A traceable text generation method, system, platform and storage medium based on structured evaluation data

Abstract

This application discloses a method, system, platform, and storage medium for generating traceable text based on structured evaluation data. The solution involves validating project metrics according to pre-designed sub-logic rules and calculating evaluation state feature information. Based on this feature information, reference knowledge information is retrieved from a knowledge base. The project metrics, evaluation state feature information, and reference knowledge information are constructed as input corpus, and a reasoning request carrying generation constraints is sent to a generative model. The generated target description text is received and traceability verification is performed. Factual information items are extracted from the target description text and compared with the original structured evaluation data for consistency. After successful verification, the target description text is combined with the evaluation graph to generate an evaluation report file containing traceability identifiers. This solution achieves high efficiency, high individualization, and high reliability in evaluation report generation.

Description

Technical Field

[0001] This application belongs to the fields of computer data processing, artificial intelligence-assisted evaluation and natural language generation technology, and specifically relates to a method, system, platform and storage medium for generating traceable text based on structured evaluation data. Background Technology

[0002] Currently, assessments of children's language behavior and learning abilities typically involve assessors entering scores item by item using paper quality forms, spreadsheets, or general information management systems. The assessors then manually complete the score summarization, assessment chart creation, pre- and post-test comparisons, interpretation of assessment results, and writing of intervention recommendations.

[0003] In existing technologies, different assessment areas include multiple levels, items, half-scores, missing items, and special scoring rules. Different assessment scales have different scoring logics; for example, the allowed score set for milestone assessment items is {0, 0.5, 1}, while the allowed score set for obstacle assessment items is {0, 1, 2, 3, 4}. Ordinary spreadsheets struggle to automatically identify incomplete data or data that does not conform to the scoring logic, resulting in high data entry error rates and frequent item omissions.

[0004] Furthermore, the same subject may undergo multiple assessments on different dates. Existing systems typically only save the results of each assessment separately, lacking a mechanism to present the data from multiple assessments in a unified coordinate system. This makes it difficult to accurately display score changes at different points in time, the evolution of test age, and ability development trends in assessment graphs. Moreover, the creation of charts and reports relies on manual labor, which is inefficient and prone to errors. Assessors need to manually calculate the total score for each area, draw milestone and obstacle maps, and copy each chart into the report. This manual operation is not only time-consuming but also prone to calculation errors, omissions, or inconsistencies between the charts and text.

[0005] Furthermore, existing automated report generation methods typically use pre-defined templates based on fixed score ranges, failing to provide truly individualized analysis and interpretation by considering the specific score distribution across different domains, unmastered items, prominent obstacles, and historical trends of the assessed individuals. Moreover, directly inputting assessment data into a general language model may produce "illusionary" content that is inconsistent with scale standards, lacks traceable factual sources, or exceeds the scope of the input data, making it difficult to meet the stringent requirements of clinical assessment for report reliability and traceability. Additionally, when assessment data is modified, existing charts or reports may not be updated synchronously, resulting in data version inconsistencies and affecting the accuracy and credibility of the assessment results.

[0006] Therefore, in order to address the aforementioned technical deficiencies, there is an urgent need to design and develop a traceable text generation method, system, platform, and storage medium based on structured assessment data. Summary of the Invention

[0007] To overcome the shortcomings and difficulties of the existing technologies, the purpose of this application is to provide a traceable text generation method, system, platform and storage medium based on structured evaluation data, so as to make full use of the point cloud encryption method based on the consistency of tunnel slice structure, template prior information and strip constraint relationship, so that the encryption result is distributed along the original sparse skeleton as a whole, and can be stably completed in local geometry, and the bottom region is specially processed.

[0008] The first objective of this application is to provide a method for generating traceable text based on structured evaluation data; the second objective of this application is to provide a system for generating traceable text based on structured evaluation data; the third objective of this application is to provide a platform for generating traceable text based on structured evaluation data; and the fourth objective of this application is to provide a computer-readable storage medium.

[0009] The primary objective of this application is achieved by the method comprising the steps of:

[0010] Create, generate, and acquire structured assessment data corresponding to the assessed object; wherein the structured assessment data contains project metrics corresponding to at least one assessment time point;

[0011] The project metrics are validated for legality according to preset scoring logic rules. After the validity validation is passed, the evaluation status feature information corresponding to the evaluated object is calculated based on the project metrics. The evaluation status feature information is used to characterize the quantitative level of the evaluated object's ability in a predetermined evaluation domain.

[0012] Based on the evaluation state feature information, reference knowledge information associated with the evaluation state feature information is retrieved from the knowledge base; the project metric, the evaluation state feature information, and the reference knowledge information are jointly constructed into input corpus, and a reasoning request carrying generation constraints is sent to the generative model; wherein, the generation constraints are used to constrain the generative model's rewriting operation of factual data in the input corpus;

[0013] The system receives the target description text generated by the generative model in response to the inference request, and performs a traceability check on the target description text; wherein, the traceability check includes: extracting factual information items to be checked from the target description text, and comparing the factual information items with the corresponding original fields in the structured evaluation data for consistency.

[0014] After the traceability verification is passed, the target description text is combined with the evaluation graphic generated based on the structured evaluation data, and an evaluation report file containing traceability identifiers is constructed.

[0015] Furthermore, the step of validating the project metrics according to preset scoring logic rules, and calculating and generating evaluation status feature information corresponding to the evaluated object based on the project metrics after the validity check is passed, also includes at least one of the following verification strategies:

[0016] Determine whether a single item metric value belongs to a preset set of valid metrics values;

[0017] Determine whether the sum of all project metrics within the same assessment domain is equal to the domain-summary metric of the assessment domain.

[0018] Determine whether there are any missing preset key item metrics in the structured evaluation data.

[0019] Furthermore, in the process of verifying the legality of the project metric value according to the preset scoring logic rules, and after the legality verification is passed, the evaluation status feature information corresponding to the evaluated object is calculated based on the project metric value. The evaluation status feature information includes the current stage measurement benchmark and the next stage measurement benchmark.

[0020] The method further includes:

[0021] The progress status of the evaluated object in at least one evaluation area is determined based on the current stage measurement benchmark;

[0022] When the progress status indicator has reached the predetermined highest measurement benchmark, the corresponding evaluation area will be associated with the completion status.

[0023] When the progress status indication does not reach the predetermined highest metric benchmark, the corresponding capability description information is extracted from the knowledge base according to the current stage metric benchmark and the next stage metric benchmark.

[0024] Furthermore, the step of recalling reference knowledge information associated with the evaluation state feature information from the knowledge base based on the evaluation state feature information; constructing the project metric, the evaluation state feature information, and the reference knowledge information together as input corpus; and sending a reasoning request carrying generation constraints to the generative model, further includes:

[0025] Based on the evaluation status feature information, corresponding retrieval constraint labels are generated;

[0026] Based on the search constraint tags, a preliminary screening is performed in the knowledge base to create a corresponding set of candidate knowledge.

[0027] Calculate the similarity between the feature vector corresponding to the generated evaluation state feature information and the vector of each knowledge fragment in the candidate knowledge set;

[0028] Based on the similarity, target knowledge fragments that meet the preset recall number are selected from the candidate knowledge set.

