A method for automatically checking the consistency of a test report

Abstract

This invention relates to the field of data processing technology, and in particular to an automatic verification method for the consistency of inspection reports. The invention involves inputting the elements to be verified from the target electronic report into a pre-trained indicator library, outputting structured indicator fields; comparing the structured indicator fields with the original data source to obtain comparison results; collecting data credibility feature information from the comparison results and calculating data credibility feature representation values; determining whether the validity of the data acquisition equipment meets the standards based on the data credibility feature representation values; responding to the validity meeting the standards, collecting factual matching degree feature information from the corresponding caliber and calculating factual matching degree feature representation values; determining whether the stability of the verification device meets the standards based on the factual matching feature representation values; if the stability does not meet the standards, calculating the matching deviation value; and determining the reasons for the instability not meeting the standards and the corresponding handling strategies based on the matching deviation value. This invention improves the accuracy of report consistency verification.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an automatic verification method for the consistency of test reports. Background Technology

[0002] In the context of compiling and reviewing periodic reports for special equipment such as pressure vessels, lifting machinery, and elevators in construction projects, the accuracy of these reports directly impacts public safety. Report compilation typically involves multiple stages, including data collection, rule referencing, and conclusion determination. Currently, report verification relies primarily on manual review or simple rule comparison. Manual review is inefficient, susceptible to subjective influences, and struggles to ensure consistency across reports and batches. While rule-based automated verification methods improve efficiency to some extent, they often only verify a single dimension, such as checking data format or comparing limit references, lacking systematic verification of the authenticity of data sources and the logical consistency of conclusions. Furthermore, existing methods often employ a one-size-fits-all approach during verification, failing to differentiate the causes of errors in different data types. This results in a lack of interpretability in verification results, hindering subsequent accurate remediation and accountability, and severely impacting report quality control and compliance audit efficiency.

[0003] Chinese Patent Publication No. CN121093916A discloses an intelligent report generation method and system based on multi-source heterogeneous data fusion, comprising the following steps: First, fact anchor construction: identifying core facts in multi-source heterogeneous data and abstracting them into corresponding fact anchors, establishing a stable identifier for each fact anchor; Second, controlled context package compilation: retrieving associated citation lists based on the fact anchors, assembling them into a controlled context package, and freezing the citation list to limit the context scope; Third, controlled statement generation: generating report statements within the scope limited by the controlled context package according to a predetermined control strategy, with each statement bound to its corresponding citation list at the sentence level; Fourth, consistency verification and conservative filtering: verifying the consistency between the generated report statements and the content of the controlled context package, performing downgrading labeling on statements with attribution confidence below a preset threshold, marking them as candidate states and filtering them; Fifth, publication binding and auditing: binding and integrating the verified report statements to generate the final report and publishing it, while simultaneously generating a structured evidence chain to preserve the report content, thereby achieving auditing. Summary of the Invention

[0004] To address this, the present invention provides an automatic verification method for the consistency of test reports, which overcomes the shortcomings of existing technologies in the automated verification of construction project reports. These shortcomings include the lack of a verification mechanism for the authenticity of data sources uploaded by testing instruments and the inability to accurately locate verification dimensions, resulting in a lack of credibility in construction project reports, difficulty in tracing accidents, and low efficiency in correcting errors in verification results.

[0005] To achieve the above objectives, the present invention provides an automatic verification method for the consistency of test reports, comprising:

[0006] Input the elements to be verified in the target electronic report into a pre-trained indicator library and output structured indicator fields;

[0007] The structured indicator fields are compared with the original data source to obtain the comparison results;

[0008] Collect the data reliability characteristics of the comparison results;

[0009] Calculate the data credibility feature representation value based on the aforementioned data credibility feature information;

[0010] The effectiveness of the data acquisition device is determined based on the comparison between the data credibility feature value and the predetermined data credibility feature threshold.

[0011] When the validity of the data acquisition equipment meets the standard, the fact matching degree feature information of the corresponding caliber is collected;

[0012] Calculate the fact matching degree feature representation value based on the aforementioned fact matching degree feature information;

[0013] The stability of the verification device is determined based on the comparison result between the fact matching degree feature representation value and the predetermined fact matching degree feature representation threshold.

[0014] In response to the fact that the stability of the verification device does not meet the standard, the matching deviation value is calculated;

[0015] Based on the matching deviation value, the reasons for the non-compliance of the stability of the verification device with the standard and the corresponding processing strategy are determined. The processing strategy includes determining the adjustment range of the calculation weight of the fact matching degree feature characterization value and determining the standardization processing of the data collected by the data acquisition device.

[0016] The data credibility features include hash value matching rate, timestamp synchronization error rate, and blockchain evidence integrity; the fact matching features include semantic matching degree and logical constraint satisfaction degree.

