A method, computer device and computer program product for unified processing of index dimensions for real-time interaction

Abstract

This application provides a method, computer device, and computer program product for unified processing of indicator definitions for real-time interaction, belonging to the field of artificial intelligence technology. The method includes: receiving and parsing a user's verbal request and generating a structured contextual state vector; based on the contextual state vector, obtaining multiple candidate denominator definitions and their basis from a definition knowledge base that match the verbal request; generating a set of constraint strategies based on the contextual state vector, the candidate denominator definitions, and their basis; based on the set of constraint strategies, performing consistency adjudication through the candidate denominator definitions to determine a unique denominator definition; generating a finalized indicator description format based on the unique denominator definition, and embedding a restriction rule for prohibiting the visual comparison of inconsistent indicators into the finalized indicator description format; rendering the finalized indicator description format and its restriction rule on a data visualization interface, and intercepting and guiding parallel display operations that violate the restriction rule in real time.

Description

Technical Field

[0001] This application belongs to the field of computer artificial intelligence technology, and in particular relates to a unified processing method for indicators for real-time interaction, computer equipment, and computer program products. Background Technology

[0002] During cross-departmental meetings, people often argue fiercely over the same data metric. For example, when looking at "success rate," Department A uses "total number of applications" as the denominator, while Department B uses "number of applications that passed the initial review" as the denominator. The results are simply incomparable. If a responsible person asks a question casually during the meeting and crams the results of two different algorithms into the same chart, it is easy to draw incorrect conclusions.

[0003] Existing solutions, such as building data dictionaries, providing prompts within the system, or using AI question answering to explain statements, mostly only remain at the "giving suggestions" stage. They cannot capture the context of the meeting (such as what topics are being discussed, who is asking questions, and which algorithms have been used previously), let alone forcibly prevent the mixing and comparison of data from different algorithms when generating charts.

[0004] Furthermore, when the system is unsure which algorithm to choose, it may repeatedly ask questions, disrupting the meeting's flow; and even after preventing erroneous comparisons, it lacks intelligent alternative display solutions. This makes it difficult to truly implement a "unified approach" in tense, real-world meetings, and the decision-making process remains risky. Summary of the Invention

[0005] This application provides a unified processing method, computer device, and computer program product for real-time interactive indicators, which can solve the current technical problem that it is impossible to automatically lock the data indicator calculation standard and forcibly prevent incorrect chart comparisons in real-time meetings.

[0006] In a first aspect, this application provides a method for unifying the metric standards for real-time interaction, the method being executed by a computing device and including the following steps:

[0007] S1: Receive and parse the interaction context information, obtain the verbal request, generate a structured context state vector, and write the context state vector as meeting history information into the aperture state cache;

[0008] S2: Based on the context state vector, obtain multiple candidate denominator calibers that match the oral request from the caliber knowledge base, as well as the basis for the candidate denominator calibers;

[0009] S3: Based on the context state vector, the candidate denominator caliber, and the basis for each candidate denominator caliber, a constraint strategy set is generated. The constraint strategy set is used to align with the meeting history information in the caliber state cache to determine a unique denominator caliber from the multiple candidate denominator calibers.

[0010] S4: Based on the set of constraint strategies, the unique denominator is determined by performing a consistency decision on the candidate denominator categories.

[0011] S5: Based on the unique denominator, generate a finalized indicator description format and embed a restriction rule for prohibiting the visual comparison of inconsistent indicators into the finalized indicator description format.

[0012] S6: Based on the finalized indicator description format and the embedded restriction rules, render the data on the data visualization interface, and intercept and guide parallel display operations that violate the restriction rules in real time.

[0013] The technical effect of this application's embodiments is that existing systems cannot automatically determine a unified standard for calculating indicators in real-time interactive scenarios and force the execution of this standard in the front-end visualization stage, leading to incorrect data comparisons. To address the above-mentioned technical deficiencies of existing systems, the technical solution of this application's embodiments achieves an automated closed loop between the "decision-making" (adjudication) and "execution" (interception) of the denominator caliber within the system, fundamentally eliminating data visualization errors caused by inconsistent standards.

[0014] In one embodiment, the interaction context information includes the spoken text, agenda, speaker role, business stage, organizational scope, and time window.

[0015] This embodiment clarifies the specific composition of "interaction context information," addressing the issue of how the system can comprehensively and accurately perceive and understand real-time meeting scenarios. By precisely defining the source of the context state, it ensures the accuracy and relevance of subsequent retrieval, adjudication, and display processes, providing rich and reliable contextual evidence for the entire method.

[0016] In one embodiment, step S2 further includes: based on the context state vector, using a retrieval-enhanced generation model to parse the spoken request into a draft index description format, and retrieving candidate denominator calibers and their basis from a caliber knowledge base to form candidate denominator calibers and the basis for the candidate denominator calibers; step S3 further includes: based on the context state vector, the plurality of candidate denominator calibers, and the basis for each candidate denominator caliber, generating a set of policy constraints, the set of policy constraints including hard constraints, preference ranking, and a first clarification threshold, and aligning the set of policy constraints with the caliber state cache to lock a preferred denominator caliber and comparison boundary.

[0017] This embodiment further clarifies the core technical means of steps S2 and S3, solving the problems of how to generate structured candidates from fuzzy verbal requests and how to dynamically generate intelligent adjudication strategies. Specific techniques such as retrieval enhancement generation and policy constraint sets are introduced to ensure the accuracy and comprehensiveness of the candidate set and to make the adjudication process adaptive and interpretable, significantly improving the system's intelligence level.

[0018] In one embodiment, step S4 further includes:

[0019] Based on the set of policy constraints, a consensus decision is made on each of the candidate denominator categories, including: the domain agent, consensus agent, and compliance agent of the multi-agent system reviewing each of the candidate denominator categories respectively; and the arbitration agent summarizing the review results of each agent and making a decision with the hard constraints as the priority principle.

[0020] In adjudicating the various review results, it is determined whether the difference between the review results of the top two candidate denominators is lower than the aforementioned clarification threshold:

[0021] If the threshold for clarification is lower than the threshold, only one clarification question will be sent to the user, and the arbitration agent will be assisted in making a ruling based on the user's answer.

[0022] If the threshold is not lower than the threshold, the arbitration agent will directly determine the unique denominator based on the review results of each agent, the set of policy constraints, and the candidate denominator caliber, and at the same time generate an alternative display scheme that can be used when a comparison conflict occurs.

[0023] Step S5 further includes: based on the unique denominator caliber and the alternative display scheme, writing the denominator caliber identifier and caliber description into the indicator description format draft to generate the final indicator description format, locking the denominator caliber identifier and caliber description, and simultaneously generating and injecting a comparison restriction rule.

[0024] This embodiment addresses the challenge of efficiently and reliably selecting the single best solution from multiple excellent candidates in complex scenarios, while avoiding the pitfalls of multiple rounds of inefficient clarification. Professional division of labor in the review process through a multi-agent system ensures the comprehensiveness (domain-specific), consistency (historical context), and compliance (regulatory requirements) of the decision. A one-time clarification threshold mechanism minimizes human-computer interaction rounds while guaranteeing the accuracy of the decision, ensuring smooth real-time meetings. The pre-generation of alternative display solutions demonstrates the system's forward-thinking approach and user experience.

[0025] In one embodiment, step S6 further includes: displaying the data on the screen based on the finalized indicator description format, intercepting parallel operations across denominator calibers according to the comparison restriction rules, guiding the adoption of the alternative display scheme, and simultaneously writing the unique denominator caliber and the interception event back to the caliber status cache and archiving it.

[0026] This embodiment addresses the disconnect between decision-making and execution, ensuring that decisions are truly implemented and forming a traceable closed loop. It combines mandatory interception, alternative guidance, and audit write-back, effectively preventing erroneous operations at the front end. Furthermore, by recording decision and interception logs, the entire process becomes auditable, reviewable, and optimizable, achieving a closed loop of governance.