[0029] Furthermore, the step of receiving the target description text generated by the generative model in response to the inference request, and performing traceability verification on the target description text, further includes:

[0030] Fact information items to be verified are extracted from the target description text; wherein, the fact information items include at least a measurement value, an item name, or a date identifier;

[0031] Determine whether the factual information item is consistent with the corresponding original information item in the structured evaluation data; if they are inconsistent, perform at least one of the following traceability assurance operations on the target description text: discard the target description text and trigger a regeneration instruction, overwrite the original information item into the target description text to replace the inconsistent factual information item, or call a preset deterministic template to replace the corresponding text paragraph in the target description text.

[0032] Furthermore, the method also includes:

[0033] Generate and acquire the evaluation status feature information of the evaluated object at multiple evaluation time points;

[0034] Based on the assessment status characteristics of each of the multiple assessment time points, trend quantitative data representing the capability changes of the assessed object is created and generated.

[0035] An evaluation trend chart containing multiple overlays is generated based on the trend quantification data, wherein each layer corresponds to a different evaluation time point.

[0036] The second objective of this application is achieved as follows: the system is used to implement the traceable text generation method based on structured evaluation data, the system comprising:

[0037] A generation unit is created to generate and acquire structured evaluation data corresponding to the evaluated object; wherein the structured evaluation data includes project metrics corresponding to at least one evaluation time point;

[0038] The verification processing unit is used to verify the legality of the project metric value according to the preset scoring logic rules, and after the legality verification is passed, to calculate and generate evaluation status feature information corresponding to the evaluated object based on the project metric value; wherein, the evaluation status feature information is used to characterize the quantitative level of the evaluated object's ability in a predetermined evaluation domain.

[0039] The recall and sending unit is used to recall reference knowledge information associated with the evaluation state feature information from the knowledge base based on the evaluation state feature information; and to construct the project metric, the evaluation state feature information and the reference knowledge information together as input corpus, and send a reasoning request carrying generation constraints to the generative model; wherein, the generation constraints are used to constrain the generative model to rewrite the factual data in the input corpus;

[0040] A receiving verification unit is configured to receive the target description text generated by the generative model in response to the inference request, and perform traceability verification on the target description text; wherein, the traceability verification includes: extracting factual information items to be verified from the target description text, and comparing the factual information items with the corresponding original fields in the structured evaluation data for consistency.

[0041] The synthesis processing unit is used to synthesize the target description text with the evaluation graphics generated based on the structured evaluation data after the traceability verification is passed, and to construct an evaluation report file containing traceability identifiers.

[0042] Furthermore, the verification processing unit also includes at least one of the following verification strategy functional modules:

[0043] The first determination module is used to determine whether a single item metric value belongs to a preset set of legal metric values;

[0044] The second determination module is used to determine whether the sum of all project metrics within the same evaluation domain is equal to the domain summary metric of the evaluation domain.

[0045] The third determination module is used to determine whether there are any missing preset key item metrics in the structured evaluation data;

[0046] The assessment status characteristic information includes the current stage measurement benchmark and the next stage measurement benchmark;

[0047] The system also includes:

[0048] The progress status determination module is used to determine the progress status of the evaluated object in at least one evaluation area based on the current stage measurement benchmark.

[0049] The fourth determination module is used to associate the corresponding evaluation area as a completed state when the progress status indication has reached a predetermined highest measurement benchmark.

[0050] The fifth determination module is used to extract corresponding capability description information from the knowledge base based on the current stage measurement benchmark and the next stage measurement benchmark when the progress status indication has not reached the predetermined highest measurement benchmark.

[0051] And / or, the recall sending unit further includes:

[0052] The first generation module is used to construct and generate corresponding retrieval constraint tags based on the evaluation state feature information;

[0053] The second generation module is used to perform preliminary screening in the knowledge base based on the retrieval constraint tags and create a corresponding candidate knowledge set.

[0054] The third generation module is used to calculate the similarity between the feature vector corresponding to the generated evaluation state feature information and the vector of each knowledge fragment in the candidate knowledge set;

[0055] The selected module is used to select target knowledge fragments that meet the preset recall quantity from the candidate knowledge set based on the similarity.

[0056] And / or, the receiving verification unit further includes:

[0057] An extraction and generation module is used to extract and generate factual information items to be verified from the target description text; wherein, the factual information items include at least a measurement value, an item name, or a date identifier;

[0058] The sixth determination module is used to determine whether the factual information item is consistent with the corresponding original information item in the structured evaluation data; wherein, if they are inconsistent, at least one of the following traceability assurance operations is performed on the target description text: discarding the target description text and triggering a regeneration instruction, overwriting the original information item into the target description text to replace the inconsistent factual information item, or calling a preset deterministic template to replace the corresponding text paragraph in the target description text;

[0059] And / or, the system further includes:

[0060] The fourth generation module is used to generate and obtain the evaluation status feature information of the evaluated object at multiple evaluation time points;

[0061] The fifth generation module is used to create trend quantitative data representing the change in the capabilities of the evaluated object based on the evaluation status characteristic information of each of the multiple evaluation time points.

[0062] The sixth generation module is used to generate an evaluation trend chart containing multiple overlays based on the trend quantification data, wherein each layer corresponds to a different evaluation time point.

[0063] The third objective of this application is achieved as follows: it includes a processor, a memory, and a traceable text generation platform control program based on structured evaluation data; wherein the traceable text generation platform control program based on structured evaluation data is executed on the processor, the traceable text generation platform control program based on structured evaluation data is stored in the memory, and the traceable text generation platform control program based on structured evaluation data implements the traceable text generation method based on structured evaluation data.

[0064] The fourth objective of this application is achieved as follows: the computer-readable storage medium stores a traceable text generation platform control program based on structured evaluation data, and the traceable text generation platform control program based on structured evaluation data implements the traceable text generation method based on structured evaluation data.

[0065] This application's solution acquires structured assessment data and performs legality verification and quantification calculations on the original project metrics according to pre-defined scoring logic rules. This generates assessment status feature information that accurately represents the ability level of the assessed object, thus providing a reliable deterministic data foundation for subsequent processing. Then, using this assessment status feature information as a retrieval constraint, semantically related reference knowledge fragments are selectively retrieved from the knowledge base. The original factual data and the retrieved knowledge fragments are combined to construct an input corpus carrying generative constraints, thereby constraining the output boundary of the generative model. This ensures that the generated descriptive text integrates individualized quantitative assessment results and professional knowledge without affecting the output of the generative model. The original factual data is rewritten without any basis. Based on this, the factual information is extracted from the descriptive text output by the generative model and compared with the original structured evaluation data. When inconsistencies are found, the output is rejected, regeneration is triggered, or the original data is forcibly overwritten to correct the discrepancies. This establishes a closed-loop verification mechanism for factual accuracy at the end of the generation process. Finally, the verified descriptive text is combined with the evaluation graphics generated based on the structured data to form an evaluation report file with embedded traceable identifiers. This ensures that the final output report not only achieves individualized analysis and expression but also has traceability and reliability from the original data to the generated results. Attached Figure Description

[0066] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0067] Figure 1 This is a schematic diagram illustrating the process steps of a traceable text generation method based on structured evaluation data according to this application.

[0068] Figure 2 This is a schematic diagram of the traceable text generation system architecture based on structured evaluation data according to this application;

[0069] Figure 3 This is a schematic diagram of the architecture of an embodiment of a traceable text generation system based on structured evaluation data according to this application;

[0070] Figure 4 This is a schematic diagram of the traceable text generation platform architecture based on structured evaluation data, as described in this application.

