Text detection method and device, equipment, storage medium and product

By building and training a quality inspection model based on performance testing, and automatically processing the content of the test files, the problems of low efficiency and insufficient accuracy in the quality inspection work of testers and managers are solved, and efficient and accurate file quality inspection is achieved.

CN119088708APending Publication Date: 2024-12-06CHINA MERCHANTS BANK
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411221837.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-02
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

When there are limited testers and managers, the analysis and quality inspection of test results rely on manual labor, which is prone to problems such as content errors, low efficiency, and omissions of problems, resulting in poor quality of test results.

Method used

A text detection method is adopted to build sample data sets by analyzing data based on performance test results and preset quality inspection rules sets, train quality inspection models, realize automated file content quality inspection, and combine it with the software platform to provide convenient quality inspection capabilities to improve Quality inspection efficiency and quality.

Benefits of technology

Through automated quality inspection models, human errors are reduced, the quality and efficiency of the test process documents are improved, real-time data marking and model enhancement are supported, the accuracy of the quality inspection models is continuously improved, and the efficiency and accuracy caused by artificial dependence are solved. Restricted issues.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119088708A_ABST
    Figure CN119088708A_ABST
Patent Text Reader

Abstract

The invention discloses a text detection method and device, equipment, a storage medium and a product. The method comprises the following steps: in response to a document quality inspection instruction, obtaining data to be subjected to quality inspection based on a performance test process document content; the data to be subjected to quality inspection are input into a pre-trained quality inspection model for quality inspection analysis, a quality inspection result is obtained, and a sample data set used for quality inspection model training is constructed based on performance test result analysis data and in combination with a preset quality inspection rule set. According to the method, the logic correctness evaluation of the text content in the specific field can be realized, the quality inspection model is integrated with a software platform, the convenient and automatic file content quality inspection capability is provided for testers and test managers, and the process file quality inspection quality and efficiency are improved. And through marking data of a real-time quality inspection result, enhancement training is continuously performed on the model, a sustainable closed loop from sample input training to model identification to new sample generation is formed, and the accuracy of the detection model is continuously improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of testing technology, and in particular to a text detection method, device, equipment, storage medium and product. Background Art

[0002] As various businesses are connected to cloud services, the scale of business systems has expanded dramatically, and the performance requirements for business systems have also increased dramatically. As a result, the demand for performance testing of a large number of business systems has increased, and a large number of testing tasks need to be completed in a relatively short period of time. Among them, the demand for quality assurance of process documents produced by testing has also increased accordingly.

[0003] For testers, in scenarios where testing human resources are limited, it is common for a single person to handle multiple testing tasks in parallel. At present, the analysis of test results is highly dependent on manual work, and is affected by multiple factors such as the tester's skill level, the amount of parallel tasks, and the difficulty of the test tasks. When analyzing test results, it is very likely that there will be problems such as content errors and inaccurate content analysis due to typos, human negligence, and insufficient knowledge reserves.

[0004] In addition, for test managers, the amount of testing tasks is large, and the quality management workload of test managers also increases accordingly. The current process document quality inspection work also relies on manual verification, which has problems such as heavy workload, low work efficiency, and risk of missing problems. Summary of the invention

[0005] The main purpose of this application is to provide a text detection method, device, equipment, storage medium and product, aiming to improve the efficiency and quality of document quality inspection during the test process.

[0006] To achieve the above purpose, the present application proposes a text detection method, which is applied to a software platform and includes:

[0007] In response to the document quality inspection instruction, obtaining the data to be quality inspected based on the content of the performance test process document;

[0008] The data to be quality inspected is input into a pre-trained quality inspection model for quality inspection analysis to obtain quality inspection results. The quality inspection model is trained based on a pre-constructed sample data set, and the sample data set is constructed based on performance test result analysis data combined with a preset quality inspection rule set.

[0009] In one embodiment, the step of inputting the data to be quality inspected into a pre-trained quality inspection model for quality inspection analysis to obtain the quality inspection result further includes:

[0010] Build a dedicated QA rule set within a preset domain for performance testing;

[0011] Analyze data based on known performance test results and construct a sample data set in combination with the quality inspection rule set;

[0012] A quality inspection model is trained based on the sample data set to obtain a trained quality inspection model.

[0013] In one embodiment, the step of constructing a dedicated quality inspection rule set within a preset domain for performance testing includes:

[0014] Obtain quality inspection factors for performance testing compiled by experts in the preset field;

[0015] A quality inspection rule set is constructed according to the quality inspection factors.

[0016] In one embodiment, the step of analyzing data based on known performance test results and constructing a sample data set in combination with the quality inspection rule set includes:

[0017] Collect and organize the scenario test result analysis data in the known performance test result analysis report;

[0018] Analyze data based on the scenario test results and construct an initial sample data set in combination with the quality inspection rule set;

[0019] The adaptive boosting algorithm Adaboost is used in combination with the initial sample data set to perform data labeling, and a calibrated positive and negative sample data set is obtained as the final sample data set for training the quality inspection model.

[0020] In one embodiment, the step of using the adaptive boosting algorithm Adaboost and combining the initial sample data set to perform data labeling to obtain a calibrated positive and negative sample data set as the final sample data set for training the quality inspection model includes:

[0021] Extracting a training sample set and a sample set to be calibrated from the initial sample data set;

[0022] Perform Adaboost model training according to the training sample set to obtain a calibration model;

[0023] The calibration model is used to classify and calibrate the sample set to be calibrated to obtain a calibrated positive and negative sample data set as the final sample data set for training the quality inspection model.