[0027] In one embodiment, step S3 generates a set of policy constraints based on the context state vector, the multiple candidate denominator calibers, and the basis for the candidate denominator calibers, including:

[0028] Read the target field, stage field, role field, query boundary field, and caliber reference field from the context state vector;

[0029] Generate hard constraints that match the target field and the stage field, and write the query boundary field into the range of the hard constraints; wherein, the hard constraints represent the set of rules that need to be followed under the context state vector;

[0030] A preference ranking is formed based on the role field and the target field; wherein, the preference ranking represents the sorting configuration of the candidate denominator caliber;

[0031] A clarification threshold is defined by combining the multiple candidate denominator calibers and the difference between the top two calibers in the ranking of each candidate denominator caliber.

[0032] A set of policy constraints is obtained, which includes the hard constraints, the preference ranking, and the first clarification threshold;

[0033] The policy constraint set is compared with the meeting history information in the caliber state cache, so that the output of the policy constraint set includes the locking priority denominator caliber and its comparison boundary.

[0034] This embodiment refines the generation logic of the policy constraint set, solving the problem of how the system transforms unstructured meeting contexts into specific, executable adjudication strategies. It reveals the internal logic of policy generation, demonstrating the rationality and transparency of the system's decision-making. Through comparison and alignment with the caliber state cache, it emphasizes the core design of maintaining the continuity of caliber within the meeting, which is crucial for achieving "real-time consistency."

[0035] In one embodiment, in step S4:

[0036] The domain agent is a processing unit that verifies the applicable scenarios and effective period based on the aforementioned knowledge base;

[0037] The consistency proxy shown is a processing unit that checks intra-session consistency and comparison boundaries based on the aforementioned caliber state cache.

[0038] The compliance agent is a processing unit that checks whether the candidate denominator caliber is compatible with the caliber knowledge base;

[0039] The arbitration agent is a processing unit that selects a unique denominator under the principle of prioritizing hard constraints;

[0040] The hard constraints are configured to prioritize the decision principle of executing hard constraints first and then preference ranking in the policy constraint set.

[0041] This embodiment clarifies the division of responsibilities among agents in a multi-agent system, addressing potential biases or insufficient capabilities of single models or modules in complex adjudication tasks. Through the collaborative work of functionally specialized agents, a robust and reliable adjudication architecture is constructed. The explicit prioritization of hard constraints ensures absolute adherence to business rules and compliance bottom lines, enhancing the overall system reliability.

[0042] In one embodiment, the unique denominator caliber in step S4 represents a single denominator definition determined under the candidate denominator caliber, the basis of the candidate denominator caliber, and the hard constraints of the policy constraint set.

[0043] The unique denominator is obtained in the following way:

[0044] The domain agent filters the candidate denominator based on the hard constraints in the policy constraint set, considering the applicable scenarios and effective periods.

[0045] The consistency agent uses the caliber state cache as a reference to filter out candidate denominator calibers that conflict with the comparison boundary of the adopted denominator caliber, and marks the candidate denominator calibers that can be displayed in parallel.

[0046] The compliance agent removes candidate denominator calibers that do not conform to the adaptation rules of the caliber knowledge base;

[0047] Based on the above results, the arbitration agent shall select a unique denominator based on the principle of prioritizing hard constraints and the aforementioned preference ranking.

[0048] When the clarification threshold is triggered, the arbitration agent raises a clarification question, updates the sorting and completes the selection after receiving the answer, and outputs a unique denominator for use in the finalization of the subsequent indicator description format.

[0049] This embodiment provides a specific and operable execution path for determining the unique denominator, solving the problems of transparency and repeatability in the final decision of arbitration. It clearly defines the complete pipeline from candidate filtering and conflict detection to final arbitration, making the "consistent ruling" no longer a black box. It organically integrates clarification into the decision-making process, ensuring efficient convergence to an optimal solution even in ambiguous situations.

[0050] Secondly, this application also provides a computer device, including a processor and a memory, the memory storing a computer program, which, when executed by the computer device, performs the steps of the unified indicator processing method for real-time interaction as described in the first aspect.

[0051] Thirdly, this application also provides a computer program product, which stores a computer program. When the computer program is executed by a computer device, it runs the unified processing method for real-time interactive indicators as described in the first aspect.

[0052] For the technical effects of the second and third aspects mentioned above, please refer to the description of the technical effects of the unified processing method for real-time interactive indicators in the first aspect mentioned above, which will not be repeated here. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application, 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 A schematic flowchart illustrating an embodiment of a method for unifying the caliber of indicators for real-time interaction provided in this application;

[0055] Figure 2 A flowchart illustrating the unique method for obtaining the denominator for this application;

[0056] Figure 3 This is a structural block diagram of an embodiment of a computer device provided in this application. Detailed Implementation

[0057] 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 pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0058] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0059] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0060] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), unless otherwise expressly and specifically defined.

[0061] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0062] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0063] In the fields of data analytics and artificial intelligence, "caliber" refers to the "calculation standard" or "statistical definition." It specifically refers to the explicit definition of the components when calculating an indicator (especially a ratio or average).

[0064] The "denominator" mentioned in this application can be illustrated by the following example:

[0065] For example, the indicator is the handling rate; the numerator is the number of events that have been handled.

[0066] Denominator: The point of contention regarding the "denominator's scope":

[0067] Scope A (which may be used by department A, for example, if department A is the marketing department, the marketing department may use Scope A to evaluate advertising effectiveness):

[0068] Conversion rate = Number of users who completed a purchase / Number of users who clicked on the ad

[0069] Meaning: The denominator is "number of users who clicked on the ad," which measures the efficiency from ad reach to final purchase.

[0070] Scope B (which may be used by department B, for example, if department B is the sales department, the sales department may use Scope B to evaluate the overall effectiveness of the platform):

[0071] Conversion rate = Number of users who completed a purchase / Number of unique website visitors

[0072] Meaning: The denominator is "unique website visitors", which measures the overall conversion rate of the entire site's traffic.

[0073] Scope C (This may be used by department C; for example, if department C is the finance department, the finance department may use Scope C for cost accounting):

[0074] Conversion rate = Number of users who completed a purchase / Number of valid leads

[0075] Meaning: The denominator is "number of valid leads", which measures the conversion efficiency from sales leads to sales.

[0076] Therefore, when a meeting presents both the "handling rate" of events in Department A (based on caliber A) and the "handling rate" of events in Department B (based on caliber B), although both indicators are called "handling rate," these two figures cannot be directly compared because the denominators are different. Forcing a comparison between them will lead to incorrect conclusions.

[0077] It is understandable that in large-scale meetings using existing technologies, the same metric often leads to misunderstandings due to differences in denominator definitions (such as UV deduplication, reach, order base, etc.), inconsistent organizational scope, and time windows. Ad-hoc queries and parallel presentations triggered by verbal statements during meetings can easily mix different denominator definitions within the same chart or table column, resulting in incomparable conclusions. Traditional data dictionaries and metric libraries are mostly governed and constrained offline outside of meetings, making it difficult to implement the "unique denominator definition" and comparison boundaries in real-time meeting interactions. They also lack guidance on alternative presentations and auditable closed loops in case of conflicts. As large-scale model-driven natural language analysis is used for ad-hoc Q&A, differences in definitions are rapidly amplified. Without in-meeting context binding and mandatory interception, erroneous comparisons are even more difficult to detect in advance.

[0078] Existing technologies primarily mitigate inconsistencies in terminology through three paths. First, data governance and a central / semantic layer for metrics: This involves establishing a unified terminology and alias mapping through data catalogs, metric dictionaries, lineage and version management. Quality rules are applied during the ETL / modeling phase, or metric selection and explanations are provided in BI, indicating the scope of application and effective period. Second, knowledge graphs and enhanced retrieval: A terminology knowledge base is built, supporting search / question-answering and returning definitions, sources, and scenario tags to assist analysts in manual alignment. Some solutions use RAG to convert verbal requests into SQL or visual queries, including terminology explanations in the responses. Third, multi-agent systems and orchestration: Planning / consistency checks / compliance review agents are used to verify generated results, reducing illusions and conflicts. Some products also set lightweight rules on the front end, such as recommending terms based on roles or providing prompts when merging tables across domains.

