Software automated testing method, system and electronic device

Abstract

This invention relates to a software automated testing method, system, and electronic device, belonging to the field of software testing technology. The method includes: obtaining initial test cases by performing correlation analysis with detailed design documents and software architecture design documents, along with the C code and model; obtaining a JSON file by performing structured parsing and semantic compression on the C code and model; using the JSON file and initial test cases as input to a large language model; the large language model outputting a target JSON file; inputting the target JSON file into a credibility assessment module to obtain a test score; if the test score is greater than or equal to a preset score, obtaining a test result list based on the target JSON file and initial test cases. This invention combines a large language model to test various functions of the software during the testing process to obtain a test result list, and uses credibility assessment for filtering, which can effectively improve test coverage and testing efficiency.

Description

Technical Field

[0001] This invention belongs to the field of software testing technology, specifically relating to a software automated testing method, system, and electronic device. Background Technology

[0002] With the widespread application of computer software systems in industrial control, internet services, mobile terminals, and enterprise information systems, the complexity of software functions, the hierarchy of business logic, and the scale of system interactions are continuously increasing. Software quality and operational stability have become crucial for the reliable operation of systems. Software testing, as a core component in ensuring software quality, verifies program functionality, performance, interfaces, and abnormal scenarios. It can identify code defects, logical errors, and deviations from requirements in advance, reducing the risk of system failures after deployment.

[0003] Traditional software testing methods typically rely primarily on manual testing, supplemented by basic automated testing tools to execute certain processes. During test execution, testers verify functionality by manually inputting test data, observing interface feedback, and recording execution results, based on requirements documents and test case designs. They then use fixed scripts to repeatedly execute tests on fixed paths and with fixed data. Because of the subjective human factors involved, traditional software testing is prone to low test coverage and inefficient testing. Summary of the Invention

[0004] To address this issue, the present invention provides a software automated testing method, system, and electronic device to solve the problems of low test coverage and low testing efficiency that are common in existing software testing methods.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a software automated testing method, comprising: After parsing and extracting comments from the C code and model, a detailed design document is generated based on the comments; software architecture design documents are obtained. Initial test cases were obtained by correlating the detailed design documents and software architecture design documents with the C code and model; The C code and model are structured and semantically compressed to obtain a JSON file; the JSON file is an intermediate representation file that can be stably processed by a large language model; The JSON file and initial test cases are used as input to the large language model, and the large language model outputs the target JSON file. Input the target JSON file into the credibility assessment module to obtain a test score; If the test score is greater than or equal to the preset score, a list of test results will be obtained based on the target JSON file and the initial test cases.

[0006] Furthermore, after parsing and extracting the comments from the C code and model, a detailed design document is generated based on the comments, including: The task launch script retrieves C code and models; The model code mapping module converts C code and natural language from the model into detailed design documents. The model code mapping module is a model code design document parsing module with a detailed design document database at its core.

[0007] Furthermore, the initial test cases are obtained by correlating the detailed design documents and software architecture design documents with the C code and model, including: The detailed design document and software architecture design document are input into the deep learning network to obtain the requirement matrix vector and architecture matrix vector; the deep learning network is a deep learning network constructed through parsing mapping of a large language model and a coding large model; The clustering results are obtained by storing semantically related demand matrix vectors and architecture matrix vectors in correspondence using a similarity clustering model; The clustering results are input into a large language model to obtain initial test cases.

[0008] Furthermore, the process of obtaining a JSON file by performing structured parsing and semantic compression on the C code and model includes: The C code and model are structured and broken down using a script to obtain a structure description file and a data file. The structure description file and data file are reconstructed to obtain an XML file; the XML file is a file that can be understood by a large language model. The XML file is semantically compressed to obtain a JSON file; the JSON file is a small XML-like semantic JSON package.