[0071] Figure 5 This is a schematic diagram of a computer-readable storage medium architecture in one embodiment of this application. Detailed Implementation

[0072] To facilitate a clearer understanding of the purpose, technical solution, and advantages of this application, the following description, in conjunction with the accompanying drawings and specific embodiments, will further illustrate this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification.

[0073] This application can also be implemented or applied through other different specific examples, and the details in this specification can also be modified and changed based on different viewpoints and applications without departing from the spirit of this application.

[0074] It should be noted that if the embodiments of this application involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0075] Furthermore, if the embodiments of this application involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" may explicitly or implicitly include at least one of those features. Secondly, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0076] Preferably, the traceable text generation method based on structured evaluation data of this application is applied in one or more terminals or servers. The terminal is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0077] The terminal can be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal can interact with the customer via a keyboard, mouse, remote control, touchpad, or voice control device.

[0078] like Figure 1 The diagram shown is a flowchart of a traceable text generation method based on structured evaluation data provided in an embodiment of this application.

[0079] In this embodiment, the traceable text generation method based on structured evaluation data can be applied to terminals or fixed terminals with display functions. The terminals are not limited to personal computers, smartphones, tablets, desktop computers or all-in-one computers with cameras, etc.

[0080] The traceable text generation method based on structured assessment data can also be applied to a hardware environment consisting of a terminal and a server connected to the terminal via a network. The network includes, but is not limited to, a wide area network (WAN), a metropolitan area network (MAN), or a local area network (LAN). The traceable text generation method based on structured assessment data in this application embodiment can be executed by a server, by a terminal, or by both a server and a terminal.

[0081] For example, for terminals that require traceable text generation based on structured assessment data, the traceable text generation function based on structured assessment data provided by the method of this application can be directly integrated into the terminal, or a client for implementing the method of this application can be installed. Alternatively, the method provided by this application can also run on servers or other devices in the form of a Software Development Kit (SDK), providing an interface for the traceable text generation function based on structured assessment data in the form of an SDK. Terminals or other devices can then implement the traceable text generation function based on structured assessment data through the provided interface. The following description, in conjunction with the accompanying drawings, further illustrates this application.

[0082] like Figure 1 As shown, this application provides a method for generating traceable text based on structured evaluation data, the method comprising the following steps:

[0083] S01. Create, generate, and obtain structured evaluation data corresponding to the evaluated object; wherein, the structured evaluation data includes project metrics corresponding to at least one evaluation time point;

[0084] S02. The project metric is validated for legality according to the preset scoring logic rules. After the legality validation is passed, the evaluation status feature information corresponding to the evaluated object is calculated based on the project metric. The evaluation status feature information is used to characterize the quantitative level of the evaluated object's ability in the predetermined evaluation domain.

[0085] S03. Based on the evaluation state feature information, retrieve reference knowledge information associated with the evaluation state feature information from the knowledge base; and construct the project metric, the evaluation state feature information, and the reference knowledge information together as input corpus, and send a reasoning request carrying generation constraints to the generative model; wherein, the generation constraints are used to constrain the generative model's rewriting operation of factual data in the input corpus;

[0086] S04. Receive the target description text generated by the generative model in response to the inference request, and perform traceability verification on the target description text; wherein, the traceability verification includes: extracting factual information items to be verified from the target description text, and comparing the factual information items with the corresponding original fields in the structured evaluation data for consistency.

[0087] S05. After the traceability verification is passed, the target description text and the evaluation graphic generated based on the structured evaluation data are synthesized and processed to generate an evaluation report file containing traceability identifiers.

[0088] The process of validating the project metrics according to preset scoring logic rules, and calculating and generating evaluation status feature information corresponding to the evaluated object based on the project metrics after the validity check is passed, also includes at least one of the following verification strategies:

[0089] S021. Determine whether a single item metric value belongs to a preset set of valid metric values;

[0090] S022. Determine whether the sum of all project metrics within the same assessment area is equal to the sum of the assessment area's metrics.

[0091] S023. Determine whether there are any missing preset key item metrics in the structured evaluation data.

[0092] The process involves validating the project metrics according to preset scoring logic rules, and after the validity check is passed, calculating and generating evaluation status feature information corresponding to the evaluated object based on the project metrics. The evaluation status feature information includes the current stage measurement benchmark and the next stage measurement benchmark.

[0093] The method further includes:

[0094] S061. Determine the progress status of the evaluated object in at least one evaluation area based on the current stage measurement benchmark;

[0095] S062. When the progress status indication has reached the predetermined highest measurement benchmark, the corresponding evaluation area is associated with the completion status.

[0096] S063. When the progress status indication does not reach the predetermined highest measurement benchmark, extract the corresponding capability description information from the knowledge base according to the current stage measurement benchmark and the next stage measurement benchmark.

[0097] The step of recalling reference knowledge information associated with the evaluation state feature information from the knowledge base based on the evaluation state feature information, constructing the project metric, the evaluation state feature information, and the reference knowledge information together as input corpus, and sending an inference request carrying generation constraints to the generative model, further includes:

[0098] S031. Construct and generate corresponding retrieval constraint tags based on the evaluation state feature information;

[0099] S032. Based on the search constraint tags, perform preliminary screening in the knowledge base to create a corresponding candidate knowledge set;

[0100] S033. Calculate the similarity between the feature vector corresponding to the generated evaluation state feature information and the vector of each knowledge fragment in the candidate knowledge set;

[0101] S034. Based on the similarity, select target knowledge fragments from the candidate knowledge set that meet the preset recall quantity.

[0102] The step of receiving the target description text generated by the generative model in response to the inference request, and performing traceability verification on the target description text, further includes:

[0103] S041. Extract and generate factual information items to be verified from the target description text; wherein, the factual information items include at least a measurement value, an item name, or a date identifier;

[0104] S042. Determine whether the factual information item is consistent with the corresponding original information item in the structured evaluation data; wherein, if they are inconsistent, perform at least one of the following traceability assurance operations on the target description text: discard the target description text and trigger a regeneration instruction, overwrite the original information item into the target description text to replace the inconsistent factual information item, or call a preset deterministic template to replace the corresponding text paragraph in the target description text.

[0105] The method further includes:

[0106] S071. Generate and obtain the evaluation status feature information of the evaluated object at each of the multiple evaluation time points;

[0107] S072. Based on the assessment status characteristic information of each of the multiple assessment time points, create and generate trend quantitative data characterizing the capability changes of the assessed object;

[0108] S073. Generate an evaluation trend chart containing multiple overlays based on the trend quantification data, wherein each layer corresponds to a different evaluation time point.

[0109] Specifically, in this application embodiment, a method and system for generating individualized assessment reports based on structured behavioral assessment data and retrieval enhancement generation technology is provided to solve problems in the prior art such as high error rate of assessment data entry, difficulty in comparing assessments at multiple time points, inability to synchronize charts and reports, lack of individualization of generated content, and difficulty in traceability of generated results.

[0110] This application's solution achieves integrated processing from assessment data collection to assessment graph and report output through structured data modeling, scoring logic verification, multi-time-point layer mapping, targeted retrieval of assessment knowledge, automatic construction of prompts, and generation of constrained text. This application not only achieves integrated processing of assessment data, assessment graphs, and assessment reports, but also reduces the probability of omissions in assessment items, errors in score calculation, inconsistencies between graphs and text, and factual errors in the generative model output through deterministic scoring rules, structured knowledge retrieval, constrained model input, and post-generation factual consistency verification. Compared to manual processing, this application can shorten the time required for assessment data aggregation, assessment graph drawing, and initial report draft generation; compared to simply calling a general generative model, this application can improve the consistency of names, dates, scores, domain names, target items, and obstacle items in the generated text with the original assessment data. Specific comparative testing methods and results are detailed in the following specific embodiments and comparative verification experiments.