[0079] The existing solutions mentioned above are relatively mature in terms of out-of-meeting governance and knowledge provision, but they mostly remain at the level of "suggestions / tips".

[0080] In real-time discussion and presentation scenarios in cross-departmental meetings and command centers, the main objective of this invention is to construct a mechanism for aligning and comparing meeting criteria using contextual state vectors as anchors. This mechanism would parse verbal requirements into constrained indicator descriptions, enabling the real-time locking of a unique denominator within the meeting, injecting and forcibly intercepting the rule of "prohibiting cross-denominator comparisons in the same graph / column," and providing executable alternative presentations and auditable write-backs in case of conflicts. This would form a consistent, interpretable, and traceable closed loop.

[0081] To address the aforementioned deficiencies, in a first aspect, embodiments of this application provide a method for unified processing of metrics for real-time interaction. The method is executed by a computing device and includes the following steps S1 to S6: S1: Receive and parse interaction context information to obtain a verbal request, generate a structured context state vector, and write the context state vector as meeting history information into the metric state cache.

[0082] It should be noted that the interaction context information in this embodiment may include the spoken text, agenda, speaker role, business stage, organizational scope, and time window;

[0083] The document includes the following components: speech text (a transcript of spoken content from the meeting); agenda (a structured description of the meeting's pre-set themes and order); speaker role (an identifier of their organizational function within the meeting); business stage (a marker of the current business process); organizational scope (a defined scope of departments or channels involved); time window (the time interval for statistical analysis of indicators); adopted denominator (the definition of the denominator used in this meeting); denominator status cache (an in-meeting cache recording the adopted denominator and comparison boundaries); and parsed contextual state vector (a set of fields that combine the above information for subsequent adjudication).

[0084] In practical application, this embodiment uses an "emergency command large screen system" as an example. For instance, in a flood control emergency command meeting, the commander (i.e., the speaker) queries data and makes decisions in front of the large screen. The commander says, "Let me see the completion rate of mass evacuation in each district and county." The system captures this speech in real time and converts it into spoken text. At the same time, the system reads the current meeting agenda as "Summary of Emergency Response Phase," the commander's role as "Commander-in-Chief," the operational phase as "Level 1 Response," the organizational scope as "all districts and counties in the city," and the time window as "during this flood disaster." The meeting previously discussed "saturation of resettlement sites," the denominator of which is "capacity of resettlement sites already in use." The system structures this information, generates a contextual state vector, and records "capacity of resettlement sites already in use" in the caliber state cache as the anchor denominator caliber for the current meeting.

[0085] S2: Based on the context state vector, obtain multiple candidate denominator calibers that match the oral request from the caliber knowledge base, as well as the basis for the candidate denominator calibers;

[0086] In this embodiment of the application, based on the context state vector, a retrieval-enhanced generation model can be used to parse the spoken request into a draft index description format, and candidate denominator calibers and their basis can be retrieved from a caliber knowledge base to form candidate denominator calibers and the basis for the candidate denominator calibers:

[0087] The verbal request can be parsed, for example, the verbal request can be mapped to a set of indicator intent fields, including indicator name, numerator definition trigger word, denominator qualifier, time window qualifier and filter condition statement;

[0088] For example, the "retrieval enhancement generation technology" can be a model that combines retrieval and generation, using the contextual state vector as the retrieval condition, calling the "caliber knowledge base" to return evidence descriptions, and generating a draft index description format and candidate denominator calibers and the basis for the candidate denominator calibers based on the evidence descriptions; the basis includes the evidence descriptions and the reasons for the adaptation.

[0089] Multiple candidate denominator definitions represent a set of candidate denominator definitions. The output represents the supporting evidence and justification for each candidate. The core components of the "evidence" are objective facts extracted from the knowledge base, such as the applicable scenario, effective period, and source basis. The "justification" is an explanatory text generated based on this and the current meeting context.

[0090] The "caliber knowledge base" can be a database that stores data on the definition of denominator calibers, such as denominator caliber definition, applicable scenario annotation, effective period, source basis and alias mapping; the candidate and basis output is a set of candidate denominator calibers and the evidence explanation and fitting reason for each candidate;

[0091] In practical applications, this embodiment of the application, based on the context (contextual state vector) of step S1, understands that the commander requires the indicator of "transfer completion rate". It can use a Retrieval Enhanced Generation (RAG) model to retrieve possible denominator definitions for "transfer completion rate" from a knowledge base, returning multiple candidates: Candidate A: Denominator = Total population to be transferred (estimated based on a risk model); Candidate B: Denominator = Population for whom transfer orders have been issued (based on command system records); Candidate C: Denominator = Actual registered disaster-affected population (based on reports from grassroots levels). Each candidate is accompanied by evidence such as its source and applicable scenarios.

[0092] S3: Based on the context state vector and the candidate denominator caliber, generate a set of constraint strategies, which are used to determine a unique denominator caliber from the multiple candidate denominator calibers.

[0093] In this embodiment of the application, a set of policy constraints can be generated based on the context state vector, the multiple candidate denominator calibers, and the basis for the candidate denominator calibers. The set of policy constraints includes hard constraints, preference ranking, and a first clarification threshold, and is aligned with the caliber state cache to lock a preferred denominator caliber and comparison boundary.

[0094] The process of generating the policy constraint set includes:

[0095] Based on multiple candidate denominator calibers and the criteria upon which these calibers are based, a high-level policy of hierarchical reinforcement learning generates hard constraints, preference rankings, and a first-clarification threshold according to the contextual state vector, and aligns these with the meeting history information in the caliber state cache. Specifically, the process involves: first, reading the target field, stage field, role field, query boundary field, and caliber reference field from the contextual state vector; then generating hard constraints matching the target field and stage field, and writing the query boundary field into the hard constraint range; next, forming a preference ranking based on the role field and the target field; subsequently, defining a first-clarification threshold by combining multiple candidate denominator calibers and the difference between the top two calibers in their rankings; finally, the resulting policy constraint set includes hard constraints, preference rankings, and a first-clarification threshold.

[0096] Finally, the set of policy constraints is compared with the meeting history information ("adopted denominator caliber" and its "comparison boundary") in the caliber state cache, and the output is a set of policy constraints that includes the lock priority denominator caliber and the comparison boundary.

[0097] Among them, hard constraints represent the set of rules that must be followed under the context state vector, including business stage restrictions, organizational scope restrictions, channel restrictions, consistency requirements with the caliber state cache, and boundary descriptions that prohibit cross-denominator caliber comparisons;

[0098] Preference ranking characterizes the ranking configuration of the candidate denominator, including speaker role preference, meeting goal preference, and historical adoption frequency reference, which is used to rank candidates without violating hard constraints;

[0099] The high-level policy of hierarchical reinforcement learning is a policy execution unit that maps the context state vector to a set of policy constraints. The technical approach can be understood as follows: First, the context state vector is used as the state input; then, the action sequence is selected as hard constraints, preference ranking is set, a clarification threshold is determined, and alignment with the caliber state cache is established; finally, the output is a set of policy constraints that locks the priority denominator caliber and comparison boundary for subsequent consistency adjudication.

[0100] It should be noted that the one-time clarification threshold is a threshold condition for triggering a single clarification. It can be understood as follows: First, the denominator of the top two ranked candidates and the basis output is read; then, the adaptation difference between the target field and the stage field of the context state vector is calibrated according to the target field and the stage field; finally, the threshold condition for triggering only one clarification is formed and written into the policy constraint set so as to identify the trigger in the subsequent consistency adjudication.