[0009] Furthermore, the JSON file and initial test cases are taken as input to the large language model, and the large language model outputs the target JSON file, including: The initial test cases are input into the large language model as prompts for the function to be tested. Use the JSON file as context input to the large language model for the function under test; After testing the function under test using a large language model, the target JSON file is output.

[0010] Furthermore, the step of using a large language model to test the function under test and then outputting a target JSON file includes: The embedding module in the large language model is used to reconstruct the prompts and context of the function to be tested, thereby obtaining semantic vectors and architectural semantic vectors. The target JSON file is obtained by segmenting the data into blocks based on semantic vectors and architectural semantic vectors through similarity mapping.

[0011] Furthermore, the step of inputting the target JSON file into the credibility evaluation module to obtain a test score includes: A Bayesian probabilistic graphical model is constructed using the credibility assessment module's incorrect responses, large model illusion phenomenon, and repetitive conversations. The reliability of the target JSON file is tested using the reliability formula of the Bayesian probabilistic graphical model to obtain a test score. The reliability formula is as follows: P(Tr|E)=P(E|Tr)×P(Tr)÷P(E); Where Tr represents the credibility state, E represents the evidence set, P() is the probability function, and P(Tr|E) is the test score.

[0012] Furthermore, the test result list obtained based on the target JSON file and the initial test cases includes: Extract test steps, expected results, and actual test results from the target JSON file; The test results list is obtained by associating and storing the test steps, expected results, and actual test results with the corresponding initial test cases.

[0013] Secondly, the present invention provides a software automated testing apparatus, comprising: The pull module is used to parse and extract comments from C code and models, and then generate detailed design documents based on the comments; it also retrieves software architecture design documents. The initial test case module is used to obtain initial test cases by performing correlation analysis with the detailed design document and software architecture design document, the C code, and the model. The compression module is used to perform structured parsing and semantic compression on the C code and model to obtain a JSON file; the JSON file is an intermediate representation file that can be stably processed by a large language model; The large language module is used to take a JSON file and initial test cases as input to the large language model, and the large language model outputs the target JSON file. The scoring module is used to input the target JSON file into the credibility assessment module to obtain a test score; The generation module generates a list of test results based on the target JSON file and the initial test cases if the test score is greater than or equal to the preset score.

[0014] Thirdly, the present invention provides an electronic device, comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform any of the aforementioned automated software testing methods.

[0015] The present invention, by adopting the above technical solution, has at least the following beneficial effects: This invention provides a software automation testing method, system, and electronic device. After parsing and extracting comments from C code and models, a detailed design document is generated based on the comments. Simultaneously, a software architecture design document is obtained. Initial test cases are derived by performing correlation analysis between the detailed design document, the software architecture design document, and the C code and model. The C code and model are then structured and semantically compressed to obtain a JSON file. This JSON file and the initial test cases are used as input to a large language model, which outputs a target JSON file. The target JSON file is then input to a reliability evaluation module to obtain a test score. If the test score is greater than or equal to a preset score, a test result list is generated based on the target JSON file and the initial test cases. This invention combines a large language model with test scores for filtering during the testing process, effectively improving test coverage and efficiency.

[0016] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating an exemplary embodiment of the present invention of a software automated testing method; Figure 2 This is a schematic block diagram of an automated software testing system according to an exemplary embodiment of the present invention; Figure 3 This is a schematic diagram of an electronic device illustrated in an exemplary embodiment of the present invention.

[0019] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0021] This invention combines a large language model to test various functions of the software during the testing process, obtaining a list of test results, and then filtering them through credibility assessment, which can effectively improve test coverage and testing efficiency. Specifically, C code refers to C program code blocks, each consisting of multiple lines of C program code; CFG (Control Flow Graph); JSON (JavaScript Object Notation) is a data storage structure; the coding large model (code generation large model); and the embedding module serves as the core support for the large language model to process structured and semi-structured data.

[0022] The methods and systems of the present invention will be described below through specific embodiments.