[0111] The method in this application embodiment includes:

[0112] S101. Establish data records for the evaluated object; receive the name, date of birth and other identification information of the evaluated object, and generate a unique identifier for the evaluated object.

[0113] S102. Establish an assessment record; receive the assessment date, test age, assessor, and assessment type, and generate a unique assessment identifier for each assessment so that different assessment time points for the same subject can be stored and associated separately.

[0114] S103. Receive structured project data; receive project scores according to pre-defined evaluation areas and project codes. Each data item includes at least the evaluated object identifier, evaluation identifier, area code, project code, and project score.

[0115] S104. Execute scoring logic verification; verify the input data based on the highest score in the corresponding field, the set of allowed scores, the half-score rules, and the project dependencies.

[0116] When an item's score exceeds the allowed range, the total score for the domain is inconsistent with the item's score, a necessary item is missing, or there is a pre-defined logical conflict between adjacent items, an exception flag is generated and the user is prompted to confirm or make corrections.

[0117] In one embodiment, let: Evaluation Area Index; Evaluation time point index; This application This application includes an index of assessment projects within its respective assessment area; This application The total number of assessment items included in each assessment area of ​​this application; This application At which time point in this application was the first application? This application is the first in the field of evaluation. The score for each item in this application; This application At which time point in this application was the first application? The total score for each evaluation area in this application.

[0118] For milestone assessment projects, the allowed score set can be set as follows:

[0119] (1)

[0120] in, This represents the set of allowed scores for milestone assessment projects.

[0121] For obstacle assessment items, the allowed score set can be set according to the scale scoring rules as follows:

[0122] (2)

[0123] in, This application represents the set of allowed scores for an obstacle assessment project.

[0124] For milestone assessment projects, when this application In this application, the item will be marked as an illegal score; for obstacle assessment items, when the item score does not belong to this application... In this application, the item was marked as an illegal score.

[0125] When a project's score does not belong to the corresponding set of allowed scores, the project is marked as an illegal score.

[0126] The rule processing module also performs validation based on the number of domain projects, the highest domain score, the order of projects, and the dependencies stored in the configuration table. The configuration table includes at least the domain code, project code, project order, allowed score, highest score, and project dependency conditions.

[0127] An anomaly flag is generated when any of the following situations are found: (1) the project score does not belong to the set of allowed scores; (2) the total score of the domain is inconsistent with the sum of the project scores; (3) a necessary project is missing; (4) the identifier of the evaluated object or the evaluation identifier does not exist; (5) the evaluation date is earlier than the date of birth; (6) the project number does not match the domain to which it belongs.

[0128] S105, Calculation of domain results; Total domain score: This application This application is the first of the evaluation time points in this application. The total score for each field in this application is calculated using the following formula:

[0129] (3)

[0130] in: This application represents the first application. This application is the first of the evaluation time points in this application. Total score for each field in this application; This application represents the first application. The total number of evaluation items included in each field of this application; This application represents the first application. This application is the first of the evaluation time points in this application. This application is the first in its field. The project score for this application.

[0131] Highest Score: The highest score is determined by the following formula:

[0132] (4)

[0133] in, This application represents the first application. At which time point in this application was the first application? The highest-scoring item in this application's field with a score greater than zero. When all items in this field have a score of zero, this application... .

[0134] Missing Items Set: Items that are located before the highest-scoring item but have a score of zero constitute the missing items set.

[0135] (5)

[0136] in, This application represents the first application. At which time point in this application was the first application? This application contains a set of item numbers within each field that are preceding the highest-scoring item but have a score of zero. The system will calculate the total field score for this application. The highest-scoring item in this application This application and the collection of missing items. This application is stored separately to avoid ignoring the differences in the distribution of projects within a field based solely on the total score of the field.

[0137] S106. The system determines the current target value and the next target value based on the total score in the domain:

[0138] (6)

[0139] (7)

[0140] in: This application represents the first application. This application is the first of the evaluation time points in this application. The current target value for this application in this field; This application represents a floor function; This application represents the first application. This application is the first of the evaluation time points in this application. The next target value in this application's field; This application represents the first application. The highest permissible target value in this application's respective fields; This application indicates that the smaller of the two values ​​is taken.

[0141] When this application In this application, the field is marked as terminated, and no further target is identified or searched; otherwise, according to this application... This application and this application This application retrieves the corresponding current capability descriptions and next-stage goals.

[0142] S107. Retrieve professional knowledge; construct retrieval criteria, wherein the retrieval criteria include at least one or more of the following: domain code, evaluation type, current target value, next target value, and obstacle item code.

[0143] The system filters knowledge fragments that match search criteria from the professional knowledge base and returns the text of the knowledge fragments along with their metadata. Metadata may include document type, domain code, project number, score, version number, and source identifier.

[0144] S108. Integrate multiple assessment time points; read data from multiple assessment dates based on the assessment identifier of the same assessed object, map each assessment data to different layers according to the assessment time sequence, and calculate the change in the total domain score between adjacent assessment time points according to the following formula:

[0145] (8)

[0146] in: This application represents the first application. This application is in the field described in this application. The total score for each evaluation point in this application; This application represents the total score for the same field at the previous assessment point in time; This application represents the change in total domain score between two adjacent assessment points.

[0147] When this application In this application, it indicates that the total score in that field increases; when this application In this application, it is indicated that the total score in that field has not changed; when this application In this application, it is indicated that the total score for that field has been reduced.

[0148] In addition to the change in the total score for each area, the system also compares the highest-scoring item in this application. Missing item set in this application And changes in half-point items, to distinguish evaluation results with the same total score but different distribution of specific items.

[0149] S109. Integrate milestone and obstacle information; associate competency areas with one or more obstacle items according to preset association rules. When the score of an obstacle item reaches a preset threshold, or the score of a certain competency area falls below a preset level, the corresponding competency objectives, obstacle explanations, and intervention references are added to the report generation input.

[0150] In another implementation, the system establishes an association rule table between capability domain codes and obstacle item codes. The association rule table includes at least the capability domain code, obstacle item code, obstacle trigger threshold, capability low score threshold, and association priority.

[0151] When an obstacle's score reaches the obstacle trigger threshold, the system adds that obstacle to the prominent obstacle set; when a capability's score falls below the capability low score threshold, the system adds that capability to the key capability set. Based on the association rule table, the system matches the capabilities in the key capability set with the items in the prominent obstacle set, and writes the matching results, corresponding capability objectives, obstacle explanations, and intervention references into the model input.

[0152] When there are no preset association rules, the system can also establish candidate associations based on the semantic similarity between knowledge fragments. However, candidate associations must be filtered through preset thresholds or confirmed by the user before they can be written into the formal report.

[0153] S110. Construct constrained model inputs; convert the basic information of the evaluated object, evaluation score, time changes, current goal, next goal, prominent obstacles, and retrieved knowledge fragments into structured inputs.

[0154] The structured input sets labels for factual data and reference knowledge, and provides restrictive instructions to the generative model, preventing the model from changing the original score, generating diagnostic conclusions not included in the evaluation data or retrieved knowledge, and outputting text according to a predetermined chapter structure.