[0101] The process of locking the priority denominator caliber and comparison boundary as a determination operation after the generation of the strategy constraint set can be understood as follows: First, the adopted denominator caliber in the caliber status cache is used as the anchor; then, the denominator caliber that is consistent with the anchor or allowed by hard constraints is selected as the priority denominator caliber from the candidate and basis output; finally, the range that prohibits cross-denominator caliber comparison is written as the comparison boundary and recorded in the strategy constraint set for use in subsequent finalization of indicator description format and injection of comparison restriction rules.

[0102] It should be noted that the difference between the top two candidates in the ranking of each candidate denominator refers to the two candidate denominators that rank first and second among all candidate denominators after the "preference ranking" in step S3. For example, suppose the meeting discusses "project success rate," and the system retrieves three candidates:

[0103] Candidate A (ranked 1st): Success rate = Number of successful projects / Total number of approved projects (Reason: Consistent with the standards adopted at the meeting and in line with the current "overall planning" stage);

[0104] Candidate B (ranked 2nd): Success rate = Number of successful projects / Total number of projects that passed the initial review (Reason: It aligns with the role preferences of the questioner (review committee)).

[0105] Candidate C (ranked 3rd): Success rate = Number of successful projects / Total number of applications submitted;

[0106] So, the "top two" here refers to candidate A and candidate B.

[0107] In practical applications, the system in this application embodiment automatically generates a set of policy constraints based on the current "Level 1 Response" stage (hard constraint: requiring data to be traceable in real time within the command system) and the commander's role (preference ranking: favoring the most authoritative and immediate data source). It determines that candidate B (the population for whom transfer orders have been issued) best fits the hard constraints and role preferences, and sets it as the preferred denominator. Simultaneously, a clarification threshold is set to prevent the confidence levels of candidates A and B from being very close. S4: Based on the constraint policy set, a consistency decision is made on the candidate denominators to determine the unique denominator.

[0108] In one embodiment, this step, based on the policy constraint set generated in S3, performs a consistency ruling on the candidate denominator caliber generated in S2. First, the domain agent, consistency agent, and compliance agent of the multi-agent system review the candidates respectively. Then, the arbitration agent summarizes the review results of each agent and makes a final ruling under the principle of hard constraints taking precedence.

[0109] During the adjudication process, the system will determine whether the evaluation difference between the top two candidates is lower than the clarification threshold set by S3.

[0110] If the answer is below the threshold, the system will only send the user a clarification question once and will assist the arbitration agent in making a ruling based on the user's answer.

[0111] If the threshold is not lower than that (or no clarification is required), the arbitration agent shall directly determine the unique denominator based on the review results of each agent, the set of policy constraints, and the candidate denominator caliber.

[0112] Ultimately, the arbitration agent outputs the award, which is a unique denominator, and simultaneously generates alternative presentation options that can be used in case of conflicting comparisons.

[0113] It should be noted that, in this application, the terms 'domain agent,' 'consistency agent,' 'compliance agent,' and 'arbitration agent' all refer to software intelligent agents or functional processing units running on computing devices and configured to perform specific tasks.

[0114] The 'Consistency Agent' is configured to verify the consistency between the denominator and the historical status within the meeting; the 'Compliance Agent' is configured to verify the compliance of the denominator; and the 'Arbitration Agent' is configured to execute the final decision based on the output of the above agents.

[0115] The domain agent is configured to verify the business applicability of the denominator caliber, and can be a processing unit that verifies the applicable scenarios and effective period based on the caliber knowledge base; the consistency agent is configured to verify the consistency between the denominator caliber and the historical state within the meeting, and can be a processing unit that checks the consistency and comparison boundaries within the meeting based on the caliber state cache; the compliance agent is a processing unit that checks whether the candidate denominator caliber is suitable for the caliber knowledge base; the arbitration agent is configured to execute the final decision based on the output of the above agents, and can be a processing unit that selects a unique denominator caliber under the principle of hard constraint priority; hard constraint priority is the decision principle of executing hard constraints first and then executing preference ranking in the policy constraint set; one-time clarification fallback is a control rule that only issues one clarification question when the one-time clarification threshold is triggered and uses the answer for convergence arbitration; the one-time clarification question is a directional question about the denominator difference.

[0116] In this embodiment, the unique denominator is a single denominator defined under the constraints of the candidate and output and policy constraint sets. The unique denominator in this embodiment is obtained in the following way:

[0117] The unique denominator is obtained in the following way:

[0118] Sub-step S401: The domain agent filters the candidate denominator based on the hard constraints in the policy constraint set, considering the applicable scenarios and effective periods.

[0119] Sub-step S402: The consistency agent uses the caliber state cache as a reference to filter out candidate denominator calibers that conflict with the comparison boundary of the adopted denominator caliber, and marks the candidate denominator calibers that can be displayed in parallel.

[0120] Sub-step S403: The compliance agent removes candidate denominator calibers that do not conform to the caliber knowledge base adaptation rules;

[0121] Sub-step S404: Based on the above results, the arbitration agent selects a unique denominator based on the principle of hard constraint priority and the aforementioned preference ranking;

[0122] Sub-step S405: When the clarification threshold is triggered, the arbitration agent raises a clarification question and updates the sorting and selection after receiving the answer, outputting a unique denominator for subsequent finalization of the indicator description format.

[0123] Furthermore, in some embodiments, the alternative display scheme for the step submission can be a screen display path provided when comparison restriction rules may prevent the merged display; "alternative display scheme" refers to one or more compliant and executable alternative data display schemes provided to the user after the system intercepts the (incorrect) parallel display operation that the user originally wanted to perform according to comparison restriction rules.

[0124] For example, an "alternative display scheme" can be implemented as follows: First, when the unique denominator caliber and the caliber state cache have inconsistent denominator calibers or there are comparison boundary conflicts, the generation of an alternative display scheme is initiated; then, the display strategy is determined according to the set of policy constraints, providing two paths: split-screen display or unified common caliber display; the former displays the indicators corresponding to different denominator calibers separately in the screen display layer and retains the comparison boundary annotations, while the latter replaces the denominator field with the common caliber in the indicator description format and generates an explanatory identifier; finally, the alternative display scheme and the unique denominator caliber are output together for subsequent injection of comparison restriction rules and execution of screen display.

[0125] In practical applications, the various organizations of the multi-agent system in this application embodiment perform the following functions:

[0126] The domain agent is used to verify whether the source (command system) of candidate B is the best practice in the current "Level 1 Response" phase.

[0127] The consistency proxy is used to check whether candidate B conflicts with the "enabled placement point capacity" in the cache of step S1 (this embodiment uses the example of no conflict for explanation).

[0128] The compliance agent is used to confirm that there are no data security and compliance risks associated with this caliber.

[0129] The arbitration agent is used to synthesize the opinions of various agents and, based on the strategy in step S3, finally determine a unique denominator: Transfer completion rate = Transferred population / Population for which transfer instructions have been issued. S5: Based on the unique denominator, a finalized indicator description format is generated, and a restriction rule is embedded in the finalized indicator description format to prevent visual comparison of inconsistent indicators;

[0130] In this embodiment of the application, based on the unique denominator caliber and alternative display scheme obtained in step S4, the denominator caliber identifier and caliber description can be written into the indicator description format draft to generate the final version of the indicator description format, and the denominator caliber identifier and caliber description can be locked, while a comparison restriction rule is generated and injected.

[0131] In practical applications, this application embodiment solidifies the adjudication result into a finalized indicator description format, explicitly including the denominator definition "population for which transfer instructions have been issued," and locks it to prevent arbitrary modification during meetings. Simultaneously, a comparison restriction rule is injected: "It is prohibited to compare this 'transfer completion rate' with completion rates calculated using other denominators (such as 'total population to be transferred') in the same chart." S6: Based on the finalized indicator description format and its embedded restriction rule, the data visualization interface is rendered, and parallel display operations that violate the restriction rule are intercepted and guided in real time.

[0132] In this embodiment of the application, the screen display can be based on the finalized index description format of step S5, and parallel operations across denominator scopes can be intercepted according to the comparison restriction rules, and an alternative display scheme can be adopted. At the same time, the unique denominator scope and the interception event are written back to the scope status cache and archived.