[0024] In one embodiment, the step of training the quality inspection model based on the sample data set to obtain the trained quality inspection model includes:

[0025] A positive sample corpus is constructed based on the positive sample data set, and a negative sample corpus is constructed based on the negative sample data set;

[0026] Based on the positive sample corpus and the negative sample corpus, word vector learning is performed using the word vector model GloVe based on word co-occurrence matrix frequency statistics, and feature vectors are obtained by feature extraction;

[0027] Based on the feature vector, a Laplace kernel function based on Euclidean distance is used to perform model training to obtain a trained quality inspection model.

[0028] In one embodiment, the method further comprises:

[0029] Embedding the trained quality inspection model into the software platform;

[0030] Obtain the manual marking results of the quality inspection results provided by the software platform, integrate the real-time data into new calibration samples according to the manual marking results, and perform enhanced training on the quality inspection model.

[0031] In one embodiment, the step of obtaining the manual marking results of the quality inspection results provided by the software platform, integrating the real-time data into a new calibration sample according to the manual marking results, and performing enhanced training on the quality inspection model includes:

[0032] The quality inspection results are displayed online, and the user decides whether to adopt the quality inspection results;

[0033] If a confirmation instruction is received from the user to make a decision using the quality inspection result, then

[0034] Determine whether the current user is a member of the expert database. If so, generate positive sample data from the current quality inspection content as new training data for the quality inspection model. Otherwise, generate a quality inspection list.

[0035] If a confirmation instruction is received from the user that the quality inspection result is not to be used for decision making, it is determined whether the current user is a member of the expert database. If so, negative sample data is generated from the current quality inspection content as new training data for the quality inspection model. Otherwise, the quality inspection process is terminated.

[0036] The new training data is used as a new calibration sample to perform enhanced training on the quality inspection model.

[0037] The present application also proposes a text detection device, the device comprising:

[0038] The stripping module is used to disassemble the content of the performance test process document in response to the document quality inspection instruction, strip the result analysis text content, and form the data to be inspected;

[0039] The quality inspection module is used to input the data to be quality inspected into a pre-trained quality inspection model for quality inspection analysis to obtain quality inspection results. The quality inspection model is trained based on a pre-constructed sample data set, and the sample data set is constructed based on performance test result analysis data combined with a preset quality inspection rule set.

[0040] The present application also proposes a text detection device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the text detection method described above.

[0041] The present application also proposes a storage medium, which is a computer-readable storage medium. A computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the text detection method described above are implemented.

[0042] The present application also proposes a computer program product, characterized in that the computer program product includes a computer program, and when the computer program is executed by a processor, the steps of the text detection method described above are implemented.

[0043] One or more technical solutions proposed in this application have at least the following technical effects: by responding to the document quality inspection instruction, the data to be inspected is obtained based on the content of the performance test process document; the data to be inspected is input into a pre-trained quality inspection model for quality inspection analysis to obtain the quality inspection result, the quality inspection model is trained based on a pre-constructed sample data set, and the sample data set is constructed based on the performance test result analysis data and a preset quality inspection rule set. Thus, through the pre-constructed quality inspection model for evaluating logical correctness in the performance test field, the logical correctness evaluation of text content in a specific field is realized, and the trained quality inspection model is integrated with the software platform to provide testers and test managers with convenient and automated file content quality inspection capabilities, ensure file quality, and at the same time, improve the efficiency of process file quality inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0046] Figure 1 A flowchart of the first embodiment of the text detection method of the present application is provided;

[0047] Figure 2 A flowchart of the second embodiment of the text detection method of the present application is provided;

[0048] Figure 3 This is a general schematic diagram of an embodiment of the text detection method of this application;

[0049] Figure 4 A schematic diagram of a data set calibration process for an embodiment of a text detection method of the present application;

[0050] Figure 5 A schematic diagram of the quality inspection model training process of an embodiment of the text detection method of the present application;

[0051] Figure 6 A complete flowchart of the text detection method provided in the embodiment of the present application;

[0052] Figure 7 This is a schematic diagram of the module structure of the text detection device according to an embodiment of the present application;

[0053] Figure 8 Schematic diagram of the device structure of the hardware operating environment involved in the text detection method in the embodiment of the present application.

[0054] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0055] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0056] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0057] The main solution of the embodiment of the present application is: by responding to the document quality inspection instruction, the data to be inspected is obtained based on the content of the performance test process document; the data to be inspected is input into a pre-trained quality inspection model for quality inspection analysis to obtain the quality inspection result, wherein the quality inspection model is trained based on a pre-constructed sample data set, and the sample data set is constructed based on the performance test result analysis data and in combination with a preset quality inspection rule set. Thus, through the pre-constructed quality inspection model for evaluating logical correctness in the performance test field, the logical correctness evaluation of text content in a specific field is realized, and the trained quality inspection model is integrated with the software platform to provide testers and test managers with convenient and automated file content quality inspection capabilities, ensure file quality, and at the same time, improve the efficiency of process file quality inspection.

[0058] Technical terms involved in the embodiments of this application:

[0059] AdaBoost, Adaptive Boosting, is a machine learning method. The adaptability of the AdaBoost method lies in that the samples that were misclassified by the previous classifier will be used to train the next classifier. The AdaBoost method is very sensitive to noisy data and abnormal data. In some problems, the AdaBoost method is less likely to overfit than most other learning algorithms. The classifier used in the AdaBoost method can effectively improve the classification effect.

[0060] The AdaBoost method is an iterative algorithm. Its principle is: add a new weak classifier in each round until a predetermined sufficiently small error rate is reached. Each training sample is assigned a weight, indicating the probability of it being selected into the training set by a certain classifier. If a sample point has been accurately classified, then the probability of it being selected in the construction of the next training set is reduced; on the contrary, if a sample point is not accurately classified, then its weight is increased. In this way, the AdaBoost method can "focus" on those samples that are more difficult to classify (more informative). In specific implementation, the weight of each sample is initially made equal. For the kth iteration operation, sample points are selected according to these weights, and then the classifier Ck is trained. Then, according to this classifier, the weight of the sample that is misclassified by it is increased, and the weight of the sample that is correctly classified is reduced. Then, the sample set with updated weights is used to train the next classifier Ck. The entire training process is carried out iteratively.