[0155] S111. Generate individualized assessment text; the generative model generates one or more of the following based on the structured input: assessment summary, explanation of each competency domain, explanation of prominent obstacles, correlation analysis between competencies and obstacles, and subsequent intervention recommendations.

[0156] S112. Perform output consistency verification; After the generative model outputs text, the system extracts the name of the evaluated object, evaluation date, domain name, domain score, obstacle item name, obstacle score, current goal and next goal from the generated text through one or more of the following methods: structured field parsing, keyword matching, regular expression matching or natural language entity recognition.

[0157] For the fields to be verified in this application The system performs consistency comparisons according to the following formula:

[0158] (9)

[0159] in: This application represents the data field to be verified; This application represents field values ​​extracted from generated text or structured model output; This application represents the corresponding field values ​​in the original structured evaluation data; This application represents the field of this application. The consistency verification results of this application.

[0160] This application This application specifies the set of fields that must be validated. The generated text passes consistency validation when the following formula is satisfied:

[0161] (10)

[0162] When any field in this application must be validated In this application, the system performs at least one of the following processes:

[0163] 1. Reject the currently generated text; 2. Rewrite the erroneous and correct field values ​​into the model input and regenerate; 3. Directly replace the erroneous field values ​​with the original structured data; 4. Replace the corresponding paragraph with a deterministic text template.

[0164] The system simultaneously verifies whether the domain name, obstacle item name, and target item in the generated text belong to a preset name set. If the model generates a non-existent domain, item, or diagnostic conclusion, the output is deemed invalid.

[0165] Each generation and verification process saves the evaluation identifier, model name, model version, knowledge base version, prompt template version, generation time, verification result, and number of retries to support result traceability.

[0166] S113. Generate Report: Combine the assessed subject information, milestone assessment map, obstacle assessment map, assessment summary, domain analysis, and intervention recommendations according to a predefined template to generate an editable document, portable document, or web-based report. The report saves the assessment identifier, knowledge base version, and generation time corresponding to this generation to support result traceability.

[0167] Specific Implementation Example 1: Generation of Children's Language Behavior Assessment Report;

[0168] This embodiment uses an Android tablet as the user terminal, FastAPI to build the data interface service, and a PostgreSQL database to store the information of the evaluated object, evaluation records, domain scores and obstacle scores. It uses the "Fill in the actual embedding model" to generate knowledge fragment vectors, uses the "Fill in the actual generation model" to generate evaluation text, and outputs a Word format report through the document generation module.

[0169] The name of the person being evaluated, A, is replaced with the anonymous identifier "A". The date of birth is [fill in], the evaluation date is [fill in], the test age is [fill in], and the evaluator is [fill in].

[0170] Assessors input milestone and obstacle assessment data via Android terminals. Taking the "Fill in a Real-World Domain" domain as an example, the scores for each item in this domain are "Fill in Item Score". The system calculates the domain's total score as "Fill in", the highest-scoring item is "Fill in", the set of missing items is "Fill in", and the half-scoring items are "Fill in". The rule processing module checks whether each item's score belongs to the allowed score set and verifies whether the domain's total score equals the sum of the item scores. This processing checked a total of "Number of Fills in" items and found "Number of Fills in" abnormal items. Abnormal items were written to the assessment database after being confirmed or modified by the assessors. The knowledge retrieval module constructs a retrieval request based on the domain code "Fill in", the current target "Fill in", the next target "Fill in", and the missing item "Fill in". First, it filters for "Number of Fills in" candidate knowledge fragments through metadata, and then selects the "Top-K" knowledge fragments with the highest semantic similarity.

[0171] The structured model input construction module writes the assessed object information, evaluation score, current goal, next goal, missing items, obstacle scores, and retrieved knowledge fragments into the structured input, and adds restrictions such as prohibiting score modification, prohibiting diagnosis generation, distinguishing half-point items, and outputting according to a predetermined structure. The generative model generates an evaluation summary, domain analysis, obstacle analysis, and intervention suggestions. The consistency verification module extracts the name, evaluation date, domain score, obstacle score, and target item from the generated text and compares it with the original structured data. If any errors are found during the first verification (no errors / specific errors), the system executes (pass / regenerate / field replacement). The graph generation module generates milestone evaluation maps and obstacle evaluation maps based on the structured data. The report generation module merges the evaluation maps with the text that passed consistency verification to generate a Word format evaluation report. The report saves the evaluation identifier (filled in), knowledge base version (filled in), model version (filled in), and generation time (filled in).

[0172] Comparative Validation Experiment: To verify the technical effectiveness of this application in terms of assessment data processing, assessment chart generation, and assessment report generation, five senior assessors with approximately ten years or more of experience in children's language behavior assessment and who assume clinical or teaching supervision responsibilities were selected for comparative testing. The test used complete and anonymized children's language behavior assessment data. The participating assessors completed the assessment data aggregation, assessment result interpretation, assessment chart creation, intervention suggestion writing, and report compilation using traditional manual methods. Simultaneously, the same structured assessment data was input into the system described in this application, which automatically completed scoring verification, domain result calculation, milestone assessment chart generation, obstacle assessment chart generation, professional knowledge retrieval, individualized assessment text generation, consistency verification, and editable report output.

[0173] Manual processing method: According to the test, when five senior assessors complete an assessment report using the traditional manual method, it takes an average of about 300 minutes to manually analyze the assessment results, review the assessment project description and write the assessment analysis and intervention recommendations; it takes an average of about 10 minutes to manually summarize the data and create the milestone assessment chart and obstacle assessment chart; and it takes an average of about 5 minutes to format, export and prepare the report for printing.

[0174] Therefore, the average total processing time for completing a full assessment report using traditional manual methods is:

[0175] (11)

[0176] in, This represents the average total processing time for completing a full assessment report using traditional manual methods.

[0177] The system's processing method for this application: Given that the evaluators have already completed the structured evaluation data entry, this system can automatically complete the following processing:

[0178] The system validates project scores, total domain scores, and missing items; calculates total scores for each domain, highest-scoring items, missing items, current goals, and next goals; generates milestone assessment maps and obstacle assessment maps; retrieves corresponding professional knowledge fragments based on assessment domains, project scores, and obstacle items; constructs constrained generative model inputs; generates assessment summaries, domain analyses, obstacle analyses, and intervention recommendations; performs consistency validation on names, assessment dates, domain scores, obstacle scores, and target items; and combines assessment maps and assessment text into an editable assessment report.

[0179] According to testing, the system takes an average of approximately 2 minutes from receiving complete structured assessment data to outputting an editable report containing assessment charts, assessment analysis, and intervention recommendations.

[0180] (12)

[0181] in, This indicates the average processing time for the system to generate a complete evaluation report.

[0182] Comparison Results: The average processing time of the traditional manual method and the system of this application are compared in the table below:

[0183] Evaluation Result Analysis and Report Writing Approximately 300 minutes Auto complete Milestone and obstacle map creation Approximately 10 minutes Auto complete Report formatting, exporting, and printing preparation Approximately 5 minutes Auto complete Average time to generate a full report Approximately 315 minutes Approximately 2 minutes

[0184] The reduction in average processing time is calculated using the following formula:

[0185] (13)

[0186] in: This represents the average processing time using traditional manual methods; This indicates the average processing time of the system in this application; This indicates the percentage reduction in average processing time after adopting the system described in this application.