[0133] In practical application, the large screen correctly displayed the "transfer completion rate" charts for each district and county based on a unified standard. At this point, if a user attempted to drag and drop completion rate data (based on candidate C denominator) reported by the civil affairs system into the same chart for comparison, the system would immediately block the operation and display a prompt, such as: "A conflict in denominator standards has been detected; comparison within the same chart is prohibited. We suggest you [split-screen display] or [use the unified standard from the command system]." Simultaneously, the final standard of this ruling and this blocking event were written back to the audit log for post-meeting review and tracing.

[0134] The technical effect of this application's embodiments is that existing systems cannot automatically determine a unified standard for calculating indicators in real-time interactive scenarios and force the execution of this standard in the front-end visualization stage, leading to incorrect data comparisons. To address the above-mentioned technical deficiencies of existing systems, the technical solution of this application's embodiments achieves an automated closed loop between the "decision-making" (adjudication) and "execution" (interception) of the denominator caliber within the system, fundamentally eliminating data visualization errors caused by inconsistent standards.

[0135] In some embodiments, for the aforementioned step S1, the context state vector is a structured output used to bind the context of the adjudication in the meeting. This can be understood as follows: First, extract the indicator intent and qualifying words from the speech text and generate the text key point field; then, perform priority mapping on the text key point field according to the agenda to determine the current target field; next, write the speaker's role and business stage into the role field and stage field; at the same time, confirm the organizational scope and time window as the query boundary field; finally, write the adopted denominator caliber into the caliber reference field and form a context state vector output for reference and constraint alignment in subsequent steps.

[0136] The initialization of the caliber status cache can be used to establish a cache for the current meeting and write the first record. It can be understood as follows: when the meeting starts, a cache record area is created and the adopted denominator caliber and its comparison boundary are written; after each screen display is completed, the current denominator caliber record is appended and the association index with the context state vector is maintained for subsequent steps to perform consistency verification and adjudication reference.

[0137] The denominator caliber has been adopted as the denominator definition used by the indicators currently displayed on the screen. It is obtained by reading from the displayed indicator description format or the previous round of adjudication records and writing it into the caliber reference field of the context state vector as the basis for the default denominator caliber in subsequent adjudications.

[0138] Furthermore, in some embodiments, for the aforementioned step S2, the indicator intent parsing is to map the spoken request to a set of indicator intent fields, including indicator name, numerator definition trigger word, denominator limit word, time window limit word and filtering condition statement; the retrieval enhancement generation model is a model that combines retrieval and generation, using the context state vector as the retrieval condition, calling the caliber knowledge base to return evidence, and generating a draft indicator description format and candidate and basis output based on the evidence;

[0139] The draft indicator description format is an initial version of the indicator description format, which may include indicator name, numerator definition, denominator placeholder, filtering conditions and time window field, for subsequent decision-making to fill in the denominator scope;

[0140] The draft of the indicator description format can be the result of the synthesis based on the retrieval enhancement generative model. In the specific implementation: first, the indicator name and numerator definition are generated on the set of indicator intent fields obtained by indicator intent parsing; then, the denominator field is marked with a denominator caliber placeholder and candidate reference identifier is attached; finally, the complete draft is formed by combining the screening conditions and the time window, and a reference relationship is established with the context state vector for use in subsequent policy constraint sets and consistency decisions.

[0141] The process of converging candidate denominator criteria into a set of candidate denominator criteria under the constraints of the context state vector involves the following steps: First, based on the target and stage fields in the context state vector, denominator criteria entries that match the current target and business stage and are in effect are selected from the denominator knowledge base. Then, the organizational scope and time window are limited according to the query boundary fields in the context state vector, excluding entries outside the scope. Subsequently, the candidates are sorted by preference according to the role field in the context state vector and compared with the denominator reference field in the context state vector, placing the adopted denominator criteria at the anchor position of the candidate set. Finally, synonymous denominator criteria are merged by aliasing and an adaptation reason for each candidate is generated, forming a candidate set that can be referenced for subsequent adjudication.

[0142] The draft index description format is a synthesis result based on the retrieval enhancement generative model. In its specific implementation: first, index names and numerator definitions are generated on the index intent field set obtained from index intent parsing; then, denominator fields are marked with denominator caliber placeholders and candidate reference identifiers are attached; finally, a complete draft is formed by combining the screening conditions and time window, and a reference relationship is established with the context state vector for use in subsequent policy constraint sets and consistency decisions.

[0143] The candidate denominator caliber and the basis for the candidate denominator caliber can be understood as a candidate-evidence pairing set. In specific implementation: first, for each candidate denominator caliber, the definition, applicable scenarios, effective period and source basis are extracted from the caliber knowledge base; then, an adaptation reason explanation is generated on the target field and stage field of the context state vector; finally, the above evidence and candidates are bound into a structured output for direct reference by the subsequent policy constraint set and consistency adjudication.

[0144] Furthermore, in some embodiments, for the aforementioned step S5, the finalized indicator description format is the final version of the indicator description format formed after the unique denominator caliber and alternative display scheme are determined in S4, including indicator name, numerator definition, denominator field, denominator caliber identifier, caliber description, filtering conditions, time window, comparison restriction rule reference, version number, and lock status.

[0145] The denominator caliber identifier is a structured identifier that uniquely locates the denominator caliber in the caliber knowledge base, including the caliber knowledge base identifier, denominator name, and version number.

[0146] The definition description is a description of the definition of the unique denominator, including the definition text, applicable scenarios, organizational scope, time window, source basis, and effective period.

[0147] Locking sets a write protection flag and version binding for the denominator field and its definition in the finalized indicator description format, restricting modifications within this meeting.

[0148] The comparison restriction rules are a set of rules that intercept screen display and query generation under comparison boundary constraints. They include rule identifiers, scope of application (chart or table column), boundary references, interception actions, and guidelines for alternative display schemes.

[0149] The comparison constraint rule injection is the operation of writing the comparison constraint rules into the finalized indicator description format and binding them to the screen display parameters.

[0150] The process of writing the denominator caliber identifier and caliber description into the indicator description format and locking it is based on the S4 unique denominator caliber and alternative display scheme. In specific implementation: First, the record corresponding to the unique denominator caliber is read from the caliber knowledge base to obtain the caliber knowledge base identifier, denominator name and version number, which are written into the denominator field of the indicator description format draft to form the denominator caliber identifier; then, the definition text, applicable scenario, organizational scope, time window, source basis and effective period are extracted from the record and written into the caliber description field of the indicator description format draft; next, a write protection flag is set and version binding is performed, and the denominator field and its caliber description are locked, limiting the lock state to the validity of this meeting; finally, the finalized indicator description format is generated for subsequent comparison restriction rule injection and screen display.

[0151] The comparison constraint rule injection is the process of generating and binding rules under the constraints of comparison boundaries and alternative display schemes. In specific implementation: First, based on the comparison boundaries and the constraint of "prohibiting comparisons across denominators in the same chart or table column", rule parameters are generated and the applicable scope (chart or table column) is determined. Then, the source basis, effective period, and rule parameters are combined to form a rule object, which includes rule identifier, boundary reference, interception action, and alternative display scheme guidance. Subsequently, the rule object is written into the rule field of the finalized indicator description format, and an interception point is registered in the screen display parameters for scenarios of chart merging and parallel generation of table columns. At the same time, constraints are registered during the query generation stage. When indicators from different denominators are fitted within the same chart or table column, an interception is triggered and an alternative display scheme is returned. Finally, the finalized indicator description format containing the injected rules is output for screen display.

[0152] The prohibition of comparing different denominators within the same chart or table column is the implementation of the rule semantics at the execution layer. In specific implementation: during the chart layout stage, the denominator identifier in the finalized description format of the indicators involved in the comparison is checked item by item. When different denominator calibers are detected within the same chart or table column, an interception action is immediately triggered to prevent the layout from being merged and to call an alternative display scheme. At the same time, the caliber description is displayed on the screen and the comparison boundary label is retained to maintain consistency within the meeting and guide the use of the alternative display scheme to complete the display.