[0061] GloVe: Global Vectors for Word Representation, a word vector model based on word representation and word co-occurrence matrix frequency statistics, is a word representation tool based on global word frequency statistics (count-based & overall statistics). It can express a word as a vector composed of real numbers. These vectors capture some semantic characteristics between words, such as similarity and analogy.

[0062] In the existing technology, for testers, in the scenario of limited testing human resources, it is common for a single person to handle multiple test tasks in parallel. At present, the analysis of test results is highly dependent on manual work, and is affected by multiple factors such as the skill level of testers, the amount of parallel tasks, and the difficulty of test tasks. When analyzing test results, it is very likely that there will be problems such as content errors and inaccurate content analysis due to typos, human negligence, insufficient knowledge reserves, etc.

[0063] In addition, for test managers, the amount of testing tasks is large, and the quality management workload of test managers also increases accordingly. The current process document quality inspection work also relies on manual verification, which has problems such as heavy workload, low work efficiency, and risk of missing problems.

[0064] The present application provides a solution to the problems that performance test result analysis may rely on testers, such as analytical logic errors, insufficient recognition capabilities, and low efficiency. The present application method aims to construct a model training sample data set by positive and negative labeling of existing sample data, and realize automatic quality inspection of process document text content by training a supervised classification model in combination with a software platform, so as to help testers conduct self-inspection in advance, reduce the occurrence of quality problems, and ensure the quality of process text output during the test process.

[0065] In view of the problems that management personnel's audit relies on manual input, the audit task volume is large, the efficiency is low, and there are problems such as human omissions, this application can help management personnel to conduct batch automated quality inspection and improve quality inspection efficiency. At the same time, this application method supports users to mark the results to indicate whether the current test results can be adopted, so as to construct new positive and negative sample data in real time, supplement the model training samples, and form a sustainable closed loop from sample input training to model recognition to new sample generation, continuously improving the accuracy of the detection model, and to a certain extent solving the problem that the limited sample data set may lead to the limited accuracy of the trained quality inspection model.

[0066] It should be noted that the execution subject of the embodiments of the present application may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, a server, a system platform (such as a software platform), etc., or an electronic device capable of realizing the above functions, etc. The following embodiments are described below taking the software platform as an example.

[0067] Based on this, the present application embodiment provides a text detection method, referring to Figure 1 , Figure 1 This is a flowchart of the first embodiment of the text detection method of the present application.

[0068] In this embodiment, the text detection method includes steps S10 to S20, each of which is explained in detail below.

[0069] like Figure 1 As shown, the first embodiment of the present application proposes a text detection method, which is applied to a software platform and includes:

[0070] Step S10, in response to the document quality inspection instruction, obtaining the data to be quality inspected based on the content of the performance test process document;

[0071] The embodiments of the present application mainly focus on the quality inspection of the content of performance testing process documents in the performance testing field, propose a logic correctness verification method in the performance testing field, and realize automatic quality inspection of performance testing process documents in combination with a software platform.

[0072] Among them, the technical solution of this embodiment mainly realizes the logical correctness evaluation of text content in a specific field through a pre-built quality inspection model for evaluating logical correctness in the performance testing field, integrates the trained quality inspection model with the software platform, and provides testers and test managers with convenient and automated file content quality inspection capabilities to ensure the quality of test process files. At the same time, it improves the quality inspection efficiency of test process files to meet the performance testing needs of the business system.

[0073] The execution subject of this embodiment may be a software platform, in which a quality inspection model for evaluating logical correctness is embedded. It should be noted that in other embodiments, the execution subject of the above text detection method may also be other system platforms or servers, etc., which is not specifically limited in this embodiment.

[0074] As an implementation method, a tester or a test manager may trigger a document quality inspection instruction on a software platform, and the software platform responds to the document quality inspection instruction and obtains the data to be quality inspected based on the content of the performance test process document.

[0075] The performance test process document may be a process document or file generated by the performance test of various business systems. The performance test process document may be captured, recorded or stored in a preset manner through a set storage path. For example, during the performance test process, the performance test process document generated by the test may be stored on the software platform. After receiving the document quality inspection instruction, the software platform obtains the content of the performance test process document locally.

[0076] The data to be inspected can be the content of the performance test process document, or can be extracted from the content of the performance test process document. The data to be inspected can be the performance test process file generated by the tester when performing performance testing on the business system, which can help the tester to conduct self-inspection in advance, reduce the occurrence of quality problems, and ensure the quality of the process text output during the test process. In addition, the data to be inspected can also be the result analysis text of the business system performance test process file, which can help the test management personnel to perform batch automated quality inspection and improve the efficiency of quality inspection.

[0077] As an implementation method, after the quality inspection begins, the software platform disassembles the content of the performance test process document, removes the result analysis text content, and forms data to be inspected as input data for the quality inspection model.

[0078] Step S20, input the data to be quality inspected into a pre-trained quality inspection model for quality inspection analysis to obtain quality inspection results. The quality inspection model is trained based on a pre-constructed sample data set, and the sample data set is constructed based on performance test result analysis data combined with a preset quality inspection rule set.

[0079] As mentioned above, this embodiment embeds a quality inspection model for evaluating logical correctness in the software platform. The quality inspection model of the software platform can be used to analyze and inspect the business system performance test process file, or the quality inspection model of the software platform can be used to inspect the result analysis text of the business system performance test process file.

[0080] The quality inspection model of this embodiment is a pre-built and trained logic correctness evaluation model, which is a supervised classification model trained by analyzing the positive and negative sample data classification and labeling of performance test results, and can realize the logic correctness evaluation of text content in a specific field.