[0017] Furthermore, the process of calculating the data credibility feature representation value based on the aforementioned data credibility feature information includes:

[0018] Extract the hash value matching rate, timestamp synchronization error rate, and blockchain evidence integrity of the comparison results between the structured indicator fields and the original data source;

[0019] The first credibility factor is determined by calculating the ratio of the hash value matching rate to a predetermined hash value matching rate threshold;

[0020] The second credibility factor is determined by calculating the ratio of the predetermined timestamp synchronization error rate threshold to the timestamp synchronization error rate.

[0021] The third credibility factor is determined by calculating the ratio of the evidence integrity to a predetermined evidence integrity threshold.

[0022] The data credibility feature value is determined by summing the first credibility factor, the second credibility factor, and the third credibility factor according to a predetermined weight ratio.

[0023] Furthermore, the process of determining whether the effectiveness of the data acquisition device meets the standard based on the comparison result between the data credibility feature representation value and the predetermined data credibility feature representation threshold includes:

[0024] If the data credibility feature value is less than or equal to the predetermined data credibility feature threshold, the validity of the data acquisition device is determined to be non-compliant with the standard.

[0025] If the data credibility feature value is greater than the predetermined data credibility feature threshold, then the validity of the data acquisition device is determined to meet the standard.

[0026] Furthermore, the process of calculating the fact matching degree feature representation value based on the aforementioned fact matching degree feature information includes:

[0027] Extract the semantic matching degree and the satisfaction degree of logical constraints at the corresponding caliber end;

[0028] The ratio of the semantic matching degree to the predetermined semantic matching degree threshold is determined as the first matching degree factor;

[0029] The ratio of the satisfaction level to the predetermined satisfaction threshold is used as the second matching factor.

[0030] The sum of the first matching factor and the second matching factor is determined as the factual matching feature value.

[0031] Furthermore, the process of determining whether the stability of the verification device meets the standard based on the comparison result between the fact matching degree feature representation value and the predetermined fact matching degree feature representation threshold includes:

[0032] If the fact matching feature value is less than or equal to the predetermined fact matching feature threshold, then the stability of the verification device is determined to be non-compliant with the standard.

[0033] If the fact matching feature value is greater than the predetermined fact matching feature threshold, then the stability of the verification device is determined to meet the standard.

[0034] Furthermore, the process of calculating the matching deviation value includes:

[0035] Extract the fact matching feature values ​​and the predetermined fact matching feature thresholds;

[0036] Calculate the difference between the predetermined fact-matching feature representation threshold and the fact-matching feature representation value;

[0037] The difference is determined as the matching deviation value.

[0038] Furthermore, the process of determining the reasons why the stability of the verification device does not meet the standard based on the matching deviation value includes:

[0039] If the matching deviation value is less than or equal to the preset matching deviation threshold, it is determined that there is a deviation in the margin calculation.

[0040] If the matching deviation value is greater than the preset matching deviation threshold, it is determined that the caliber end is inconsistent.

[0041] Furthermore, the process of matching the corresponding processing strategy based on the stated cause includes:

[0042] If the cause is a deviation in the margin calculation, then determine the adjustment range of the calculation weight of the fact matching degree feature representation value;

[0043] If the cause is inconsistency in the caliber, then the data collected by the data acquisition equipment should be standardized.

[0044] Furthermore, the adjustment range of the calculation weight for determining the factual matching degree feature representation value based on the matching deviation value is wherein the adjustment range of the calculation weight is related to the matching deviation value.

[0045] Furthermore, the standardization process for the data collected by the data acquisition device is determined based on the matching deviation value, wherein the standardized data is related to the matching deviation value.

[0046] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention provides an automatic verification method for the consistency of inspection reports. The method involves inputting the elements to be verified in the target electronic report into a pre-trained indicator library, outputting structured indicator fields, comparing these structured indicator fields with the original data source to obtain comparison results, collecting data credibility feature information from the comparison results, calculating data credibility feature representation values ​​based on these data credibility feature information, and comparing these data credibility feature representation values ​​with a predetermined data credibility feature representation threshold to determine whether the validity of the data acquisition device meets the standard. Only when the validity of the data acquisition device meets the standard is the corresponding factual matching degree feature information further collected. Based on this factual matching degree feature information, a factual matching degree feature representation value is calculated. Based on the comparison result of the factual matching degree feature representation value and the predetermined factual matching feature representation threshold, the stability of the verification device is determined to meet the standard. If the stability of the verification device does not meet the standard, a matching deviation value is calculated, and based on the matching deviation value, the reason for the non-compliance of the verification device's stability and the corresponding processing strategy are determined. This invention overcomes the shortcomings of existing technologies in the automated verification of construction project reports. These shortcomings include the lack of a verification mechanism to verify the authenticity of the data source uploaded by the testing instruments, the single verification dimension, and the inability to accurately locate the cause of errors. Consequently, the verification results lack interpretability, and the error correction efficiency is low. This leads to a lack of credibility in the quality of construction project reports, difficulty in tracing safety accidents, and the accumulation of regulatory compliance risks.