[0153] The finalized indicator description format combines a unique denominator caliber, caliber description, and comparison restriction rules. In its implementation: the denominator caliber identifier and caliber description are written and locked based on the draft indicator description format; comparison restriction rules are injected and rule binding is completed; the version number and lock status are recorded; and a structured output that can be directly called for screen display is generated.

[0154] Regarding the unified indicator processing method for real-time interaction described in the first aspect, a specific implementation method in one embodiment may be:

[0155] In step S1, the system parses the internal information into contextual state vectors in a structured manner and initializes the caliber state cache. To facilitate subsequent citation and constraint alignment, the elements entering this step are first symbolically represented: the spoken text is denoted as a vector sequence. The agenda is a structured object. The spokesperson's role is as an identifier. The business stage is labeled. The scope of the organization is a collection The time window is an interval The denominator has been adopted as the unique identifier. During the parsing process, the system generates a set of fields for adjudication, including text key points, target fields, stage fields, role fields, query boundary fields, and caliber reference fields, represented as follows: , , , , and The aforementioned fields, as components of the contextual state vector, are used to bind the adjudication context during the meeting and support subsequent candidate retrieval and policy constraints.

[0156] Specifically, the system first bases its decisions on the spoken text. Extract the intent and qualifiers of the indicators to form the key text fields. ; then according to the agenda right Perform priority mapping to determine the current target field. And the spokesperson role Write role field Business phase Write stage field The system will also include the organizational scope. With time window Combined fields are confirmed as query boundary fields. This limits the effective boundaries of candidate retrieval and decision-making. For those already using a denominator approach... The system writes it into the target reference field. This serves as the default denominator for calculations within the meeting, ensuring consistency and uniformity of the denominator across the same meeting period. The above fields are assembled into a contextual state vector:

[0157]

[0158] This contextual state vector is directly referenced as a structured output in subsequent steps.

[0159] During the initialization of the caliber status cache, the system creates the caliber status cache at the meeting venue level. This is then written into the first record to solidify the denominator caliber anchor position and comparison boundary within the meeting. The comparison boundary is denoted as set. (Used to mark the scope and rule boundaries where cross-denominator comparisons are prohibited), the initialization record can be represented as: The record is stored in and the context state vector Create a related index (denoted as) This data is used for subsequent candidate screening, strategy constraint set generation, and consistency adjudication for intra-meeting consistency verification and boundary comparison. Through the above parsing, assembly, and cache initialization, step S1 forms a reusable contextual state vector output and an intra-meeting caliber state baseline, supporting subsequent steps in referencing the denominator caliber, adhering to comparison boundaries, and selecting display paths.

[0160] In step S2, the system relies on the context state vector Conduct analysis of the intent of indicators and gather candidate denominator definitions, and generate a draft of indicator description format and output candidates and supporting evidence.

[0161] First, the system in the text key field The above performs indicator intent parsing to obtain the indicator intent field set. This includes the indicator name, numerator definition trigger word, denominator qualifier, time window qualifier, and filter condition statement. This set and the target field... Stage Field Role fields The system jointly constructs generation conditions to drive subsequent retrieval and generation. To support the generated results, the system creates a draft index description format. Its fields include numerator definitions and denominator field placeholders. Filtering criteria Time window The candidate denominator is referenced by an identifier; where the time window and filtering conditions are respectively influenced by the query boundary fields. Interval and range constraints in (i.e.) and The denominator field is reserved for subsequent candidate injection.

[0162] Furthermore, the system calls the large model, namely the retrieval-enhanced generative model. , in context state vector Using the caliber knowledge base as search criteria The system retrieves entries for the denominator scope, their applicable scenarios, and effective dates, and generates fillable fragments, candidates, and supporting evidence for a draft indicator description format based on the retrieved evidence. To standardize candidate representation, the system uses a set of candidate denominator scopes. The denominator is defined as the retrieved data, in terms of the evidence set. Each candidate is represented by evidence fields such as the source of evidence, applicable scenario, and effective date, and is presented as a candidate-evidence pairing set. As candidates and criteria for output. During the generation process, the system will... The denominator qualifiers and time window qualifiers in the search query are mapped to search constraints, which are then overlaid with the target field. With stage field Contextual constraints are used to avoid retrieving denominators that do not match the current business stage; the role field is also referenced. Candidates are labeled with display preferences for subsequent sorting and adjudication. (Generated) Placeholders in the denominator field Connected to The reference identifiers form a structured output that allows for linkage between "drafts" and "candidates".

[0163] In the candidate denominator aggregation stage, the system relies on the query boundary fields. The defined organizational scope With time window ,right Perform range filtering and reference fields by scope. The denominator diameter has been adopted in the calculation. As an anchor point, candidate merging is performed with priority locking. Specifically, the system first eliminates candidates that are not present in the list. or exceed The candidates are: first, merge synonymous denominators with aliases, retaining unique identifiers and integrating relevant evidence; second, consolidate the candidates. again in the role of spokesperson Candidates are labeled according to their preferences for meeting objectives, forming a ranking mapping. Finally, it will be with Candidates that are consistent or fall within the permissible range of the contextual state vector are placed at the forefront of the set so that subsequent policy constraint sets can reference them. Based on this convergence result, the system outputs a draft of the hooked index description format. And candidate and basis output And maintain placeholders for the denominator field in the draft. Binding with candidate reference identifiers to support step S3 in locking the priority denominator caliber under hard constraints and preference ordering.

[0164] Based on this, step S2 completes the closed loop of indicator intent parsing, retrieval enhancement generation, and candidate denominator aggregation, and outputs... This allows for direct invocation of subsequent policy constraint sets and consistency decisions.

[0165] In step S3, the system is based on the context state vector And the candidate and basis output of step S2 With the set of candidate denominators The policy constraint set is generated by the high-level policy of hierarchical reinforcement learning and is cached with the caliber state. Comparison Boundary Set Alignment. The set of policy constraints is denoted as... It includes hard constraints, preference ranking, first clarification threshold, priority denominator caliber, and comparison boundary, which are used to limit the selection space of candidates and the conditions for triggering clarification within the meeting.

[0166] In practice, the system first constructs hard constraints. Based on the target field With stage field Define the business phase and target scope; based on query boundary fields. Introducing organizational scope With time window Boundary conditions; and cached with caliber state. Provided comparison boundaries The comparison constraints between candidates are verified. Then, a preference ranking is generated without violating the hard constraints. : By role field Role preferences and target fields of meeting target preferences Perform label sorting and reference fields by scope. The denominator has been adopted As the priority of anchor position enhancement candidates, the top two candidates are formed after sorting. and .

[0167] Regarding the threshold for clarification The system calculates a threshold based on the principle of prioritizing hard constraints. This threshold is adaptively set to the context state vector and is used to determine whether to only ask a clarification question when the difference between the top two ranked items is less than this threshold. The specific calculation formula is as follows:

[0168]

[0169]

[0170] in: This is to clarify the threshold; For hard constraints, feasibility indicators (feasibility take) Not feasible ); Rank by preferences The top two candidates; These are weight parameters; Quantify the speaker's role preferences; Quantify preferences for different business stages; Anchor consistency indicator function; Alignment measure for organizational scope or time window; Candidates The applicable organizational scope and effective time window; To query the organizational range and time window in the boundary field; To compare boundary conflict measures (when a conflict occurs in the comparison of denominators, the nearest value is taken). ).

