, with a change of 0.03, lower than the threshold of 0.05, indicating that the scores tend to stabilize; The score ranking results remain consistent across multiple stages. For example, the ranking in the first stage is A>B>C, and the ranking in the second stage remains A>B>C, without change. Upon successful convergence, the subsequent scoring phase is immediately terminated, and the planning domain with the best score in the candidate list of target financial business planning domains is determined as the target financial business planning domain. In the event of scoring anomalies or ties, a warning log is output, or a final decision is made according to preset priority rules.

[0035] Step S15: Load the target financial business planning domain into the initial artificial intelligence model to obtain the target financial business question-and-answer model, so that the user terminal can use the target financial business question-and-answer model to inquire about financial business.

[0036] In this embodiment, after determining the target financial business planning domain, the planning domain loading and runtime constraint injection operations are performed to load the target financial business planning domain into the initial artificial intelligence model. The planning domain loading process includes: parsing and verifying the definition file of the target financial business planning domain to ensure its syntactic correctness, completeness, and consistency; initializing the corresponding state space in the artificial intelligence runtime environment according to the state definition of the planning domain; registering the action interfaces defined in the planning domain to the planning engine so that the planning engine can recognize and call these actions; injecting the business rules, permission constraints, risk control rules, and other constraint rules in the planning domain into the runtime environment; constructing a runtime context containing information such as the current planning domain state space, available actions, and constraint rules; and adjusting the configuration parameters of the planning engine according to the characteristics of the current planning domain to optimize planning performance. After loading, the initial artificial intelligence model evolves into a specialized target financial business question-and-answer model. When the user interacts with the target financial business question-and-answer model, the obtained financial business inquiry service will strictly follow the preset financial business rules, adapt to the user's role and permissions, closely follow the user's intent, and operate within the boundaries of controllable risk and reasonable execution cost. The target financial business Q&A model can provide accurate, compliant, and personalized answers and operational guidance for specific business scenarios, thereby supporting users to efficiently complete financial business consultations and transactions.

[0037] As described above, this embodiment first obtains an initial AI model and determines the dimensions of preset constraints, while collecting constraint information related to these dimensions, such as user intent, financial business planning domain capabilities, permission constraints, and risk costs. Next, the initial model is adjusted at the rule level using this constraint information to form a set of candidate financial business planning domains, and a target financial business planning domain that meets the requirements is selected through a preset scoring function. Finally, the selected target financial business planning domain is integrated into the initial model to generate the final target financial business question-and-answer model for user-side business consultation. In this way, by constructing an AI-based multi-dimensional scoring model for financial business planning domains, the selection process of financial business planning domains is modeled as a computable, sortable, and adjudicable decision-making process. The target financial business planning domain is determined when the scoring results meet the convergence conditions, thus providing a stable and controllable technical constraint foundation for the subsequent planning and reasoning process.

[0038] As can be seen from the previous embodiment, this application discloses a financial business inquiry method based on an artificial intelligence model, which can stably and efficiently determine the target financial business planning domain. Next, taking intelligent wealth management in banks as an example, the financial business inquiry method based on an artificial intelligence model will be described in detail.

[0039] When a user with ordinary customer privileges asks the initial AI model, "What are some suitable stable investment options for me recently?", the system first analyzes the user's query to generate an initial product list, such as money market funds, bond funds, structured deposits, and trust products. Then, products ineligible for sale are filtered out based on financial business criteria; high-threshold products only available to private banking clients are eliminated based on role and permission criteria; products with clear stability attributes are prioritized through semantic matching based on user intent; and finally, high-risk, high-management-fee products are scored and reduced based on risk and cost. After multiple rounds of convergence, only two low-to-medium risk bond funds with suitable minimum investment amounts and reasonable fees are retained and loaded into the initial AI model to obtain the target financial business Q&A model. When the user subsequently consults through the target financial business Q&A model, they will only receive precise answers and comparisons related to these two products, rather than broad and vague investment recommendations.