[0023] Please see Figure 1 , Figure 1 This is a flowchart illustrating an exemplary embodiment of the software automated testing method of the present invention. See also: Figure 1 The method includes: Step S11: After parsing and extracting comments from the C code and model, generate a detailed design document based on the comments; obtain the software architecture design document; Step S12: Based on the detailed design document and software architecture design document, perform correlation analysis with the C code and model to obtain initial test cases; Step S13: After performing structured parsing and semantic compression on the C code and model, a JSON file is obtained; the JSON file is an intermediate representation file that can be stably processed by a large language model; Step S14: Take the JSON file and initial test cases as input to the large language model, and the large language model outputs the target JSON file; Step S15: Input the target JSON file into the credibility evaluation module to obtain the test score; Step S16: If the test score is greater than or equal to the preset score, then obtain the test result list based on the target JSON file and the initial test cases.

[0024] It should be noted that the technical solution provided in this embodiment can be used in practice as a mini-program or a plugin within an existing testing system or application, or as a standalone application that implements testing functions through external interfaces. Applicable scenarios include, but are not limited to, software testing.

[0025] It is understood that the method provided in this embodiment combines a large language model to test various functions of the software during the testing process to obtain a list of test results, and then filters them through a credibility assessment, which can effectively improve test coverage and testing efficiency.

[0026] In practice, step S11, "parse and extract comments from C code and model, and then generate detailed design documents based on the comments; obtain software architecture design documents", includes: pulling C code and model through a task startup script; converting the natural language in C code and model into detailed design documents through a model code mapping module; the model code mapping module is a model code design document parsing module with a detailed design document database as its core.

[0027] It should be noted that the process of retrieving C code and models, and obtaining detailed design documents and software architecture design documents linked to the C code, is implemented through a task startup script. Since the detailed design documents do not break down code blocks, and the correct mapping between requirements and code is a prerequisite for independent testing, this step aims to map the retrieved detailed design documents and code. This method employs an enhanced retrieval-driven model-code mapping module, namely the detailed design document parsing module. This module, centered on a detailed design document database, is composed of an independent AI large language model agent responsible for converting the natural language in the detailed design documents into a formalized language that the large language model can fully understand and that avoids ambiguity. It outputs the detailed design documents and the mapping relationship between the detailed design documents and the C code / model, with the mapping relationship serving as an index.

[0028] In practice, step S12, "obtaining initial test cases by performing correlation analysis with the detailed design document and software architecture design document, along with the C code and model," includes: inputting the detailed design document and software architecture design document into a deep learning network to obtain requirement matrix vectors and architecture matrix vectors; the deep learning network is a deep learning network constructed through parsing mapping of an AI large language model and a coding large model; storing semantically related requirement matrix vectors and architecture matrix vectors through a similarity clustering model to obtain clustering results; and inputting the clustering results into the large language model to obtain initial test cases.

[0029] It should be noted that the process involves extracting requirement slices, indexes, and software architecture design documents from the generated detailed design documents. The software architecture design documents include architecture control flow diagrams, which include sequence diagrams (Native-XML), class diagrams (Markdown), etc. A deep learning network is constructed by parsing and mapping the AI ​​large language model (processing requirement slices and architecture control flow diagrams from the detailed design documents) with a deep symbol retrieval + reinforcement learning coding large model. The deep learning network constructs embedding tuples for the detailed design documents and software architecture design documents, respectively, and outputs requirement matrix vectors and architecture matrix vectors, which are then fed into a similarity clustering model. The similarity clustering model matches semantically related requirement vectors and architecture vectors together, and the clustering results of the output vectors are then divided into blocks and fed into the large language model to write and obtain initial test cases.

[0030] In practice, step S13, "obtaining a JSON file by performing structured parsing and semantic compression on the C code and model," includes: parsing and splitting the C code and model using a script to obtain a structure description file and a data file; reconstructing the structure description file and the data file to obtain an XML file; the XML file is a file that a large language model can understand; and semantically compressing the XML file to obtain a JSON file; the JSON file is a minimally sized XML-like semantic JSON package.