[0171] The system controls clarification triggers based on the aforementioned threshold: when based on the evidence set The calculated gap between the first two fits is lower than At that time, only one clarification question is issued and the answer is used to update the sorting map. Otherwise, subsequent locking is determined directly based on hard constraints and preference ordering. In the alignment with the caliber state cache, the system uses... As an internal anchor point, it should be selected in accordance with hard constraints and without triggering comparison boundaries. Candidates as priority denominator and write it into the policy constraint set. Final output This allows step S4 to be directly referenced in multi-agent consensus adjudication and first-time clarification fallback, while maintaining consistency with the draft indicator description format. Denominator field placeholders The binding is used to maintain intra-session consistency when subsequently locking the priority denominator caliber and comparison boundaries.

[0172] In step S4, the system uses a context state vector Candidates and output based on criteria Candidate denominator set Policy constraint set and caliber status cache Comparison Boundary Set As input, a multi-agent consensus decision and a fallback clarification are initiated. The decision-making process includes domain delegate, consensus delegate, compliance delegate, and arbitration delegate, adhering to the principle of prioritizing hard constraints. The goal of the internal decision-making process is to determine a unique denominator without triggering comparison boundaries, while simultaneously preparing alternative presentation schemes to support the finalization of the indicator description format and the injection of comparison restriction rules.

[0173] Specifically, the domain agent considers each candidate based on its own candidate data and the source data and effective period in the output. Verification of applicable scenarios and effective period was conducted to obtain domain review quantification. Consistent proxy uses caliber state caching. Compare boundary sets For comparison, the denominator of the candidates and those already used in the meeting was checked. The consistency and compliance of the comparison boundary are quantified by the consistency review. Compliance agents examine candidates' compliance from both legal and internal policy perspectives, resulting in quantifiable compliance assessments. Arbitration representation under hard constraints The above reviews are summarized under constraints and ranked according to preferences. One-time clarification threshold Compare boundary sets The arbitration process prioritizes hard constraints and provides a fallback option of clarification.

[0174] Arbitration agents use the following single arbitration selection formula to determine the unique denominator:

[0175]

[0176]

[0177] in: It is the only denominator. This serves as a hard constraint feasibility indicator. , , The evaluation criteria for domain proxy, consistency proxy, and compliance proxy are quantified respectively. Sorting by preference Derived sorting weights; To match the denominator diameter already adopted Anchor position consistency indication; Alignment metrics for organizational scope or time window, where , Candidates The applicable organizational scope and effective time window, , From the query boundary field; The reliability of evidence is calculated based on the source and the effective date. A cross-denominator conflict measure related to the comparison boundary set; , These are weight parameters; For indicator functions; The difference between the first two arbitrations before clarification was triggered (by...) (The difference between the maximum and second largest values ​​is obtained). This is to clarify the threshold; To clarify the answer to a question For candidates The amount of a single correction.

[0178] In a clarification and safeguarding process, if The arbitration representatives only raised one clarifying question, focusing on the difference in the denominators of the top two candidates; each representative then responded based on their answers. Update the components related to ambiguity in the mapping or ranking weights of each review quantification, in the formula. This term is used to transform the single response into an additivity correction, thereby completing convergence arbitration without changing the priority of hard constraints. If This amendment is ineffective, and arbitration will be conducted directly based on hard constraints and multi-agency review to make the selection.

[0179] when Once determined, the consistency proxy... and Comparing boundary sets The final comparison will be performed above: If If the indication may trigger a conflict between different denominator standards, the arbitration agent will generate an alternative display scheme according to the agreement in the policy constraint set, including a split-screen display path and a common standard display path. The former is used to separate the indicators corresponding to different denominator standards in the screen display layer and retain the comparison boundary label, while the latter is used to replace the denominator field with the common standard in the indicator description format and generate an explanatory label.

[0180] Then the system will Output along with alternative presentation schemes, while maintaining the same format as the draft indicator description. Denominator field placeholders The binding relationship is used in step S5 to finalize the indicator description format and inject comparison restriction rules. Through the above multi-agent consensus adjudication and first clarification fallback, step S4 achieves the unique determination of the denominator under the priority of hard constraints, and provides an executable display path for possible comparison boundary conflicts.

[0181] In step S5, the system uses a unique denominator. Alternative display solutions As the core input, combined with the contextual state vector Draft of indicator description format Compare boundary sets With candidates and based on output Complete the finalization of the indicator description format and the injection of comparison restriction rules.

[0182] First, read from the caliber knowledge base... The corresponding denominator caliber identifier (including caliber knowledge base identifier, denominator name and version number) and caliber description (including definition text, applicable scenarios, organizational scope, time window, source basis and effective period) should be written into the caliber. The denominator field and the scope description field are used to generate the final draft of the indicator description format. Write protection locks are set for the denominator field and its scope description, and the lock status is recorded. The version number is used to ensure consistency with internal references. Meanwhile, the system... Record comparison boundary references Guidelines for alternative display solutions, used for execution binding of the screen display layer.

[0183] In the comparison restriction rule injection stage, a set of rule objects is constructed based on the scope of application (same chart or same table column). and for each display unit (Representing a chart or table column) The calculation rule triggers a score to determine whether to prohibit comparisons across denominators within the unit and to inject interception and guidance. To ensure that rule generation is consistent with the meeting and strictly aligned with comparison boundaries, a comprehensive measure is taken around denominator differences, comparison boundary conflicts, inconsistencies in organizational scope and time windows, and evidence uncertainty, resulting in a single trigger formula as follows:

[0184]

[0185]

[0186] in: Indicates the display unit (same chart or same table column) within the applicable scope; In the locked state (locked) ); Weights for applicable scope (distinguishing between charts and table columns); for A set of pairwise combinations of internal indicator items; This is the denominator of the two terms in the combination; To compare boundary conflict measures; for An internally designed set of indicators displayed in parallel; Alignment measure for organizational scope or time window; For indicator items The applicable organizational scope and effective time window; From query boundary fields ; for The set of denominator definitions involved; The reliability of evidence regarding the source and effective date; These are the weight parameters.

[0187] System set trigger threshold ,when At that time, comparison restriction rules are generated and injected, including rule identifier and scope of application. Boundary references The guidelines for interception actions and alternative display solutions will include the source basis and effective date, which will be written into the rules. The rule field; simultaneously register intercept points in the screen display parameters so that the same chart or the same table column is intercepted when indicators with different denominators are to be displayed side by side, and according to Choose between split-screen display path or common-view display path.

[0188] When the rule is triggered, the interception action takes effect at the screen display layer: if there are requirements for cross-denominator caliber but not related to the comparison boundary set. If a conflict occurs, the system will refuse to merge the layouts and will execute either a split-screen display path (displaying indicators corresponding to different denominator definitions separately while retaining comparison boundary labels) or a common-definition display path (replacing the denominator field with the common definition in the indicator description format and generating a description identifier) ​​based on the alternative display scheme. After the injection is complete, As a structured output, it includes denominator caliber identifier, caliber description, lock status, rule object and version number, which can be directly called by the screen display layer and referenced in the generation stage. This ensures that cross-denominator comparisons are prohibited in the same chart or table column within the meeting, and supports the traceability and effectiveness control of the rules with source basis and effective period.

[0189] In step S6, the front end is finalized in the indicator description format. For rendering purposes, the denominator caliber identifier, caliber description, lock status, and rule object set are read. Boundary references and alternative display solutions and with a single denominator Populate the denominator field, and output the indicator name, denominator field, and scope description in the screen display layer according to the field order. Simultaneously, display comparison boundary annotations at the component level, and mark the lock state to indicate that the denominator field and scope description cannot be modified. Maintain the context state vector during rendering. Consistent query boundaries ensure organizational scope With time window The instructions and interface prompts are aligned.

[0190] During the interaction interception phase, user actions on the same chart or the same column of a table are identified as display units. The merge attempt, based on the front end The comparison constraint rules already injected in the collection of rule objects Search The rules identify and read the boundary references. Scope of application. When a rule indicates that comparisons across denominators are prohibited within a display unit, the front end immediately intercepts the comparison, prevents layout merging, and provides an alternative display solution suggestion: [Following the provided text]. The system guides users to choose between a split-screen display path or a common-standard display path. The former displays the indicators corresponding to different denominator standards separately on the screen and retains comparison boundary labels, while the latter applies the common standard to the denominator field before displaying the data and includes a standard description label to indicate the basis for the replacement and the effective period. If the rule is not triggered, the parallel display will proceed normally, but the standard description and comparison boundary label will still be displayed in the component sidebar to indicate the consistency boundary within the meeting.