[0047] In particular, this invention inputs the elements to be verified in the target electronic report into a pre-trained indicator library, outputs structured indicator fields, and simultaneously collects the hash value matching rate, timestamp synchronization error rate, and blockchain evidence integrity of the comparison results between the structured indicator fields and the original data source to obtain data credibility feature information. The data credibility feature information is processed into a first credibility factor, a second credibility factor, and a third credibility factor, and summed according to a predetermined weight ratio to determine the data credibility feature representation value. This can convert multi-dimensional data into dimensionless data credibility feature representation values, thereby improving the accuracy of automatic verification of the original information collection.

[0048] In particular, this invention obtains factual matching degree feature information by collecting the semantic matching degree and the satisfaction degree of logical constraints at the corresponding caliber end, and decomposes the factual matching degree feature information into a first matching degree factor and a second matching degree factor, and sums them to determine the factual matching degree feature representation value. This can standardize the parsing process of feature representation value, unify the calculation standard, and provide reliable data support for subsequent judgment steps, thereby improving the stability of the verification device and the accuracy of judgment.

[0049] In particular, this invention first compares the data credibility feature representation value with a predetermined data credibility feature representation threshold to determine whether the validity of the data acquisition device meets the standard. If the data credibility feature representation value is greater than the predetermined data credibility feature representation threshold, the validity of the data acquisition device is determined to meet the standard. Then, based on the comparison result of the fact matching feature representation value and the predetermined fact matching feature representation threshold, the stability of the verification device is determined to meet the standard. This allows for layered verification of the authenticity of the data source first, and then verification of the consistency of the report conclusions, thereby improving the diagnostic efficiency of report consistency verification for abnormal situations.

[0050] In particular, the present invention calculates the difference between a predetermined fact matching feature representation threshold and a fact matching feature representation value when the stability of the verification device does not meet the standard, using this difference as a matching deviation value. Based on the comparison result between the matching deviation value and the preset matching deviation threshold, the cause of the verification device's instability not meeting the standard is located, and a differentiated processing strategy is matched. This can accurately locate the specific cause of the report consistency anomaly and provide a targeted processing method, thereby improving the pertinence of the report consistency verification anomaly repair. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating the steps of the automatic consistency verification method for test reports according to an embodiment of the present invention.

[0052] Figure 2 This is a logic diagram for determining whether the validity of a data acquisition device conforms to a standard based on the data reliability feature characterization value in an embodiment of the present invention.

[0053] Figure 3 This is a logic diagram illustrating whether the stability of the verification device based on the fact matching degree feature characterization value conforms to the standard in an embodiment of the present invention.

[0054] Figure 4 This is a logic diagram illustrating how the stability of the verification device fails to meet the standard based on the matching deviation value, and the corresponding processing strategy, according to an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0056] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0057] Please see Figure 1The diagram shows the steps of an automatic verification method for the consistency of test reports according to an embodiment of the present invention. The present invention provides an automatic verification method for the consistency of test reports, comprising:

[0058] Step S1: Input the elements to be verified in the target electronic report into the pre-trained indicator library and output the structured indicator fields;

[0059] Step S2: Compare the structured indicator fields with the original data source to obtain the comparison results;

[0060] Step S3: Collect the data reliability feature information of the comparison results;

[0061] Step S4: Calculate the data credibility feature representation value based on the data credibility feature information;

[0062] Step S5: Determine whether the effectiveness of the data acquisition device meets the standard based on the comparison result between the data credibility feature characterization value and the predetermined data credibility feature characterization threshold.

[0063] Step S6: In response to the data acquisition device meeting the validity standard, collect the fact matching degree feature information of the corresponding caliber end;

[0064] Step S7: Calculate the fact matching degree feature representation value based on the fact matching degree feature information;

[0065] Step S8: Determine whether the stability of the verification device meets the standard based on the comparison result between the fact matching degree feature representation value and the predetermined fact matching degree feature representation threshold.

[0066] Step S9: Calculate the matching deviation value in response to the fact that the stability of the verification device does not meet the standard;

[0067] Step S10: Based on the matching deviation value, determine the reasons why the stability of the verification device does not meet the standard and the corresponding processing strategy. The processing strategy includes determining the adjustment range of the calculation weight of the fact matching degree feature characterization value and determining the standardization processing of the data collected by the data acquisition device.

[0068] The data credibility features include hash value matching rate, timestamp synchronization error rate, and blockchain evidence integrity; the fact matching features include semantic matching degree and logical constraint satisfaction degree.

[0069] In this embodiment, each time steps S1 to S9 are executed, the single verification process is automatically recorded, generating an auditable traceability link, providing complete and tamper-proof evidence support for subsequent quality review and compliance audit.