[0081] The sample data set used for quality inspection model training in this embodiment is constructed based on performance test result analysis data and in combination with a preset quality inspection rule set.

[0082] Among them, as an implementation method, the construction process of the quality inspection model includes: building a quality inspection rule set, calibrating sample data and building a sample data set, training the quality inspection model, and further enhancing and supplementing the quality inspection model through real-time quality inspection results as new labeled data samples.

[0083] The quality inspection rule set is a special quality inspection rule set within the scope of the performance testing field, so as to realize the quality inspection and evaluation of the text content of the test process in a specific field.

[0084] The sample data set is constructed by combining the quality inspection rule set with the performance test result analysis data. When constructing the sample data set, the sample data also needs to be calibrated. By marking the positive and negative characteristics of the sample data, the final sample data set for model training is constructed. Among them, as an implementation method, in order to make the sample labeling more accurate in the scenario of limited data sets, the adaptive boosting algorithm Adaboost can be used for data labeling.

[0085] After obtaining the quality inspection data in the performance test process file, the data to be inspected is input into the pre-trained quality inspection model through the software platform for quality inspection analysis to obtain the quality inspection result. The quality inspection result can be displayed to users, such as testers or test managers or corresponding clients.

[0086] This embodiment uses the above scheme, specifically by responding to the document quality inspection instruction, obtaining the data to be inspected based on the content of the performance test process document; inputting the data to be inspected into a pre-trained quality inspection model for quality inspection analysis to obtain the quality inspection result, wherein the quality inspection model is trained based on a pre-constructed sample data set, and the sample data set is constructed based on the performance test result analysis data and a preset quality inspection rule set. Thus, through the pre-constructed quality inspection model for evaluating logical correctness in the performance test field, the logical correctness evaluation of text content in a specific field is achieved, the software platform capability is provided, the trained quality inspection model is integrated with the software platform, and convenient and automated file content quality inspection capabilities are provided for testers and test managers, ensuring the quality of the file, while improving the efficiency of process file quality inspection.

[0087] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the first embodiment can refer to the above introduction, and will not be repeated in the following. Figure 2 Before step S20, the text detection method further includes step S00.

[0088] Step S00, constructing a quality inspection model, wherein the quality inspection model is trained based on a pre-constructed sample data set, wherein the sample data set is constructed based on performance test result analysis data combined with a preset quality inspection rule set.

[0089] The quality inspection model of this embodiment is a logic correctness evaluation model, which is a supervised classification model trained by analyzing the positive and negative sample data classification and labeling of performance test results, and can realize the logic correctness evaluation of text content in a specific field.

[0090] The sample data set used for quality inspection model training in this embodiment is constructed based on performance test result analysis data and in combination with a preset quality inspection rule set.

[0091] Among them, as an implementation method, the construction process of the quality inspection model includes: building a quality inspection rule set, calibrating sample data and building a sample data set, training the quality inspection model, and further enhancing and supplementing the quality inspection model through real-time quality inspection results as new labeled data samples.

[0092] The text checking method of this embodiment can refer to Figure 3 As shown in the figure, a logic correctness verification method in the performance field is proposed, which realizes the automatic quality inspection of performance test process files by combining with the software platform. The specific implementation process is as follows Figure 3As shown, the construction of the quality inspection model includes: quality inspection rule set construction, positive and negative sample data calibration, and quality inspection model training. Furthermore, after the text quality inspection, the results are labeled, and the labeled data is used as a new labeled data sample to enhance and supplement the quality inspection model to improve the accuracy of the quality inspection model.

[0093] The quality inspection rule set is a special quality inspection rule set within the scope of the performance testing field, so as to realize the quality inspection and evaluation of the text content of the test process in a specific field.

[0094] A sample data set is constructed by analyzing the data based on the quality inspection rule set and the performance test results. When constructing the sample data set, the sample data also needs to be calibrated. By labeling the sample data with positive and negative characteristics, the final sample data set for model training is constructed.

[0095] Among them, as an implementation method, in order to make sample labeling more accurate in the scenario of limited data sets, the adaptive boosting algorithm Adaboost can be used for data labeling.

[0096] It should be noted that this embodiment, by analyzing data based on performance test results and combining a preset quality inspection rule set to construct a sample data set, can realize the logical correctness evaluation of text content in a specific field, integrate the trained quality inspection model with the software platform, and provide testers and test managers with convenient and automated file content quality inspection capabilities to ensure the quality of test process files. At the same time, it improves the quality inspection efficiency of test process files and meets the performance testing requirements of business systems.

[0097] Specifically, in a feasible implementation, the step S00 of building a quality inspection model may include steps A11 to A13, each of which is described in detail below:

[0098] Step A11, constructing a dedicated quality inspection rule set within a preset domain for performance testing;

[0099] In a feasible implementation, the step of constructing a dedicated quality inspection rule set within a preset domain for performance testing may include:

[0100] Obtain quality inspection factors for performance testing sorted out by experts in the preset field, and build a quality inspection rule set based on the quality inspection factors.

[0101] Specifically, for performance testing, a dedicated quality inspection rule set is constructed within the scope of the field. First, the quality inspection factors are sorted out by field experts. The quality inspection factors may include resource categories and performance categories. Resource categories include: server CPU utilization, memory occupancy, IO rate, network rate, etc. Performance categories include: response TPS, average response time, error rate, data loss rate, etc. Then, a quality inspection rule set is constructed based on the quality inspection factors, such as: CPU utilization greater than 85% may have a resource bottleneck, and the test does not meet expectations when the response TPS is less than the expected performance indicator.