[0040] See Figure 2 As shown in the figure, this invention discloses a financial business inquiry device based on an artificial intelligence model, which may specifically include: The constraint dimension determination module 11 is used to determine the preset constraint dimensions corresponding to the initial artificial intelligence model; The constraint information acquisition module 12 is used to acquire constraint information corresponding to the preset constraint dimension; wherein, the constraint information includes any one or more of user intent information, financial business planning domain capability information, permission constraint information, and financial business risk cost information. The set acquisition module 13 is used to apply rule constraints to the initial artificial intelligence model based on the constraint information for each of the preset constraint dimensions, so as to obtain a set of candidate financial business planning domains that meet the first preset conditions. The planning domain determination module 14 is used to determine the target financial business planning domain that meets the second preset condition from the set of candidate financial business planning domains. The model acquisition module 15 is used to load the target financial business planning domain into the initial artificial intelligence model to obtain the target financial business question-and-answer model, so that the user terminal can use the target financial business question-and-answer model to inquire about financial business.

[0041] As described above, this application first obtains an initial AI model and determines the dimensions of preset constraints, while collecting constraint information related to these dimensions, such as user intent, financial business planning domain capabilities, permission constraints, and risk costs. Next, it uses this constraint information to adjust the initial model at the rule level, forming a set of candidate financial business planning domains, and then filters out the target financial business planning domains that meet the requirements through a preset scoring function. Finally, the selected target financial business planning domains are integrated into the initial model to generate the final target financial business question-and-answer model for user-side business consultation. In this way, by constructing an AI-based multi-dimensional scoring model for financial business planning domains, the selection process of financial business planning domains is modeled as a computable, sortable, and adjudicable decision-making process. The target financial business planning domain is determined when the scoring results meet the convergence conditions, thus providing a stable and controllable technical constraint foundation for the subsequent planning and reasoning process.

[0042] In some specific implementations, the preset constraint dimensions include financial business constraint dimensions, role and permission constraint dimensions, user intent constraint dimensions, and financial business risk and execution cost constraint dimensions. Accordingly, the set acquisition module 13 may specifically include: The hierarchy determination unit is used to determine the rule convergence hierarchy corresponding to each of the preset constraint dimensions; The sequence acquisition unit is used to acquire the constraint order for applying rule constraints to the initial artificial intelligence model according to the rule convergence level; The request acquisition unit is used to acquire financial business inquiry requests sent by the user client. A set determination unit is used to determine an initial set of financial business planning domains based on the financial business inquiry request. The set acquisition unit is used to apply rule constraints to the initial set of financial business planning domains according to the constraint order of each of the preset constraint dimensions, based on the constraint information and the preset scoring function, to obtain the current set of candidate financial business planning domains that meet the first preset condition.

[0043] In some specific implementations, the base set determination unit may specifically include: The first feature vector generation subunit is used to perform semantic parsing on the financial business inquiry request through a preset artificial intelligence semantic coding model to generate the corresponding target business semantic feature vector. The mapping table determination subunit is used to determine the preset mapping table; the preset mapping table is used to store the mapping relationship between the preset financial business planning domain and the business semantic feature vector. The set determines the sub-unit, which is used to determine the initial financial business planning domain set corresponding to the target business semantic feature vector from the preset financial business planning domain set based on the preset mapping table.

[0044] In some specific implementations, the set acquisition unit may specifically include: The first scoring acquisition subunit is used to score the initial financial business planning domain set according to a preset scoring function in the financial business constraint dimension, so as to obtain the first score corresponding to each financial business planning domain in the initial financial business planning domain set. The first scoring judgment subunit is used to determine whether the first score corresponding to each financial business planning domain in the initial financial business planning domain set is greater than a preset threshold. The first set generates sub-units, which, if so, integrate the corresponding financial business planning domain into the current candidate financial business planning domain set.

[0045] In some specific implementations, the set acquisition unit may specifically include: The information acquisition subunit is used to acquire the user's permission constraint information; The attribute determination subunit is used to determine the corresponding role attribute based on the permission constraint information in the role and permission constraint dimension. The scoring generation subunit is used to determine the matching degree between the role attribute and each financial business planning domain in the current candidate financial business planning domain set, and to generate a corresponding second score based on the matching degree using the preset scoring function; The second scoring acquisition subunit is used to weight and fuse the second scores of each financial business planning domain with the corresponding second scores to obtain the comprehensive score of each financial business planning domain. The second scoring and judgment subunit is used to determine whether the comprehensive score corresponding to each financial business planning domain in the current candidate financial business planning domain set is greater than the preset threshold. The second set generation sub-unit is used to determine, if not, that the corresponding financial business planning domain does not meet the preset role constraint conditions, and to remove the corresponding financial business planning domain from the current candidate financial business planning domain set to generate a new current candidate financial business planning domain set.