[0070] It is understood that the target electronic report, referring to a safety assessment report in fields such as construction engineering and special equipment, is compiled based on on-site collected data, standard limits, and judgment rules. The elements to be verified in the target electronic report are the key data items to be verified, including at least: test values ​​such as wall thickness, load, and vibration amplitude; standard limit references such as specific clauses of national and industry standards; conclusion judgments such as qualified, unqualified, or requiring re-inspection; and metadata such as testing time, equipment number, and personnel signature. Specific elements to be verified can be set by those skilled in the art based on the actual testing scenario.

[0071] In this embodiment, the pre-trained indicator library is a mapping model trained through machine learning. It takes natural language text from the target electronic report as input and outputs structured indicator fields in a unified format. The construction process of the pre-trained indicator library includes:

[0072] S11. Collect 1,000 audited test reports in the fields of building engineering and special equipment, and have experts in the field annotate the key test parameters, limit references and conclusion judgment items in each report to form an annotated dataset;

[0073] S12. A bidirectional encoder representation model is used as the basic algorithm. The model is trained by taking the original text fragments of the report in the labeled dataset as input and the corresponding structured labels as output until the labeling accuracy of the model on the validation set reaches more than 95%.

[0074] S13. The trained model is packaged into an indicator library and deployed in the verification device to automatically extract and output structured indicator fields from the target electronic report.

[0075] It is understood that the output structured index field refers to converting the text description in the target electronic report into a unified comparison format. For example, converting the wall thickness of 12.5mm into a parameter named "wall thickness," with a value of 12.5 and a unit of mm. The structuring process does not change the original core semantic content such as values, units, time, and conclusions; it only adjusts the expression form. Therefore, the hash values ​​can be directly compared.

[0076] Understandably, the comparison result refers to the set of matching scores obtained by performing hash calculations, timestamp comparisons, and blockchain verification on each field of the structured indicator fields and the original data source fields.

[0077] As can be understood, the hash value matching rate refers to the proportion of overlap between the structured indicator fields of the elements to be verified in the target electronic report and the corresponding fields in the original data source, collected through a hash calculation unit. The calculation formula is as follows: .

[0078] Understandably, the timestamp synchronization error rate refers to the percentage difference between the generation timestamp of the element to be verified in the target electronic report and the generation timestamp of the corresponding element in the original data source, collected through the time synchronization unit. Calculation formula: The maximum allowable time deviation is set at 100 milliseconds. The lower the timestamp synchronization error rate, the higher the synchronization between the generation time of the elements to be verified in the target electronic report and the original data source, and the better the compliance of the data generation sequence.

[0079] Understandably, the completeness of evidence storage in blockchain refers to the percentage of complete overlap between the evidence storage data of the target electronic report's elements to be verified in the blockchain node and the locally stored evidence data, collected through the blockchain evidence verification unit. Calculation formula: The higher the integrity of the blockchain-stored evidence information of the elements to be verified in the target electronic report, the better, indicating that there are no missing or tampered elements in the evidence storage chain.

[0080] As can be understood, semantic matching degree refers to the degree of overlap between the textual semantic content of the target structured indicator field and the corresponding elements in the original data source, collected through a natural language processing semantic comparison unit. Calculation formula: The greater the semantic matching degree, the higher the consistency between the target structured indicator field and the text semantic content of the original data source, and the better the matching degree of the content meaning.

[0081] It is understandable that the satisfaction degree of logical constraints refers to the degree to which the target structured indicator fields conform to preset business logic rules, which is collected through the logical rule verification unit. Calculation formula: The preset business logic rules transform the national and industry standard rules requirements in building engineering and special equipment inspection reports into computer-readable fields. The greater the satisfaction of logical constraints, the higher the degree of conformity of the target structured indicator fields to the preset business logic rules, and the stronger the compliance of the data logic. The national and industry standard rules specifically include, but are not limited to: GB / T50189 "Evaluation Standard for Green Buildings", GB50300 "Unified Standard for Acceptance of Construction Quality of Building Engineering", GB / T30559 "Energy Performance of Elevators, Escalators and Moving Walks", and TSGT7001 "Rules for Supervision and Periodic Inspection of Elevators". Corresponding fields are extracted from the normative appendices and main text clauses.

[0082] This invention improves the accuracy of report consistency verification results by inputting the elements to be verified in the target electronic report into a pre-trained indicator library, outputting structured indicator fields, collecting data credibility feature information from the comparison results of the structured indicator fields and the original data source, calculating data credibility feature representation values ​​based on the data credibility feature information, and comparing the data credibility feature representation values ​​with predetermined data credibility feature representation thresholds. Only when the data acquisition device's validity meets the standard is the corresponding factual matching feature information further collected. Based on the factual matching feature information, a factual matching feature representation value is calculated. Based on the comparison result of the factual matching feature representation value with the predetermined factual matching feature representation threshold, the stability of the verification device is determined. If the stability of the verification device does not meet the standard, a matching deviation value is calculated. Based on the matching deviation value, the reasons for the non-compliance of the verification device's stability and corresponding processing strategies are determined, thereby improving the accuracy of report consistency verification results.