[0187] Substitute the test results into the formula above:

[0188] (14)

[0189] Therefore, compared with the traditional manual method, the system of this application reduces the average processing time of a complete evaluation report from about 315 minutes to about 2 minutes, which is a reduction of about 99.37%.

[0190] The processing speed multiplier is calculated using the following formula:

[0191] (15)

[0192] in, This indicates the ratio of the processing speed of the system in this application to the traditional manual method.

[0193] Substitute the test results into the formula above:

[0194] (16)

[0195] Therefore, under the conditions of this comparative test, the report generation and processing speed of the system in this application is approximately 157.5 times that of the traditional manual method.

[0196] The test results above show that this application shortens the processing time required for assessment data aggregation, graphics production, and report generation through integrated processing of structured assessment data processing, deterministic scoring verification, automatic graphics generation, professional knowledge retrieval, constrained text generation, and consistency verification.

[0197] Furthermore, the user terminal of this application can be a tablet computer, mobile phone, computer, or web terminal. The backend system can be deployed on a local server, an internal server, a cloud server, or a hybrid deployment environment. The generative model can be a locally deployed language model, a cloud-based language model, or a model routing system composed of multiple language models. Knowledge retrieval can employ structured database queries, keyword retrieval, vector retrieval, knowledge graph retrieval, or a combination thereof. The current and next objectives can be determined based on the domain total score, specific item scores, the highest consecutive score item, or unmastered items. The evaluation graph can be a matrix diagram, bar chart, line chart, radar chart, color overlay map, or other graphs capable of representing changes across multiple evaluations. Reports can be generated in Word, PDF, HTML, JSON, or other machine-readable formats. The correlation between milestones and obstacles can be determined through preset rules, statistical models, machine learning models, semantic similarity models, or hybrid methods. The generated text can be generated by the language model or jointly generated by rule templates and the language model. The system can also receive voice, images, videos, or observation records and convert them into structured evaluation data. This application is not limited to a specific assessment scale, but can also be applied to other child development, language, behavior, or education assessment tools that consist of multiple domains, items, scores, and assessment time points.

[0198] To achieve the purpose of this application, such as Figure 3 As shown, a traceable text generation system based on structured assessment data is also provided. This system is applied to the traceable text generation method based on structured assessment data. The system includes:

[0199] A generation unit is created to generate and acquire structured evaluation data corresponding to the evaluated object; wherein the structured evaluation data includes project metrics corresponding to at least one evaluation time point;

[0200] The verification processing unit is used to verify the legality of the project metric value according to the preset scoring logic rules, and after the legality verification is passed, to calculate and generate evaluation status feature information corresponding to the evaluated object based on the project metric value; wherein, the evaluation status feature information is used to characterize the quantitative level of the evaluated object's ability in a predetermined evaluation domain.

[0201] The recall and sending unit is used to recall reference knowledge information associated with the evaluation state feature information from the knowledge base based on the evaluation state feature information; and to construct the project metric, the evaluation state feature information and the reference knowledge information together as input corpus, and send a reasoning request carrying generation constraints to the generative model; wherein, the generation constraints are used to constrain the generative model to rewrite the factual data in the input corpus;

[0202] A receiving verification unit is configured to receive the target description text generated by the generative model in response to the inference request, and perform traceability verification on the target description text; wherein, the traceability verification includes: extracting factual information items to be verified from the target description text, and comparing the factual information items with the corresponding original fields in the structured evaluation data for consistency.

[0203] The synthesis processing unit is used to synthesize the target description text with the evaluation graphics generated based on the structured evaluation data after the traceability verification is passed, and to construct an evaluation report file containing traceability identifiers.

[0204] The verification processing unit further includes at least one of the following verification strategy functional modules:

[0205] The first determination module is used to determine whether a single item metric value belongs to a preset set of legal metric values;

[0206] The second determination module is used to determine whether the sum of all project metrics within the same evaluation domain is equal to the domain summary metric of the evaluation domain.

[0207] The third determination module is used to determine whether there are any missing preset key item metrics in the structured evaluation data;

[0208] The assessment status characteristic information includes the current stage measurement benchmark and the next stage measurement benchmark;

[0209] The system also includes:

[0210] The progress status determination module is used to determine the progress status of the evaluated object in at least one evaluation area based on the current stage measurement benchmark.

[0211] The fourth determination module is used to associate the corresponding evaluation area as a completed state when the progress status indication has reached a predetermined highest measurement benchmark.

[0212] The fifth determination module is used to extract corresponding capability description information from the knowledge base based on the current stage measurement benchmark and the next stage measurement benchmark when the progress status indication has not reached the predetermined highest measurement benchmark.

[0213] And / or, the recall sending unit further includes:

[0214] The first generation module is used to construct and generate corresponding retrieval constraint tags based on the evaluation state feature information;

[0215] The second generation module is used to perform preliminary screening in the knowledge base based on the retrieval constraint tags and create a corresponding candidate knowledge set.

[0216] The third generation module is used to calculate the similarity between the feature vector corresponding to the generated evaluation state feature information and the vector of each knowledge fragment in the candidate knowledge set;

[0217] The selected module is used to select target knowledge fragments that meet the preset recall quantity from the candidate knowledge set based on the similarity.

[0218] And / or, the receiving verification unit further includes:

[0219] An extraction and generation module is used to extract and generate factual information items to be verified from the target description text; wherein, the factual information items include at least a measurement value, an item name, or a date identifier;

[0220] The sixth determination module is used to determine whether the factual information item is consistent with the corresponding original information item in the structured evaluation data; wherein, if they are inconsistent, at least one of the following traceability assurance operations is performed on the target description text: discarding the target description text and triggering a regeneration instruction, overwriting the original information item into the target description text to replace the inconsistent factual information item, or calling a preset deterministic template to replace the corresponding text paragraph in the target description text;

[0221] And / or, the system further includes:

[0222] The fourth generation module is used to generate and obtain the evaluation status feature information of the evaluated object at multiple evaluation time points;

[0223] The fifth generation module is used to create trend quantitative data representing the change in the capabilities of the evaluated object based on the evaluation status characteristic information of each of the multiple evaluation time points.

[0224] The sixth generation module is used to generate an evaluation trend chart containing multiple overlays based on the trend quantification data, wherein each layer corresponds to a different evaluation time point.