[0102] Step A12, analyzing data based on known performance test results and building a sample data set in combination with the quality inspection rule set;

[0103] In a feasible implementation, the step of analyzing data based on known performance test results and building a sample data set in combination with the quality inspection rule set may include:

[0104] Collect and organize the scenario test result analysis data in the known performance test result analysis report;

[0105] Analyze data based on the scenario test results and construct an initial sample data set in combination with the quality inspection rule set;

[0106] The adaptive boosting algorithm Adaboost is used in combination with the initial sample data set to perform data labeling, and a calibrated positive and negative sample data set is obtained as the final sample data set for training the quality inspection model.

[0107] When the adaptive boosting algorithm Adaboost is used in combination with the initial sample data set for data labeling, a training sample set and a sample set to be calibrated are extracted from the initial sample data set, and the Adaboost model is trained according to the training sample set to obtain a calibration model. Then, the sample set to be calibrated is classified and calibrated by the calibration model to obtain a calibrated positive and negative sample data set as the final sample data set for training the quality inspection model.

[0108] As an implementation method, the specific implementation process of constructing a sample data set is as follows:

[0109] First, collect and organize the scenario test result analysis data in the performance test result analysis report, and build a sample data set based on the quality inspection rule set obtained in the previous steps.

[0110] Then, a certain amount of data sets are separated from the sample data sets as training sample sets for the calibration model, and the remaining data sets are used as sample sets to be calibrated.

[0111] Since the sample data set depends on the current project inventory, it is impossible to build a sample set of more than one million in the short term. In order to make sample labeling more accurate in the scenario of limited data sets, this embodiment uses the adaptive boosting algorithm Adaboost for data labeling. The Adaboost algorithm will give higher weights to misclassified samples in each iteration, so that the model will pay more attention to these negative samples in subsequent iterations.

[0112] The dataset calibration process is as follows: Figure 4 As shown, the specific process includes:

[0113] First, a certain amount of data sets are separated from the sample data sets as training sample sets for the calibration model, and the remaining data sets are used as sample sets to be calibrated.

[0114] Among them, construct the training data set T = {(x1,y1),(x2,y2)…(x N ,y N )}, where T is the current iteration and N is the number of samples.

[0115] Then, the weight distribution of the training data is initialized, and the training samples are initially given equal weights. Then the initial weight distribution of training samples is:

[0116]

[0117] Then, multiple rounds of iterative training are performed to calculate the basic classifier H in each iteration t The weight of (x) in the final classifier:

[0118]

[0119] According to the obtained weights, the weight distribution of the next round of iterative training data set is obtained to obtain a new weight distribution:

[0120]

[0121] Combine each weak classifier to get the final classifier:

[0122]

[0123] The trained final separator, also known as the calibration model, is used to classify and calibrate the sample set to be calibrated, and the calibrated positive and negative sample data are obtained as training samples for the quality inspection model in the next step, thereby obtaining the final sample data set.

[0124] Step A13: training a quality inspection model based on the sample data set to obtain a trained quality inspection model.

[0125] Specifically, the quality inspection model is trained based on the final sample data set obtained in the previous step to obtain a trained quality inspection model.

[0126] Wherein, as an implementation manner, the step of training the quality inspection model based on the sample data set to obtain the trained quality inspection model may include:

[0127] First, a positive sample corpus is constructed based on the positive sample data set, and a negative sample corpus is constructed based on the negative sample data set;

[0128] Then, based on the positive sample corpus and the negative sample corpus, word vector learning is performed using the word vector model GloVe based on word co-occurrence matrix frequency statistics, and feature vectors are obtained by feature extraction;

[0129] Finally, based on the feature vector, a Laplace kernel function based on Euclidean distance is used to perform model training to obtain a trained quality inspection model.

[0130] Among them, the quality inspection model is mainly used to verify the logical correctness of the text. That is, the quality inspection model is required to effectively identify the similarities between the input text content and the positive and negative samples, so as to identify the error analysis close to the negative samples.

[0131] The detailed process of the quality inspection model training steps in this embodiment can be referred to Figure 5 As shown, the specific process is as follows:

[0132] In this embodiment, the training of the quality inspection model can be implemented using the Laplacian kernel function Laplacian Kernel based on the Euclidean distance. The Laplacian Kernel function provides a method for measuring the similarity between two points or vectors. The specific calculation method is:

[0133]

[0134] Among them, x and y are two vectors in the feature space, ‖xy‖ represents the Manhattan distance between vectors, σ is used to control the width of the kernel function, and K(x,y) represents the similarity between two points or vectors.

[0135] In order to better perform text feature extraction, this embodiment adopts the word vector model GloVe based on word co-occurrence matrix frequency statistics, and learns word vectors through the global word co-occurrence matrix to better capture the semantics and statistical information of the vocabulary.

[0136] Among them, feature extraction includes two main steps. First, it is necessary to construct a co-occurrence matrix. According to the final sample data set obtained in the previous steps, the co-occurrence matrix X(N,N) is constructed, where N is the number of corpus words constructed by the sample data set. The frequency of two words X is obtained by sliding window traversal. ij. Then construct the loss function, the loss function model is as follows:

[0137]

[0138] Among them, v i , v j is the word vector of word i and word j in the initial stage, f(x ij ), b i 、b j It is a weighted term and bias term to enhance the robustness of the model. The loss function J is minimized through training to obtain the final feature vector for Laplacian Kernel model training, and finally a trained quality inspection model is obtained.