[0083] Specifically, the process of calculating the data credibility feature representation value based on the data credibility feature information includes:

[0084] Extract the hash value matching rate, timestamp synchronization error rate, and blockchain evidence integrity of the comparison results between the structured indicator fields and the original data source;

[0085] The first credibility factor is determined by calculating the ratio of the hash value matching rate to a predetermined hash value matching rate threshold;

[0086] The second credibility factor is determined by calculating the ratio of the predetermined timestamp synchronization error rate threshold to the timestamp synchronization error rate.

[0087] The third credibility factor is determined by calculating the ratio of the evidence integrity to a predetermined evidence integrity threshold.

[0088] The data credibility feature value is determined by summing the first credibility factor, the second credibility factor, and the third credibility factor according to a predetermined weight ratio.

[0089] In this embodiment, the single acquisition period is preset, and the preferred single acquisition period is 100 milliseconds.

[0090] In this embodiment, the predetermined hash value matching rate threshold is preset. Specifically, hash value matching rate samples within three historical collection periods are predetermined, and the predetermined hash value matching rate threshold is determined based on the average value of the hash value matching rate samples. The threshold is determined within the range of [93%, 99%]. In this embodiment, the predetermined hash value matching rate threshold is preferably 97%.

[0091] In this embodiment, the predetermined timestamp synchronization error rate threshold is preset. Specifically, timestamp synchronization error rate samples within three historical acquisition periods are predetermined, and the predetermined timestamp synchronization error rate threshold is determined based on the average value of the timestamp synchronization error rate samples. The threshold is determined within the range of [8%, 12%]. In this embodiment, the predetermined timestamp synchronization error rate threshold is preferably 11%.

[0092] In this embodiment, the predetermined evidence integrity threshold is preset, wherein evidence integrity samples within three historical collection periods are predetermined, and the predetermined evidence integrity threshold is determined based on the average value of the evidence integrity samples, within the range [90%, 96%]. In this embodiment, the predetermined evidence integrity threshold is preferably 92%.

[0093] In this embodiment, the predetermined weight ratio is set in advance and selected based on the contribution of hash value matching rate, timestamp synchronization error rate and blockchain evidence integrity to the data acquisition device validity judgment standard. The preferred predetermined weight ratio in this embodiment is 6:7:7, that is, the data credibility feature characterization value is equal to the sum of 0.3 times the first credibility factor, 0.35 times the second credibility factor and 0.35 times the third credibility factor.

[0094] This invention extracts hash value matching rate, timestamp consistency, and blockchain evidence integrity, and calculates the corresponding first credibility factor, second credibility factor, and third credibility factor respectively. Then, it sums them according to a predetermined weight ratio to obtain the data credibility feature characterization value. This can uniformly convert multi-dimensional data credibility feature information into quantifiable numerical values, thereby improving the accuracy of data sources and the accuracy of calculation results in the report consistency verification process.

[0095] Please see Figure 2 As shown, this is a logic diagram for determining whether the validity of a data acquisition device meets the standard based on the data reliability feature characterization value in an embodiment of the present invention. The process of determining whether the validity of a data acquisition device meets the standard based on the comparison result of the data reliability feature characterization value and the predetermined data reliability feature characterization threshold includes:

[0096] If the data credibility feature value is less than or equal to the predetermined data credibility feature threshold, the validity of the data acquisition device is determined to be non-compliant with the standard.

[0097] If the data credibility feature value is greater than the predetermined data credibility feature threshold, then the validity of the data acquisition device is determined to meet the standard.

[0098] In this embodiment, the predetermined data credibility feature representation threshold is preset. Specifically, the average value of the data credibility feature representation values ​​over six historical collection periods is predetermined. The predetermined data credibility feature representation threshold is determined based on the average value of the data credibility feature representation values ​​and is determined within the range [0.95, 1.15]. In this embodiment, the predetermined data credibility feature representation threshold is preferably 1.04.

[0099] Understandably, the validity of data acquisition equipment meets the standard if the average hash value matching rate is greater than 97% over six consecutive acquisition cycles, the average timestamp synchronization error rate is less than or equal to 11%, and the average evidence integrity is greater than 92%, thus meeting the criteria for determining the authenticity of the data source.

[0100] This invention compares the real-time calculated data reliability feature value with a predetermined data reliability feature threshold set based on historical collection cycles. This allows for direct differentiation of the validity status of data collection devices. When the device validity does not meet the standard, a parameter calibration process is triggered, which filters reliable data sources for subsequent verification stages. This improves the accuracy of the data collected during the report consistency verification process.

[0101] Specifically, the process of calculating the fact matching degree feature representation value based on the fact matching degree feature information includes:

[0102] Extract the semantic matching degree and the satisfaction degree of logical constraints at the corresponding caliber end;

[0103] The ratio of the semantic matching degree to the predetermined semantic matching degree threshold is determined as the first matching degree factor;

[0104] The ratio of the satisfaction level to the predetermined satisfaction threshold is used as the second matching factor.