[0225] Specifically, in this application, the system includes: a user terminal for receiving information about the assessed individuals and assessment project data; a data interface module for converting data submitted by the terminal into a predetermined data format; an assessment data storage module for storing information about the assessed individuals, assessment date, test age, assessors, project scores, domain total scores, and obstacle scores; a rule processing module for performing score validity verification, project integrity verification, domain total score calculation, and target project determination; a graphics generation module for converting data from one or more assessment time points into assessment graphs; and a professional knowledge preprocessing and storage module for segmenting assessment scale descriptions, assessment project definitions, ability descriptions, obstacle explanations, and intervention reference materials according to assessment domains, assessment projects, or complete semantic paragraphs, and establishing metadata for each knowledge segment. The metadata includes at least one or more of the following: document type, assessment type, domain code, project number, corresponding score, obstacle project code, version number, and source identifier. The knowledge retrieval module constructs search criteria based on domain encoding, current score, current goal, next goal, missing items, obstacle items, and assessment type. It first filters candidate knowledge fragments based on metadata, then determines the knowledge fragments relevant to the assessment result through keyword matching, vector similarity calculation, or a combination of both. The structured model input construction module writes the basic information of the assessed object, assessment time, item score, total domain score, missing items, changes over multiple time points, current goal, next goal, prominent obstacles, and retrieved knowledge fragments into predetermined data fields, and adds model instructions prohibiting modification of the original data, prohibiting the generation of models without diagnostic conclusions, and restricting the output format. The generative model invocation module sends structured model input to a locally or remotely deployed generative language model and receives assessment text generated according to a predetermined chapter structure. This embodiment does not require parameter fine-tuning as a necessary condition, but rather restricts model output through enhanced retrieval generation, structured prompt templates, output format constraints, and post-generation verification. The consistency verification module extracts name, date, score, domain name, project number, current goal, next goal, and obstacle project from the generated text or structured output, and compares them item by item with the original structured assessment data. When inconsistencies are found, one or more of the following actions are taken: rejecting the output, regenerating, replacing the erroneous field with the original data, or calling a deterministic text template. The graph generation module directly generates milestone assessment graphs, obstacle assessment graphs, and multi-timepoint comparison graphs based on the structured data in the assessment database and preset graph mapping rules. This graph generation process does not rely on a generative language model. The report generation module combines the assessed object information, assessment graphs, consistency-verified assessment text, and intervention references into an editable document, a fixed-format document, or a web page report, and saves the corresponding assessment identifier, knowledge base version, prompt template version, model version, and generation time.

[0226] In the system solution embodiment of this application, the specific details of the method steps involved in the generation of traceable text based on structured evaluation data have been described above. That is to say, the functional modules in the system are used to implement the steps or sub-steps in the above method embodiment, and will not be repeated here.

[0227] To achieve the above objectives, this application also provides a traceable text generation platform based on structured assessment data, such as... Figure 4 As shown, it includes a processor, a memory, and a traceable text generation platform control program based on structured evaluation data; wherein, the traceable text generation platform control program based on structured evaluation data is executed by the processor, and the traceable text generation platform control program based on structured evaluation data is stored in the memory. The traceable text generation platform control program based on structured evaluation data implements the steps of the traceable text generation method based on structured evaluation data. The specific details of the steps have been described above and will not be repeated here.

[0228] In this embodiment, the built-in processor of the traceable text generation platform based on structured assessment data can be composed of integrated circuits. For example, it can be composed of a single packaged integrated circuit, or multiple integrated circuits packaged with the same or different functions. This includes combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor connects various components using various interfaces and lines, and executes programs or units stored in memory, as well as calling data stored in memory, to perform various functions of traceable text generation based on structured assessment data and process data.

[0229] The memory is used to store program code and various data, installed in a traceable text generation platform based on structured evaluation data, and enables high-speed, automatic access to programs or data during operation. The memory includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0230] To achieve the above objectives, this application also provides a computer-readable storage medium, such as... Figure 5 As shown, the computer-readable storage medium stores a control program for a traceable text generation platform based on structured assessment data. The control program for the traceable text generation platform based on structured assessment data implements the steps of the traceable text generation method based on structured assessment data. The specific details of the steps have been described above and will not be repeated here.

[0231] In the description of the embodiments of this application, it should be noted that any process or method description in the flowcharts or otherwise described herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order according to the functions involved, as should be understood by those skilled in the art to which the embodiments of this application pertain.

[0232] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processing module, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, a “computer-readable medium” can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, the computer-readable medium can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0233] This application's solution acquires structured assessment data and performs legality verification and quantification calculations on the original project metrics according to pre-defined scoring logic rules. This generates assessment status feature information that accurately represents the ability level of the assessed object, thus providing a reliable deterministic data foundation for subsequent processing. Then, using this assessment status feature information as a retrieval constraint, semantically related reference knowledge fragments are selectively retrieved from the knowledge base. The original factual data and the retrieved knowledge fragments are combined to construct an input corpus carrying generative constraints, thereby constraining the output boundary of the generative model. This ensures that the generated descriptive text integrates individualized quantitative assessment results and professional knowledge without affecting the output of the generative model. The original factual data is rewritten without any basis. Based on this, the factual information is extracted from the descriptive text output by the generative model and compared with the original structured evaluation data. When inconsistencies are found, the output is rejected, regeneration is triggered, or the original data is forcibly overwritten to correct the discrepancies. This establishes a closed-loop verification mechanism for factual accuracy at the end of the generation process. Finally, the verified descriptive text is combined with the evaluation graphics generated based on the structured data to form an evaluation report file with embedded traceable identifiers. This ensures that the final output report not only achieves individualized analysis and expression but also has traceability and reliability from the original data to the generated results.

[0234] In other words, this application achieves intelligent generation of evaluation reports from raw data to professional knowledge by constructing a closed-loop technology chain consisting of structured data acquisition, deterministic rule verification and quantitative evaluation, domain knowledge-oriented retrieval enhancement, constrained generative text reasoning, and automatic fact consistency verification. In terms of efficiency, this solution completely replaces the cumbersome operations of traditional manual scoring, charting, data retrieval, and report writing with automated processes, significantly reducing labor costs and shortening processing cycles. In terms of personalization, this solution breaks through the existing paradigm of templated output corresponding to fixed score ranges. Based on the unique ability distribution map, skill deficiency set, and dynamic change trends of the assessed individuals, it accurately retrieves semantically related reference knowledge from a professional knowledge base, achieving personalized and precise analysis and suggestion generation. In terms of reliability, this solution constructs a feedback verification mechanism of generation, extraction, comparison, and correction, verifying the factual information in the generated text against the original structured data item by item. Inconsistent content is rejected, regenerated, or definitively replaced, fundamentally suppressing the risks of factual deviation and content illusion commonly found in general language models in vertical applications. Simultaneously, the traceable meta-information embedded in the report allows each output to be linked to the original data version, knowledge base version, and model version, ensuring that the generated results have complete interpretability and traceability in clinical compliance and professional auditing scenarios. This application organically integrates the determinism of computer processing with the generalization ability of generative artificial intelligence, providing a systematic technical solution for the field of evaluation report generation that combines high efficiency, high degree of individualization, and high reliability.

[0235] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for generating traceable text based on structured assessment data, characterized in that, The method includes: Create, generate, and acquire structured assessment data corresponding to the assessed object; wherein the structured assessment data contains project metrics corresponding to at least one assessment time point; The project metrics are validated for legality according to preset scoring logic rules. After the validity validation is passed, the evaluation status feature information corresponding to the evaluated object is calculated based on the project metrics. The evaluation status feature information is used to characterize the quantitative level of the evaluated object's ability in a predetermined evaluation domain. Based on the evaluation state feature information, reference knowledge information associated with the evaluation state feature information is retrieved from the knowledge base; the project metric, the evaluation state feature information, and the reference knowledge information are jointly constructed into input corpus, and a reasoning request carrying generation constraints is sent to the generative model; wherein, the generation constraints are used to constrain the generative model's rewriting operation of factual data in the input corpus; The system receives the target description text generated by the generative model in response to the inference request, and performs a traceability check on the target description text; wherein, the traceability check includes: extracting factual information items to be checked from the target description text, and comparing the factual information items with the corresponding original fields in the structured evaluation data for consistency. After the traceability verification is passed, the target description text is combined with the evaluation graphic generated based on the structured evaluation data, and an evaluation report file containing traceability identifiers is constructed.