[0139] This embodiment uses the above scheme, specifically by constructing a quality inspection model, the quality inspection model is trained based on a pre-constructed sample data set, the sample data set is constructed based on performance test result analysis data and combined with a preset quality inspection rule set, responds to document quality inspection instructions, obtains data to be inspected based on the content of the performance test process document; inputs the data to be inspected into the trained quality inspection model for quality inspection analysis, and obtains quality inspection results. Thus, through the pre-constructed quality inspection model for evaluating logical correctness in the performance test field, the logical correctness evaluation of text content in a specific field is achieved, the software platform capability is provided, the trained quality inspection model is integrated with the software platform, and convenient and automated file content quality inspection capabilities are provided for testers and test managers, ensuring file quality, and at the same time, improving the efficiency of file quality inspection in the performance test process. Among them, when building the quality inspection model, the adaptive boosting algorithm Adaboost is used for data labeling, which can make the sample labeling more accurate in the scenario of limited data sets. In addition, when training the quality inspection model, the text feature extraction adopts the word vector model GloVe based on the frequency statistics of the word co-occurrence matrix. The word vector is learned through the global word co-occurrence matrix to better capture the semantic and statistical information of the vocabulary. The training of the quality inspection model is implemented using the Laplacian kernel function Laplacian Kernel based on the Euclidean distance to enhance the robustness and accuracy of the model.

[0140] Based on the second embodiment of the present application, in the third embodiment of the present application, the same or similar contents as those in the second embodiment can be referred to the above description, and will not be repeated hereafter. On this basis, after step S00, the text detection method further includes step S30, and after step S20, the text detection method further includes step S40.

[0141] Step S30, embedding the trained quality inspection model into the software platform;

[0142] Step S40, obtaining the manual marking results of the quality inspection results provided by the software platform, integrating the real-time data into new calibration samples according to the manual marking results, and performing enhanced training on the quality inspection model.

[0143] As mentioned above, the text detection method of this application is a logic correctness verification method in the performance field, which combines the software platform to realize the automatic quality inspection of the performance test process file. The specific implementation process is as follows: Figure 3 As shown, the construction of the quality inspection model includes: quality inspection rule set construction, positive and negative sample data calibration, and quality inspection model training. Furthermore, after the text quality inspection, the results are labeled, and the labeled data is used as a new labeled data sample to enhance and supplement the quality inspection model to improve the accuracy of the quality inspection model.

[0144] Specifically, after the quality inspection model is trained through the above-mentioned embodiment, the trained quality inspection model is embedded in the software platform as the initial quality inspection model, and the user operation is online through the software platform to improve process automation and operational convenience. At the same time, the software platform provides manual labeling of quality inspection results, and integrates real-time data into new calibration samples according to the labeling results, as a supplement to the model training samples, and continuously improves the accuracy of the model. Among them, in order to ensure the accuracy of the calibration samples, the calibration data conversion can be performed only for the members of the expert database.

[0145] As an implementation method, the step of obtaining the manual marking results of the quality inspection results provided by the software platform, integrating the real-time data into a new calibration sample according to the manual marking results, and performing enhanced training on the quality inspection model may include:

[0146] The quality inspection results are displayed online, and the user decides whether to adopt the quality inspection results;

[0147] If a confirmation instruction is received from the user to make a decision using the quality inspection result, then

[0148] Determine whether the current user is a member of the expert database. If so, generate positive sample data from the current quality inspection content as new training data for the quality inspection model. Otherwise, generate a quality inspection list.

[0149] If a confirmation instruction is received from the user that the quality inspection result is not to be used for decision making, it is determined whether the current user is a member of the expert database. If so, negative sample data is generated from the current quality inspection content as new training data for the quality inspection model. Otherwise, the quality inspection process is terminated.

[0150] The new training data is used as a new calibration sample to perform enhanced training on the quality inspection model.

[0151] The complete process of the text detection method in this embodiment can be referred to Figure 6As shown, the user platform operation process is as follows:

[0152] At the beginning of the quality inspection, the software platform disassembles the document content and strips the result analysis text content as the input data of the quality inspection model. The quality inspection model obtained in the previous steps is used to perform quality inspection and analysis on the text content, and the quality inspection results are displayed online. The user makes a decision on whether to adopt the result. If the user adopts it, it is determined whether the current user is a member of the expert library. If so, the current quality inspection content generates positive sample data as new training data for the quality inspection model. Otherwise, a quality inspection list is directly generated. If the user does not adopt it, it is determined whether the current user is a member of the expert library. If so, the current quality inspection content generates negative sample data as new training data for the quality inspection model. Otherwise, the quality inspection ends directly.

[0153] This embodiment uses the above scheme, specifically by constructing a quality inspection model, the quality inspection model is trained based on a pre-constructed sample data set, the sample data set is constructed based on performance test result analysis data and combined with a preset quality inspection rule set; embedding the trained quality inspection model into the software platform; responding to the document quality inspection instruction, obtaining the data to be inspected based on the content of the performance test process document; inputting the data to be inspected into the trained quality inspection model for quality inspection analysis to obtain the quality inspection result; obtaining the manual marking result of the quality inspection result provided by the software platform, integrating the real-time data into a new calibration sample according to the manual marking result, and performing enhanced training on the quality inspection model. Thus, through the pre-constructed quality inspection model for evaluating logical correctness in the performance test field, the logical correctness evaluation of text content in a specific field is realized, the software platform capability is provided, the trained quality inspection model is integrated with the software platform, and convenient and automated file content quality inspection capabilities are provided for testers and test managers to ensure the quality of the file, and at the same time, the efficiency of file quality inspection in the performance test process is improved. Among them, when building the quality inspection model, the adaptive boosting algorithm Adaboost is used for data labeling, which can make the sample labeling more accurate in the scenario of limited data sets. In addition, when training the quality inspection model, the text feature extraction adopts the word vector model GloVe based on the frequency statistics of the word co-occurrence matrix. The word vector is learned through the global word co-occurrence matrix to better capture the semantic and statistical information of the vocabulary. The training of the quality inspection model is implemented using the Laplacian kernel function Laplacian Kernel based on the Euclidean distance to enhance the robustness and accuracy of the model. By labeling the real-time quality inspection results, new positive and negative sample data are formed, and the model is continuously enhanced and trained, forming a sustainable closed loop from sample input training to model recognition to new sample generation, and continuously improving the accuracy of the detection model.