[0046] In some specific implementations, the set acquisition unit may specifically include: The information decomposition subunit is used to decompose the current user intent information into several atomic sub-intents using the intent decomposition algorithm of the preset artificial intelligence semantic encoding model in the user intent constraint dimension. An intent sequence generation subunit is used to establish the logical relationship between the atomic sub-intents and generate a structured atomic sub-intent sequence based on the logical relationship; The intent parsing subunit is used to parse the decomposed atomic sub-intents into intent semantic feature vectors based on the atomic sub-intent sequence; The second feature vector generation subunit is used to obtain the action set, state transition rules and inherent constraints of the current candidate financial business planning domain set, and to semantically encode the action set, state transition rules and inherent constraints through the preset artificial intelligence semantic encoding model to generate the financial business planning domain capability feature vector of each financial business planning domain in the current candidate financial business planning domain set. The similarity determination subunit is used to determine the similarity between the intent semantic feature vector and the financial business planning domain capability feature vector; The scoring determination subunit is used to determine the third score of each financial business planning domain in the current candidate financial business planning domain set based on the similarity using the preset scoring function; The third scoring acquisition subunit is used to weight and fuse the third scores of each financial business planning domain with the corresponding comprehensive scores to obtain a new comprehensive score. The third scoring and judgment subunit is used to determine whether the new comprehensive score of each financial business planning domain in the current candidate financial business planning domain set is greater than the preset threshold. The third set generation subunit is used to determine, if not, that the corresponding financial business planning domain does not meet the preset user intent constraint conditions, and to remove the corresponding financial business planning domain from the current candidate financial business planning domain set to generate a new current candidate financial business planning domain set.

[0047] In some specific implementations, the set acquisition unit may specifically include: The cost information acquisition subunit is used to acquire financial business risk cost information for each financial business planning domain in the current set of candidate financial business planning domains, based on the dimensions of financial business risk and execution cost constraints. The financial business risk cost information includes preset risk level, compliance attributes, and execution cost indicators. The scoring function acquisition sub-unit is used to quantify the financial business risk cost information into a penalty factor and introduce it into the preset scoring function to obtain the target penalty scoring function; The fourth scoring acquisition subunit is used to reduce the comprehensive score of each financial business planning domain in the current candidate financial business planning domain set using the target punitive scoring function, so as to obtain the reduced comprehensive score. The fourth scoring and judgment subunit is used to determine whether the reduced comprehensive score corresponding to each financial business planning domain in the current candidate financial business planning domain set is greater than the preset threshold. The fourth set generation sub-unit is used to determine, if not, that the corresponding financial business planning domain does not meet the preset risk and cost constraints, and to remove the corresponding financial business planning domain from the current candidate financial business planning domain set to generate a new current candidate financial business planning domain set.

[0048] In some specific implementations, the planning domain determination module 14 may specifically include: The list generation unit is used to sort the candidate financial business planning domain set according to a preset order and generate a candidate list of target financial business planning domains. The planning domain judgment unit is used to determine whether the financial business planning domains in the candidate list of target financial business planning domains meet the preset rule convergence conditions; the preset rule convergence conditions include any one or more of the preset score difference conditions, preset change range stable conditions, and preset multi-dimensional score consistency conditions. The first planning domain determination unit is used to directly determine the financial business planning domain that meets the preset highest score condition in the candidate list of target financial business planning domains as the target financial business planning domain if the condition is met. The second planning domain determination unit is used to, if not, re-execute the step of applying rule constraints to the initial financial business planning domain set based on the constraint information and the preset scoring function according to the constraint order of each preset constraint dimension, to obtain the current candidate financial business planning domain set that satisfies the first preset condition.

[0049] Furthermore, embodiments of this application also disclose an electronic device, Figure 3 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.