[0191] During the auditable write-back phase, the frontend generates an interception event record when the operation is completed or an interception occurs, and writes it into the denominator caliber cache along with the record. The inner anchor position will be updated to Record and display unit Rule identifiers and boundary references The triggering reason and an alternative display solution summary are then archived into the audit log collection (including timestamp, operator ID, and version number) to support post-meeting traceability and reuse in subsequent meetings. After the above write-back is completed, the front-end retains and... The binding relationship is used to directly read the lock state and rule object when the session is reopened, continuing the adopted denominator caliber and comparison boundary, ensuring continuous interception of cross-denominator caliber operations during screen display and interaction, and providing executable alternative display solutions.

[0192] Secondly, such as Figure 3 As shown in the figure, this application provides a computer device, which may be a desktop computer, a laptop computer, a smartphone, etc.

[0193] The computer device may include a processor 101, a memory 102, and a computer program stored in the memory 102 and executable on the at least one processor 101;

[0194] The computer device can be connected to a display screen; of course, the computer device itself can also have a display screen.

[0195] The display screen can show the speaker's verbal request and visualize the relevant descriptive information of the unique denominator caliber finally determined by the unified processing method for real-time interactive indicators according to the embodiments of this application.

[0196] Thirdly, embodiments of this application also provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0197] This application also provides a computer program product that, when run on the aforementioned computer device, enables the processor of the aforementioned computer device to execute the steps described in the first aspect of the method embodiment.

[0198] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0199] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for unifying indicator definitions for real-time interaction, the method being executed by a computing device, characterized in that... Includes the following steps: S1: Receive and parse the interaction context information, obtain the verbal request, generate a structured context state vector, and write the context state vector as meeting history information into the aperture state cache; S2: Based on the context state vector, obtain multiple candidate denominator calibers that match the oral request from the caliber knowledge base, as well as the basis for the candidate denominator calibers; S3: Based on the context state vector, the candidate denominator caliber, and the basis for each candidate denominator caliber, a constraint strategy set is generated. The constraint strategy set is used to align with the meeting history information in the caliber state cache to determine a unique denominator caliber from the multiple candidate denominator calibers. S4: Based on the set of constraint strategies, the unique denominator is determined by making a consistency decision on the candidate denominator calibers. S5: Based on the unique denominator, generate a finalized indicator description format and embed a restriction rule for prohibiting the visual comparison of inconsistent indicators into the finalized indicator description format. S6: Based on the finalized indicator description format and the embedded restriction rules, render the data in the data visualization interface, and intercept and guide parallel display operations that violate the restriction rules in real time. Step S2 further includes: based on the context state vector, using a retrieval enhancement generation model to parse the oral request into a draft index description format, and retrieving candidate denominator calibers and their basis from a caliber knowledge base to form candidate denominator calibers and the basis for the candidate denominator calibers; Step S3 further includes: generating a set of policy constraints based on the context state vector, the multiple candidate denominator calibers, and the basis for each candidate denominator caliber. The set of policy constraints includes hard constraints, preference ranking, and a first clarification threshold. The set of policy constraints is aligned with the caliber state cache to lock a preferred denominator caliber and comparison boundary. The clarification threshold is a threshold condition that triggers a single clarification.

2. The method of claim 1, wherein, The interaction context information includes the speech text, agenda, speaker role, business stage, organizational scope, and time window.

3. The method of claim 1, wherein, Step S4 further includes: Based on the set of policy constraints, a consensus decision is made on each of the candidate denominator categories, including: the domain agent, consensus agent, and compliance agent of the multi-agent system reviewing each of the candidate denominator categories respectively; and the arbitration agent summarizing the review results of each agent and making a decision with the hard constraints as the priority principle. In adjudicating the various review results, it is determined whether the difference between the review results of the top two candidate denominators is lower than the aforementioned clarification threshold: If the result is below the threshold for clarification, a clarification question will be sent to the user, and the arbitration agent will be assisted in making a ruling based on the user's answer. If the threshold is not lower than the threshold, the arbitration agent determines the unique denominator based on the review results of each agent, the set of policy constraints, and the candidate denominator caliber, and simultaneously generates an alternative display scheme that can be used when a comparison conflict occurs. Step S5 further includes: based on the unique denominator caliber and the alternative display scheme, writing the denominator caliber identifier and caliber description into the indicator description format draft to generate the final indicator description format, locking the denominator caliber identifier and caliber description, and simultaneously generating and injecting a comparison restriction rule.

4. The method of claim 3, wherein, Step S6 further includes: displaying the data on the screen based on the finalized index description format, intercepting parallel operations across denominator calibers according to the comparison restriction rules, guiding the adoption of the alternative display scheme, and writing the unique denominator caliber and the interception event back to the caliber status cache and archiving it.

5. The method of claim 1, wherein, In step S3, a set of policy constraints is generated based on the contextual state vector, the multiple candidate denominator calibers, and the basis for the candidate denominator calibers, including: Read the target field, stage field, role field, query boundary field, and caliber reference field from the context state vector; Generate hard constraints that match the target field and the stage field, and write the query boundary field into the range of the hard constraints; wherein, the hard constraints represent the set of rules that need to be followed under the context state vector; A preference ranking is formed based on the role field and the target field; wherein, the preference ranking represents the sorting configuration of the candidate denominator caliber; A clarification threshold is defined by combining the multiple candidate denominator calibers and the difference between the top two calibers in the ranking of each candidate denominator caliber. A set of policy constraints is obtained, which includes the hard constraints, the preference ranking, and the first clarification threshold; The policy constraint set is compared with the meeting history information in the caliber state cache, so that the output of the policy constraint set includes the locking priority denominator caliber and its comparison boundary.

6. The method of claim 3, wherein, In step S4: The domain agent is a processing unit that verifies the applicable scenarios and effective period based on the aforementioned knowledge base; The consistency proxy shown is a processing unit that checks intra-session consistency and comparison boundaries based on the aforementioned caliber state cache. The compliance agent is a processing unit that checks whether the candidate denominator caliber is compatible with the caliber knowledge base; The arbitration agent is a processing unit that selects a unique denominator under the principle of prioritizing hard constraints; The hard constraints are configured to prioritize the decision principle of executing hard constraints first and then preference ranking in the policy constraint set.

7. The method of claim 4, wherein, The unique denominator caliber in step S4 represents a single denominator definition determined under the candidate denominator caliber, the basis of the candidate denominator caliber, and the hard constraints of the policy constraint set. The unique denominator is obtained in the following way: The domain agent filters the candidate denominator based on the hard constraints in the policy constraint set, considering the applicable scenarios and effective periods. The consistency agent uses the caliber state cache as a reference to filter out candidate denominator calibers that conflict with the comparison boundary of the adopted denominator caliber, and marks the candidate denominator calibers that can be displayed in parallel. The compliance agent removes candidate denominator calibers that do not conform to the adaptation rules of the caliber knowledge base; Based on the above results, the arbitration agent shall select a unique denominator based on the principle of prioritizing hard constraints and the aforementioned preference ranking. When the clarification threshold is triggered, the arbitration agent raises a clarification question, updates the sorting and completes the selection after receiving the answer, and outputs a unique denominator for use in the finalization of the subsequent indicator description format.

8. A computer device, comprising: It includes a processor and a memory, the memory storing a computer program that, when executed by the computer device, performs the steps of the unified index processing method for real-time interaction as described in any one of claims 1 to 7.

9. A computer program product storing a computer program, characterized in that, When the computer program is executed by the computer device, it runs the unified processing method for real-time interactive indicators as described in any one of claims 1 to 7.

Images