[0154] Compared with the prior art, the scheme of this embodiment aims at the problems that performance result analysis relies on testers, which may have analysis logic errors, insufficient recognition ability, low efficiency, etc. The method of the present invention constructs a model training sample data set by positive and negative labeling of existing sample data, and realizes automatic quality inspection of process document text content by training a supervised classification model in combination with a software platform, helping testers to conduct self-inspection in advance, reduce the occurrence of quality problems, and ensure the quality of process text output during the test process.

[0155] In view of the problems that management personnel's audit relies on manual input, the audit task is large, the efficiency is low, and there are problems such as human omissions, the present invention can help management personnel to conduct batch automated quality inspection and improve the efficiency of quality inspection. At the same time, the method of the present invention supports users to mark the results to indicate whether the current test results can be adopted, so as to construct new positive and negative sample data in real time, supplement the model training samples, and form a sustainable closed loop from sample input training to model recognition to new sample generation, continuously improve the accuracy of the detection model, and to a certain extent solve the problem that the limited sample data set may lead to the limited accuracy of the trained quality inspection model.

[0156] In addition, the present application also proposes a text detection device, such as Figure 7 As shown, the device comprises:

[0157] An acquisition module 10, for obtaining the data to be inspected based on the content of the performance test process document in response to the document quality inspection instruction;

[0158] The quality inspection module 20 is used to input the data to be quality inspected into a pre-trained quality inspection model for quality inspection analysis to obtain quality inspection results. The quality inspection model is trained based on a pre-constructed sample data set, and the sample data set is constructed based on performance test result analysis data combined with a preset quality inspection rule set.

[0159] The text detection device provided by the present application adopts the text detection method in the above-mentioned embodiment, can realize the logical correctness evaluation of text content in a specific field, provide software platform capabilities, integrate the trained quality inspection model with the software platform, and provide testers and test managers with convenient and automated file content quality inspection capabilities to ensure file quality. At the same time, it improves the efficiency of file quality inspection during the performance testing process, and supports sample data calibration and supplementation to realize a self-enhanced closed-loop process from training to quality inspection to enhanced training, thereby improving the accuracy of quality inspection evaluation. Compared with the prior art, the beneficial effects of the text detection device provided by the present application are the same as the beneficial effects of the text detection method provided by the above-mentioned embodiment, and the other technical features in the text detection device are the same as the features disclosed in the above-mentioned embodiment method, which will not be repeated here.

[0160] An embodiment of the present application further proposes a text detection device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the text detection method described above.

[0161] Reference below Figure 8 , which shows a schematic diagram of the structure of a text detection device suitable for implementing the embodiment of the present application. The text detection device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 8 The text detection device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0162] like Figure 8 As shown, the text detection device may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the xxx device are also stored. The processing device 1001, ROM1002, and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the text detection device to communicate wirelessly or wired with other devices to exchange data. Although the text detection device with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have alternatively.

[0163] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0164] The text detection device provided by the present application adopts the text detection method in the above embodiment, can realize the logical correctness evaluation of text content in a specific field, provide software platform capabilities, integrate the trained quality inspection model with the software platform, and provide testers and test managers with convenient and automated file content quality inspection capabilities to ensure file quality. At the same time, it improves the efficiency of file quality inspection during the performance testing process, and supports sample data calibration and supplementation to realize a self-enhanced closed-loop process from training to quality inspection to enhanced training, thereby improving the accuracy of quality inspection evaluation. Compared with the prior art, the beneficial effects of the text detection device provided by the present application are the same as the beneficial effects of the text detection method provided in the above embodiment, and the other technical features in the text detection device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0165] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0166] An embodiment of the present application further proposes a storage medium, which is a computer-readable storage medium. A computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the text detection method described above are implemented.

[0167] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0168] The computer-readable storage medium may be included in the text detection device; or may exist independently without being assembled into the text detection device.

[0169] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the text detection device, the text detection device: responds to the document quality inspection instruction and obtains the data to be quality inspected based on the content of the performance test process document; inputs the data to be quality inspected into a pre-trained quality inspection model for quality inspection analysis to obtain quality inspection results, wherein the quality inspection model is trained based on a pre-constructed sample data set, and the sample data set is constructed based on the performance test result analysis data and in combination with a preset quality inspection rule set.

[0170] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0171] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0172] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0173] The readable storage medium provided in the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned text detection method, and can realize the logical correctness evaluation of text content in a specific field, provide software platform capabilities, integrate the trained quality inspection model with the software platform, and provide testers and test managers with convenient and automated file content quality inspection capabilities to ensure file quality. At the same time, it improves the efficiency of file quality inspection during the performance testing process, and supports sample data calibration and supplementation, realizing a self-enhanced closed-loop process from training to quality inspection to enhanced training, and improving the accuracy of quality inspection evaluation. Compared with the prior art, the beneficial effects of the storage medium provided in the present application are the same as the beneficial effects of the text detection method provided in the above-mentioned embodiment, and the other technical features in the storage medium are the same as the features disclosed in the above-mentioned embodiment method, which will not be repeated here.

[0174] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps of the text detection method described above are implemented.