[0050] Figure 3 This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the financial business inquiry method based on an artificial intelligence model disclosed in any of the foregoing embodiments. Alternatively, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0051] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0052] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0053] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the artificial intelligence model-based financial business inquiry method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0054] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned disclosed financial business inquiry method based on an artificial intelligence model. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0055] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0056] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0057] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0058] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0059] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A financial business inquiry method based on an artificial intelligence model, characterized in that, The method comprises the following steps: determining preset constraint dimensions corresponding to an initial artificial intelligence model; obtaining constraint information corresponding to the preset constraint dimensions; wherein the constraint information comprises any one or several of user intention information, financial business planning domain capability information, permission constraint information and financial business risk cost information; performing rule constraint on the initial artificial intelligence model in each of the preset constraint dimensions based on the constraint information, to obtain a candidate financial business planning domain set satisfying a first preset condition; determining a target financial business planning domain satisfying a second preset condition from the candidate financial business planning domain set; loading the target financial business planning domain to the initial artificial intelligence model to obtain a target financial business question and answer model, so that a user end uses the target financial business question and answer model to inquire about a financial business. 2.The AI model-based financial service inquiry method of claim 1, wherein, The preset constraint dimensions comprise financial business constraint dimensions, role and permission constraint dimensions, user intention constraint dimensions, financial business risk and execution cost constraint dimensions. Correspondingly, the rule constraint on the initial artificial intelligence model in each of the preset constraint dimensions based on the constraint information to obtain a candidate financial business planning domain set satisfying a first preset condition comprises: determining rule convergence levels corresponding to each of the preset constraint dimensions; obtaining a constraint order for performing rule constraint on the initial artificial intelligence model according to the rule convergence levels; obtaining a financial business inquiry request sent by a user end; determining an initial financial business planning domain set based on the financial business inquiry request; performing rule constraint on the initial financial business planning domain set based on the constraint information and a preset scoring function according to the constraint order of each of the preset constraint dimensions, to obtain a current candidate financial business planning domain set satisfying a first preset condition. 3.The AI model-based financial service inquiry method of claim 2, wherein, The determination of the initial financial business planning domain set based on the financial business inquiry request comprises: performing semantic analysis on the financial business inquiry request by a preset artificial intelligence semantic coding model to generate a corresponding target business semantic feature vector; determining a preset mapping table; the preset mapping table is used to store a mapping relationship between preset financial business planning domains and business semantic feature vectors; determining an initial financial business planning domain set corresponding to the target business semantic feature vector from the preset financial business planning domain set based on the preset mapping table. 4.The AI model-based financial service inquiry method of claim 2, wherein, The rule constraint on the initial financial business planning domain set based on the constraint information and a preset scoring function according to the constraint order of each of the preset constraint dimensions to obtain a current candidate financial business planning domain set satisfying a first preset condition comprises: scoring the initial financial business planning domain set according to a preset scoring function in the financial business constraint dimension, to obtain a first score corresponding to each of the initial financial business planning domain set; determining whether the first score corresponding to each of the initial financial business planning domain set is greater than a preset threshold value; if yes, integrating the corresponding financial business planning domain into the current candidate financial business planning domain set. 5.The AI model-based financial service inquiry method of claim 4, wherein, The step of applying rule constraints to the initial set of financial business planning domains based on the constraint information and a preset scoring function according to the constraint order of each preset constraint dimension, to obtain a current set of candidate financial business planning domains that satisfy the first preset condition, includes: Obtain permission constraint information from the user's client; In the role and permission constraint dimension, the corresponding role attributes are determined based on the permission constraint information; Determine the matching degree between the role attribute and each financial business planning domain in the current candidate financial business planning domain set, and generate a corresponding second score based on the matching degree using the preset scoring function; The second score of each financial business planning domain is weighted and fused with the corresponding first score to obtain the comprehensive score of each financial business planning domain. Determine whether the comprehensive score corresponding to each financial business planning domain in the current candidate financial business planning domain set is greater than the preset threshold. If not, the corresponding financial business planning domain is determined to not meet the preset role constraint conditions, and the corresponding financial business planning domain is removed from the current candidate financial business planning domain set to generate a new current candidate financial business planning domain set. 6.The AI model-based financial service inquiry method of claim 5, wherein, The step of applying rule constraints to the initial set of financial business planning domains based on the constraint information and a preset scoring function according to the constraint order of each preset constraint dimension, to obtain a current set of candidate financial business planning