[0105] The sum of the first matching factor and the second matching factor is determined as the factual matching feature value.

[0106] It is understandable that the scope includes a report generation unit, a hash calculation unit, a time synchronization unit, a blockchain evidence verification unit, a semantic comparison unit, and a logical rule verification unit.

[0107] In this embodiment, the predetermined semantic matching threshold is preset, wherein semantic matching samples within three historical collection periods are predetermined, and the predetermined semantic matching threshold is determined based on the average value of the semantic matching samples, within the range [90%, 96%]. In this embodiment, the predetermined semantic matching threshold is preferably 93%.

[0108] In this embodiment, the predetermined satisfaction threshold is preset, wherein satisfaction samples within three historical collection periods are predetermined, and the predetermined satisfaction threshold is determined based on the average value of the satisfaction samples, within the range [87%, 94%]. In this embodiment, the predetermined satisfaction threshold is preferably 89%.

[0109] This invention extracts the semantic matching degree and the satisfaction degree of logical constraints at the corresponding caliber end, calculates the first matching degree factor and the second matching degree factor respectively, and then sums them to obtain the fact matching degree feature representation value. This can unify the feature parameters of the two dimensions of text semantics and business logic into quantifiable values, thereby improving the accuracy of the fact matching degree evaluation of the report consistency verification.

[0110] Please see Figure 3 As shown, this is a logic diagram for determining whether the stability of the verification device meets the standard based on the fact matching degree feature representation value in an embodiment of the present invention. The process of determining whether the stability of the verification device meets the standard based on the comparison result of the fact matching degree feature representation value and the predetermined fact matching degree feature representation threshold includes:

[0111] If the fact matching feature value is less than or equal to the predetermined fact matching feature threshold, then the stability of the verification device is determined to be non-compliant with the standard.

[0112] If the fact matching feature value is greater than the predetermined fact matching feature threshold, then the stability of the verification device is determined to meet the standard, and the various elements to be verified in the target electronic report are consistent with the corresponding elements in the original data source in terms of content, semantics, and logic.

[0113] Understandably, the verification device is an automated data processing system deployed on a testing cloud platform, comprising: a hash calculation unit, a time synchronization unit, a blockchain evidence verification unit, a semantic comparison unit, a logical rule verification unit, a weight adjustment unit, and a data standardization unit. Specifically, the hash calculation unit calculates the hash values ​​of structured indicator fields and original data source fields, and outputs the hash value matching rate; the time synchronization unit compares the timestamps of the target electronic report with those of the original data source, and outputs the timestamp synchronization error rate; the blockchain evidence verification unit verifies the consistency between the evidence data stored in the blockchain and the locally stored evidence data, and outputs the evidence integrity; the semantic comparison unit calculates the semantic similarity between structured indicator fields and the original data source using a natural language processing model, and outputs the semantic matching degree; the logical rule verification unit verifies whether the structured indicator fields meet preset business logic constraints, and outputs the logical constraint satisfaction degree; the weight adjustment unit dynamically adjusts the calculation weights of the factual matching degree feature representation values ​​based on the matching deviation value; and the data standardization unit performs unified conversion of the format, units, and definitions of the raw data uploaded by the data acquisition equipment.

[0114] In this embodiment, the predetermined fact matching feature representation threshold is preset. Specifically, the average value of the fact matching feature representation values ​​over six historical collection periods is predetermined. The predetermined fact matching feature representation threshold is determined based on the average value of the fact matching feature representation values ​​and is determined within the range [1.95, 2.09]. In this embodiment, the predetermined fact matching feature representation threshold is preferably 2.03.

[0115] This invention improves the accuracy of report consistency verification results by comparing the real-time obtained fact matching degree feature representation value with the fact matching degree feature representation threshold determined based on historical multi-collection cycles.

[0116] Specifically, the process of calculating the matching deviation value includes:

[0117] Extract the fact matching feature values ​​and the predetermined fact matching feature thresholds;

[0118] Calculate the difference between the predetermined fact-matching feature representation threshold and the fact-matching feature representation value;

[0119] The difference is determined as the matching deviation value.

[0120] Understandably, the matching deviation value is used to intuitively reflect the numerical difference between the actual matching degree characteristic value and the predetermined threshold. It is directly used to locate the specific type of problem where the stability of the verification device fails to meet the standard, and to provide a quantitative basis for the subsequent determination of the cause of the anomaly.

[0121] This invention extracts fact matching feature values ​​and corresponding predetermined fact matching feature thresholds, and calculates the difference to obtain a matching deviation value. This can transform the non-standard stability of the verification device into a quantifiable deviation value, thereby improving the accuracy of locating abnormal problems and the reliability of the judgment basis when verifying report consistency.