[0175] The embodiments of the present application can realize the logical correctness evaluation of text content in a specific field, provide software platform capabilities, integrate the trained quality inspection model with the software platform, provide testers and test managers with convenient and automated file content quality inspection capabilities, ensure file quality, and at the same time, improve the efficiency of file quality inspection during performance testing, and support sample data calibration and supplementation, realize a self-enhanced closed-loop process from training to quality inspection to enhanced training, and improve the accuracy of quality inspection evaluation. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as the beneficial effects of the text detection method provided by the above embodiments, and the other technical features in the computer program product are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0176] One or more technical solutions proposed in this application have at least the following technical effects: by responding to the document quality inspection instruction, the data to be inspected is obtained based on the content of the performance test process document; the data to be inspected is input into a pre-trained quality inspection model for quality inspection analysis to obtain the quality inspection result, the quality inspection model is trained based on a pre-constructed sample data set, and the sample data set is constructed based on the performance test result analysis data and a preset quality inspection rule set. Thus, through the pre-constructed quality inspection model for evaluating logical correctness in the performance test field, the logical correctness evaluation of text content in a specific field is realized, and the trained quality inspection model is integrated with the software platform to provide testers and test managers with convenient and automated file content quality inspection capabilities, ensure file quality, and at the same time, improve the efficiency of process file quality inspection. Among them, when building the quality inspection model, the adaptive boosting algorithm Adaboost is used for data labeling, which can make the sample labeling more accurate in the scenario of limited data sets. In addition, when training the quality inspection model, the text feature extraction adopts the word vector model GloVe based on the frequency statistics of the word co-occurrence matrix. The word vector is learned through the global word co-occurrence matrix to better capture the semantic and statistical information of the vocabulary. The training of the quality inspection model is implemented using the Laplacian kernel function LaplacianKernel based on the Euclidean distance to enhance the robustness and accuracy of the model; by labeling the real-time quality inspection results, new positive and negative sample data are formed, and the model is continuously enhanced and trained, forming a sustainable closed loop from sample input training to model recognition to new sample generation, and continuously improving the accuracy of the detection model.

[0177] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A text detection method, characterized in that: The method is applied to a software platform, and the method comprises: In response to the document quality inspection instruction, obtaining the data to be quality inspected based on the content of the performance test process document; The data to be quality inspected is input into a pre-trained quality inspection model for quality inspection analysis to obtain quality inspection results. The quality inspection model is trained based on a pre-constructed sample data set, and the sample data set is constructed based on performance test result analysis data combined with a preset quality inspection rule set.

2. The method according to claim 1, characterized in that Before the step of inputting the data to be inspected into a pre-trained quality inspection model for quality inspection analysis and obtaining the quality inspection result, the step further includes: Build a dedicated quality inspection rule set within a preset domain for performance testing; Analyze data based on known performance test results and construct a sample data set in combination with the quality inspection rule set; A quality inspection model is trained based on the sample data set to obtain a trained quality inspection model.

3. The method according to claim 2, characterized in that The step of constructing a dedicated quality inspection rule set within a preset domain for performance testing includes: Obtain quality inspection factors for performance testing compiled by experts in the preset field; A quality inspection rule set is constructed according to the quality inspection factors.

4. The method according to claim 2, characterized in that The step of analyzing data based on known performance test results and building a sample data set in combination with the quality inspection rule set includes: Collect and organize the scenario test result analysis data in the known performance test result analysis report; Analyze data based on the scenario test results and construct an initial sample data set in combination with the quality inspection rule set; The adaptive boosting algorithm Adaboost is used in combination with the initial sample data set to perform data labeling, and a calibrated positive and negative sample data set is obtained as the final sample data set for training the quality inspection model.

5. The method according to claim 4, characterized in that The step of using the adaptive boosting algorithm Adaboost and combining the initial sample data set to perform data labeling to obtain a calibrated positive and negative sample data set as the final sample data set for training the quality inspection model includes: Extracting a training sample set and a sample set to be calibrated from the initial sample data set; Perform Adaboost model training according to the training sample set to obtain a calibration model; The calibration model is used to classify and calibrate the sample set to be calibrated to obtain a calibrated positive and negative sample data set as the final sample data set for training the quality inspection model.

6. The method according to claim 5, characterized in that The step of training the quality inspection model based on the sample data set to obtain a trained quality inspection model comprises: A positive sample corpus is constructed based on the positive sample data set, and a negative sample corpus is constructed based on the negative sample data set; Based on the positive sample corpus and the negative sample corpus, word vector learning is performed using the word vector model GloVe based on word co-occurrence matrix frequency statistics, and feature vectors are obtained by feature extraction; Based on the feature vector, a Laplace kernel function based on Euclidean distance is used to perform model training to obtain a trained quality inspection model.

7. The method according to any one of claims 2 to 6, characterized in that The method further comprises: Embedding the trained quality inspection model into the software platform; Obtain the manual marking results of the quality inspection results provided by the software platform, integrate the real-time data into new calibration samples according to the manual marking results, and perform enhanced training on the quality inspection model.

8. The method according to claim 7, characterized in that The steps of obtaining the manual marking results of the quality inspection results provided by the software platform, integrating the real-time data into new calibration samples according to the manual marking results, and performing enhanced training on the quality inspection model include: The quality inspection results are displayed online, and the user decides whether to adopt the quality inspection results; If a confirmation instruction is received from the user to make a decision using the quality inspection result, then Determine whether the current user is a member of the expert database. If so, generate positive sample data from the current quality inspection content as new training data for the quality inspection model. Otherwise, generate a quality inspection list. If a confirmation instruction is received from the user that the quality inspection result is not to be used for decision making, it is determined whether the current user is a member of the expert database. If so, negative sample data is generated from the current quality inspection content as new training data for the quality inspection model. Otherwise, the quality inspection process is terminated. The new training data is used as a new calibration sample to perform enhanced training on the quality inspection model.

9. A text detection device, characterized in that: The device comprises: The stripping module is used to disassemble the content of the performance test process document in response to the document quality inspection instruction, strip the result analysis text content, and form the data to be inspected; The quality inspection module is used to input the data to be quality inspected into a pre-trained quality inspection model for quality inspection analysis to obtain quality inspection results. The quality inspection model is trained based on a pre-constructed sample data set, and the sample data set is constructed based on performance test result analysis data and combined with a preset quality inspection rule set.

10. A text detection device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the text detection method according to any one of claims 1 to 8.

11. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the text detection method according to any one of claims 1 to 8 are implemented.

12. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the text detection method according to any one of claims 1 to 8 are implemented.