domains that satisfy the first preset condition, includes: In the user intent constraint dimension, the intent decomposition algorithm of the preset artificial intelligence semantic encoding model is used to decompose the current user intent information into several atomic sub-intents; Establish the logical relationships between the atomic sub-intents, and generate a structured sequence of atomic sub-intents based on the logical relationships; Based on the atomic sub-intention sequence, the decomposed atomic sub-intention is parsed into an intention semantic feature vector; The action set, state transition rules, and inherent constraints of the current candidate financial business planning domain set are obtained, and the action set, state transition rules, and inherent constraints are semantically encoded through the preset artificial intelligence semantic encoding model to generate the financial business planning domain capability feature vector of each financial business planning domain in the current candidate financial business planning domain set. Determine the similarity between the intent semantic feature vector and the financial business planning domain capability feature vector; The third score of each financial business planning domain in the current candidate financial business planning domain set is determined based on the similarity using the preset scoring function. The third score of each financial business planning domain is weighted and merged with the corresponding comprehensive score to obtain a new comprehensive score; Determine whether the new comprehensive score of each financial business planning domain in the current candidate financial business planning domain set is greater than the preset threshold; If not, the corresponding financial business planning domain is determined to not meet the preset user intent constraints, and the corresponding financial business planning domain is removed from the current candidate financial business planning domain set to generate a new current candidate financial business planning domain set. 7.The AI model-based financial service inquiry method of claim 6, wherein, The step of applying rule constraints to the initial set of financial business planning domains based on the constraint information and a preset scoring function according to the constraint order of each preset constraint dimension, to obtain a current set of candidate financial business planning domains that satisfy the first preset condition, includes: In the dimension of financial business risk and execution cost constraints, obtain the financial business risk cost information of each financial business planning domain in the current set of candidate financial business planning domains; wherein the financial business risk cost information includes preset risk level, compliance attribute and execution cost indicator; The financial business risk cost information is quantified into a penalty factor and introduced into the preset scoring function to obtain the target penalty scoring function; The comprehensive score of each financial business planning domain in the current candidate financial business planning domain set is reduced using the target punitive scoring function to obtain the reduced comprehensive score; Determine whether the reduced comprehensive score corresponding to each financial business planning domain in the current candidate financial business planning domain set is greater than the preset threshold. If not, the corresponding financial business planning domain is determined to not meet the preset risk and cost constraints, and the corresponding financial business planning domain is removed from the current candidate financial business planning domain set to generate a new current candidate financial business planning domain set. 8.The AI model-based financial service inquiry method according to any one of claims 2 to 7, characterized in that, The step of determining the target financial business planning domain that meets the second preset condition from the set of candidate financial business planning domains includes: The candidate financial business planning domain set is sorted according to a preset order to generate a candidate list of target financial business planning domains; Determine whether the financial business planning domains in the candidate list of target financial business planning domains meet the preset rule convergence conditions; the preset rule convergence conditions include any one or more of the preset conditions for large score differences, preset conditions for stable change amplitude, and preset conditions for consistent multi-dimensional scores. If so, the financial business planning domain that meets the preset highest score condition in the candidate list of target financial business planning domains will be directly determined as the target financial business planning domain. If not, then the step of applying rule constraints to the initial set of financial business planning domains according to the constraint order of each of the preset constraint dimensions, based on the constraint information and the preset scoring function, to obtain the current set of candidate financial business planning domains that satisfies the first preset condition is executed again. 9.A financial service inquiry device based on an artificial intelligence model, characterized by, include: The constraint dimension determination module is used to determine the preset constraint dimensions corresponding to the initial artificial intelligence model; The constraint information acquisition module is used to acquire constraint information corresponding to the preset constraint dimension; wherein, the constraint information includes any one or more of the following: user intent information, financial business planning domain capability information, permission constraint information, and financial business risk cost information. The set acquisition module is used to apply rule constraints to the initial artificial intelligence model based on the constraint information for each of the preset constraint dimensions, so as to obtain a set of candidate financial business planning domains that satisfy the first preset condition. The planning domain determination module is used to determine the target financial business planning domain that meets the second preset condition from the set of candidate financial business planning domains. The model acquisition module is used to load the target financial business planning domain into the initial artificial intelligence model to obtain the target financial business question-and-answer model, so that the user terminal can use the target financial business question-and-answer model to conduct financial business inquiries.

10. An electronic device, comprising: include: Memory, used to store computer programs; A processor for executing the computer program to implement the financial business inquiry method based on an artificial intelligence model as described in any one of claims 1 to 8.

11. A computer readable storage medium, characterized in that, Used to store computer programs; wherein, when the computer programs are executed by a processor, they implement the financial business inquiry method based on an artificial intelligence model as described in any one of claims 1 to 8.

Images