[0122] Please see Figure 4 As shown, this is the logical determination of the reasons why the stability of the verification device does not meet the standard based on the matching deviation value and the corresponding processing strategy in an embodiment of the present invention. The process of determining the reasons why the stability of the verification device does not meet the standard based on the matching deviation value includes:

[0123] If the matching deviation value is less than or equal to the preset matching deviation threshold, it is determined that there is a deviation in the margin calculation.

[0124] If the matching deviation value is greater than the preset matching deviation threshold, it is determined that the caliber end is inconsistent.

[0125] In this embodiment, the predetermined matching deviation threshold is preset, wherein the average value of the matching deviation values ​​in the past 8 collection periods is predetermined, and the predetermined matching deviation threshold is determined based on the average value of the matching deviation values, which is determined within the range [0.01, 0.08]. In this embodiment, the predetermined matching deviation threshold is preferably 0.05.

[0126] It is understandable that the deviation in margin calculation indicates a discrepancy between the internal calculation parameters of the verification device and the preset business rules, and does not involve any abnormalities in the data acquisition and data transmission process.

[0127] Understandably, inconsistencies in the standards indicate differences in data transmission standards between the data acquisition unit, report generation unit, data storage unit, and verification unit, affecting the consistency of the electronic report generation results.

[0128] This invention compares the matching deviation value with a matching deviation threshold set according to the historical acquisition cycle. Based on the magnitude of the matching deviation value, it can distinguish the abnormal type corresponding to the failure of the core device to meet the stability standard. The causes of the abnormality are divided into two categories: deviation in margin calculation and inconsistency at the caliber end. The scope of the problem and the generation link corresponding to different abnormalities are clarified, thereby improving the accuracy of the location of the cause of the abnormality in the report consistency verification.

[0129] Specifically, the process of matching the corresponding processing strategy based on the stated reason includes:

[0130] If the cause is a deviation in the margin calculation, then determine the adjustment range of the calculation weight of the fact matching degree feature representation value;

[0131] If the cause is inconsistency in the caliber, then the data collected by the data acquisition equipment should be standardized.

[0132] It is understandable that by determining the adjustment range of the calculation weight of the fact matching degree feature representation value, the internal judgment logic of the verification device can be adaptively optimized, so that the fact matching degree feature representation value is more in line with the current data characteristics, thereby improving the credibility of the verification results.

[0133] It is understandable that by standardizing the data collected by the data acquisition device, the format and semantics of different data sources can be unified, reducing consistency deviations caused by differences in caliber, thereby improving the compatibility and processing accuracy of the verification device for multi-source heterogeneous data.

[0134] The embodiments of the present invention, through the synergistic effect of processing strategies, can specifically address the stability issues of the verification device from two dimensions: internal calculation weight adjustment and external data standardization. This can correct deviations at the calculation logic level and standardize differences at the data input level, thereby forming a closed-loop anomaly handling mechanism, improving the accuracy of report consistency verification and the efficiency of anomaly repair.

[0135] Specifically, the adjustment range of the calculation weight for determining the factual matching degree feature representation value based on the matching deviation value is wherein the adjustment range of the calculation weight is related to the matching deviation value.

[0136] In this embodiment, the formula for calculating the weight of the fact matching degree feature representation value is as follows: ;in, The adjusted weights of the first matching factor. The adjusted weights of the second matching factor. The weight of the first matching factor before adjustment is 0.5, and the preferred value in this embodiment is 0.5. The matching deviation value is selected within the range [0.01, 0.05], with a preferred value of 0.03. The correction coefficient is selected based on the mapping relationship required for actual verification, and the preferred value in the example is 25.

[0137] By introducing a weight adjustment mechanism related to the matching deviation value, this embodiment of the invention can dynamically optimize the calculation weight of the fact matching degree feature representation value when a deviation in the margin calculation is detected, enabling the verification device to have adaptive adjustment capability, thereby improving the response sensitivity and judgment accuracy of the verification logic to actual data fluctuation scenarios.

[0138] Specifically, the standardization process for the data collected by the data acquisition device is determined based on the matching deviation value, wherein the standardized data is related to the matching deviation value.

[0139] In this embodiment, the formula for calculating the standardized data is: ;in, This is the correction amount for the hash value matching rate. The hash value matching rate is a baseline correction value, preferably 1% in this embodiment. The value representing the matching deviation is selected within the range [0.05, 0.08] in this embodiment, with a preferred value of 0.07. The calibration coefficient is selected based on the standardization target requirements of the caliber end data, and the preferred value in the example is 0.2.

[0140] This invention, by dynamically adjusting the data collected at the caliber end based on the matching deviation value, can implement differentiated data correction for the degree of caliber inconsistency, enabling the verification device to achieve unified standardization of multi-source data in the data input stage, thereby reducing the risk of verification misjudgment caused by caliber differences and enhancing the robustness and interpretability of the overall verification process.