2. The method for generating traceable text based on structured assessment data according to claim 1, characterized in that, The process of validating the project metrics according to preset scoring logic rules, and calculating and generating evaluation status feature information corresponding to the evaluated object based on the project metrics after the validity check is passed, also includes at least one of the following verification strategies: Determine whether a single item metric value belongs to a preset set of valid metrics values; Determine whether the sum of all project metrics within the same assessment domain is equal to the domain-summary metric of the assessment domain. Determine whether there are any missing preset key item metrics in the structured evaluation data.

3. A method for generating traceable text based on structured assessment data according to claim 1 or 2, characterized in that, The process involves validating the project metrics according to preset scoring logic rules, and after the validity check is passed, calculating and generating evaluation status feature information corresponding to the evaluated object based on the project metrics. The evaluation status feature information includes the current stage measurement benchmark and the next stage measurement benchmark. The method further includes: The progress status of the evaluated object in at least one evaluation area is determined based on the current stage measurement benchmark; When the progress status indicator has reached the predetermined highest measurement benchmark, the corresponding evaluation area will be associated with the completion status. When the progress status indication does not reach the predetermined highest metric benchmark, the corresponding capability description information is extracted from the knowledge base according to the current stage metric benchmark and the next stage metric benchmark.

4. The method for generating traceable text based on structured assessment data according to claim 1, characterized in that, The step of recalling reference knowledge information associated with the evaluation state feature information from the knowledge base based on the evaluation state feature information, constructing the project metric, the evaluation state feature information, and the reference knowledge information together as input corpus, and sending an inference request carrying generation constraints to the generative model, further includes: Based on the evaluation status feature information, corresponding retrieval constraint labels are generated; Based on the search constraint tags, a preliminary screening is performed in the knowledge base to create a corresponding set of candidate knowledge. Calculate the similarity between the feature vector corresponding to the generated evaluation state feature information and the vector of each knowledge fragment in the candidate knowledge set; Based on the similarity, target knowledge fragments that meet the preset recall number are selected from the candidate knowledge set.

5. A method for generating traceable text based on structured assessment data according to claim 1 or 4, characterized in that, The step of receiving the target description text generated by the generative model in response to the inference request, and performing traceability verification on the target description text, further includes: Fact information items to be verified are extracted from the target description text; wherein, the fact information items include at least a measurement value, an item name, or a date identifier; Determine whether the factual information item is consistent with the corresponding original information item in the structured evaluation data; if they are inconsistent, perform at least one of the following traceability assurance operations on the target description text: discard the target description text and trigger a regeneration instruction, overwrite the original information item into the target description text to replace the inconsistent factual information item, or call a preset deterministic template to replace the corresponding text paragraph in the target description text.

6. The method for generating traceable text based on structured assessment data according to claim 1, characterized in that, The method further includes: Generate and acquire the evaluation status feature information of the evaluated object at multiple evaluation time points; Based on the assessment status characteristics of each of the multiple assessment time points, trend quantitative data representing the capability changes of the assessed object is created and generated. An evaluation trend chart containing multiple overlays is generated based on the trend quantification data, wherein each layer corresponds to a different evaluation time point.

7. A traceable text generation system based on structured assessment data, characterized in that, The system is applied to the traceable text generation method based on structured evaluation data as described in any one of claims 1 to 6, and the system comprises: A generation unit is created to generate and acquire structured evaluation data corresponding to the evaluated object; wherein the structured evaluation data includes project metrics corresponding to at least one evaluation time point; The verification processing unit is used to verify the legality of the project metric value according to the preset scoring logic rules, and after the legality verification is passed, to calculate and generate evaluation status feature information corresponding to the evaluated object based on the project metric value; wherein, the evaluation status feature information is used to characterize the quantitative level of the evaluated object's ability in a predetermined evaluation domain. The recall and sending unit is used to recall reference knowledge information associated with the evaluation state feature information from the knowledge base based on the evaluation state feature information; and to construct the project metric, the evaluation state feature information and the reference knowledge information together as input corpus, and send a reasoning request carrying generation constraints to the generative model; wherein, the generation constraints are used to constrain the generative model to rewrite the factual data in the input corpus; A receiving verification unit is configured to receive the target description text generated by the generative model in response to the inference request, and perform traceability verification on the target description text; wherein, the traceability verification includes: extracting factual information items to be verified from the target description text, and comparing the factual information items with the corresponding original fields in the structured evaluation data for consistency. The synthesis processing unit is used to synthesize the target description text with the evaluation graphics generated based on the structured evaluation data after the traceability verification is passed, and to construct an evaluation report file containing traceability identifiers.

8. A traceable text generation system based on structured assessment data according to claim 7, characterized in that, The verification processing unit further includes at least one of the following verification strategy functional modules: The first determination module is used to determine whether a single item metric value belongs to a preset set of legal metric values; The second determination module is used to determine whether the sum of all project metrics within the same evaluation domain is equal to the domain summary metric of the evaluation domain. The third determination module is used to determine whether there are any missing preset key item metrics in the structured evaluation data; The assessment status characteristic information includes the current stage measurement benchmark and the next stage measurement benchmark; The system also includes: The progress status determination module is used to determine the progress status of the evaluated object in at least one evaluation area based on the current stage measurement benchmark. The fourth determination module is used to associate the corresponding evaluation area as a completed state when the progress status indication has reached a predetermined highest measurement benchmark. The fifth determination module is used to extract corresponding capability description information from the knowledge base based on the current stage measurement benchmark and the next stage measurement benchmark when the progress status indication has not reached the predetermined highest measurement benchmark. And / or, the recall sending unit further includes: The first generation module is used to construct and generate corresponding retrieval constraint tags based on the evaluation state feature information; The second generation module is used to perform preliminary screening in the knowledge base based on the retrieval constraint tags and create a corresponding candidate knowledge set. The third generation module is used to calculate the similarity between the feature vector corresponding to the generated evaluation state feature information and the vector of each knowledge fragment in the candidate knowledge set; The selected module is used to select target knowledge fragments that meet the preset recall quantity from the candidate knowledge set based on the similarity. And / or, the receiving verification unit further includes: An extraction and generation module is used to extract and generate factual information items to be verified from the target description text; wherein, the factual information items include at least a measurement value, an item name, or a date identifier; The sixth determination module is used to determine whether the factual information item is consistent with the corresponding original information item in the structured evaluation data; wherein, if they are inconsistent, at least one of the following traceability assurance operations is performed on the target description text: discarding the target description text and triggering a regeneration instruction, overwriting the original information item into the target description text to replace the inconsistent factual information item, or calling a preset deterministic template to replace the corresponding text paragraph in the target description text; And / or, the system further includes: The fourth generation module is used to generate and obtain the evaluation status feature information of the evaluated object at multiple evaluation time points; The fifth generation module is used to create trend quantitative data representing the change in the capabilities of the evaluated object based on the evaluation status characteristic information of each of the multiple evaluation time points. The sixth generation module is used to generate an evaluation trend chart containing multiple overlays based on the trend quantification data, wherein each layer corresponds to a different evaluation time point.

9. A traceable text generation platform based on structured assessment data, characterized in that, The system includes a processor, a memory, and a traceable text generation platform control program based on structured evaluation data; wherein the traceable text generation platform control program based on structured evaluation data is executed on the processor, the traceable text generation platform control program based on structured evaluation data is stored in the memory, and the traceable text generation platform control program based on structured evaluation data implements the traceable text generation method based on structured evaluation data as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a traceable text generation platform control program based on structured evaluation data, which implements the traceable text generation method based on structured evaluation data as described in any one of claims 1 to 6.

Images