[0141] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. An automatic verification method for the consistency of test reports, characterized in that, include: Input the elements to be verified in the target electronic report into a pre-trained indicator library and output structured indicator fields; The structured indicator fields are compared with the original data source to obtain the comparison results; Collect the data reliability characteristics of the comparison results; Calculate the data credibility feature representation value based on the aforementioned data credibility feature information; The effectiveness of the data acquisition device is determined based on the comparison between the data credibility feature value and the predetermined data credibility feature threshold. When the validity of the data acquisition equipment meets the standard, the fact matching degree feature information of the corresponding caliber is collected; Calculate the fact matching degree feature representation value based on the aforementioned fact matching degree feature information; The stability of the verification device is determined based on the comparison result between the fact matching degree feature representation value and the predetermined fact matching degree feature representation threshold. In response to the fact that the stability of the verification device does not meet the standard, the matching deviation value is calculated; Based on the matching deviation value, the reasons for the non-compliance of the stability of the verification device with the standard and the corresponding processing strategy are determined. The processing strategy includes determining the adjustment range of the calculation weight of the fact matching degree feature characterization value and determining the standardization processing of the data collected by the data acquisition device. The data credibility features include hash value matching rate, timestamp synchronization error rate, and blockchain evidence integrity; the fact matching features include semantic matching degree and logical constraint satisfaction degree.

2. The automatic verification method for consistency of test reports according to claim 1, characterized in that, The process of calculating the data credibility feature representation value based on the aforementioned data credibility feature information includes: Extract the hash value matching rate, timestamp synchronization error rate, and blockchain evidence integrity of the comparison results between the structured indicator fields and the original data source; The first credibility factor is determined by calculating the ratio of the hash value matching rate to a predetermined hash value matching rate threshold; The second credibility factor is determined by calculating the ratio of the predetermined timestamp synchronization error rate threshold to the timestamp synchronization error rate. The third credibility factor is determined by calculating the ratio of the evidence integrity to a predetermined evidence integrity threshold. The data credibility feature value is determined by summing the first credibility factor, the second credibility factor, and the third credibility factor according to a predetermined weight ratio.

3. The automatic verification method for consistency of test reports according to claim 1, characterized in that, The process of determining whether the effectiveness of the data acquisition device meets the standard based on the comparison result of the data reliability feature characterization value and the predetermined data reliability feature characterization threshold includes: If the data credibility feature value is less than or equal to the predetermined data credibility feature threshold, the validity of the data acquisition device is determined to be non-compliant with the standard. If the data credibility feature value is greater than the predetermined data credibility feature threshold, then the validity of the data acquisition device is determined to meet the standard.

4. The automatic verification method for consistency of test reports according to claim 1, characterized in that, The process of calculating the fact matching degree feature representation value based on the aforementioned fact matching degree feature information includes: Extract the semantic matching degree and the satisfaction degree of logical constraints at the corresponding caliber end; The ratio of the semantic matching degree to the predetermined semantic matching degree threshold is determined as the first matching degree factor; The ratio of the satisfaction level to the predetermined satisfaction threshold is used as the second matching factor. The sum of the first matching factor and the second matching factor is determined as the factual matching feature value.

5. The automatic verification method for consistency of test reports according to claim 1, characterized in that, The process of determining whether the stability of the verification device meets the standard based on the comparison result between the fact matching feature value and the predetermined fact matching feature threshold includes: If the fact matching feature value is less than or equal to the predetermined fact matching feature threshold, then the stability of the verification device is determined to be non-compliant with the standard. If the fact matching feature value is greater than the predetermined fact matching feature threshold, then the stability of the verification device is determined to meet the standard.

6. The automatic verification method for consistency of test reports according to claim 1, characterized in that, The process of calculating the matching deviation value includes: Extract the fact matching feature values ​​and the predetermined fact matching feature thresholds; Calculate the difference between the predetermined fact-matching feature representation threshold and the fact-matching feature representation value; The difference is determined as the matching deviation value.

7. The automatic verification method for consistency of test reports according to claim 1, characterized in that, The process of determining the reasons why the stability of the verification device does not meet the standard based on the matching deviation value includes: If the matching deviation value is less than or equal to the preset matching deviation threshold, it is determined that there is a deviation in the margin calculation. If the matching deviation value is greater than the preset matching deviation threshold, it is determined that the caliber end is inconsistent.

8. The automatic verification method for consistency of test reports according to claim 7, characterized in that, The process of matching the corresponding processing strategy based on the aforementioned reasons includes: If the cause is a deviation in the margin calculation, then determine the adjustment range of the calculation weight of the fact matching degree feature representation value; If the cause is inconsistency in the caliber, then the data collected by the data acquisition equipment should be standardized.

9. The automatic verification method for consistency of test reports according to claim 8, characterized in that, The adjustment range of the calculation weight for determining the factual matching degree feature representation value based on the matching deviation value, wherein the adjustment range of the calculation weight is related to the matching deviation value.

10. The automatic verification method for consistency of test reports according to claim 8, characterized in that, The standardization process for the data collected by the data acquisition device is determined based on the matching deviation value, wherein the standardized data is related to the matching deviation value.

Images