[0031] It should be noted that the process involves pulling C code and the model, using a script to split and unpack the C code and the model to obtain a structure description file and a data file; through algorithm analysis, the structure description file and the data file are combined and reconstructed into an XML file that the large language model can understand; the XML file obtained in the previous step is semantically compressed into a token using an overfitting model. The token is a JSON file package with minimal XML-like semantics, which is then given to the large model for code or model analysis.

[0032] Specifically, when inputting C code and the model along with the generated initial test cases into the large model, it is necessary to fully consider the contextual capabilities of the large language model. To maximize the utilization of the large model's context token, this method proposes a context compression technique, which reorganizes the code into a simplified JSON file package containing key information. This achieves extreme compression of contextual semantics without affecting the interpretation of the large model. The agent, through the A2A protocol, first groups and tests the initial test cases corresponding to the functions with the JSON file packages of the C code mapped to them.

[0033] In practice, step S14, "using the JSON file and initial test cases as input to the large language model, and the large language model outputting the target JSON file," includes: inputting the initial test cases as prompts for the function under test into the large language model; inputting the JSON file as context for the function under test into the large language model; and outputting the target JSON file after testing the function under test using the large language model.

[0034] Specifically, the embedding module in the large language model is used to reconstruct the prompts and context of the function to be tested to obtain semantic vectors and architectural semantic vectors; based on the semantic vectors and architectural semantic vectors, the target JSON file is obtained by segmenting the data through similarity mapping.

[0035] It should be noted that the initial test cases are retrieved as prompts; the JSON file is used as context; a large model is input for testing; and the output results are output in JSON format to obtain the target JSON file.

[0036] In practice, step S15, "inputting the target JSON file into the credibility assessment module to obtain a test score," includes: constructing a Bayesian probabilistic graphical model using incorrect answers, large model illusions, and repetitive conversations from the credibility assessment module; and conducting a credibility test on the target JSON file based on the reliability formula of the Bayesian probabilistic graphical model to obtain a test score. The reliability formula is: P(Tr|E)=P(E|Tr)×P(Tr)÷P(E); where Tr represents the credibility state, E represents the evidence set, P() is the probability function, and P(Tr|E) is the fault tolerance weight coefficient.

[0037] It should be noted that the TrustScore algorithm, based on the Trustworthiness Dynamic Evaluation System, trains a prediction model using historical error logs to obtain thresholds for different C code and models as preset scores. A Bayesian probabilistic graphical model is then constructed to evaluate the reliability of the test results: P(Tr|E) = P(E|Tr) × P(Tr) ÷ P(E), where Tr represents the trustworthiness state, and the evidence set E includes observational indicators such as historical test consistency and code change frequency. Depending on the specific code or model test project, different fault tolerance weight coefficients are set. The target JSON file is input into the Bayesian probabilistic graphical model to obtain the test score. If the test score is greater than the corresponding preset score, a test result list is obtained based on the target JSON file and the initial test cases.

[0038] In practice, step S16, "obtaining a test result list based on the target JSON file and initial test cases," includes: extracting test steps, expected results, and actual test results from the target JSON file; and storing the test steps, expected results, and actual test results in association with the corresponding initial test cases to obtain the test result list.

[0039] It should be noted that the test results are input into the dynamic trustworthiness evaluation model (TrustScore) for result trustworthiness evaluation. If the test score is greater than or equal to the preset score, a Markdown format list containing the initial test cases, test steps, expected results and actual test results will be output.

[0040] In practice, this method is based on a cloud-deployed intelligent agent cluster, where each intelligent agent monitors independent subtasks (code retrieval / document parsing / model compression, etc.) and dynamically adjusts the scheduling strategy based on real-time resource load (CPU / memory / bandwidth) and task priority.

[0041] In practice, this method also includes multimodal collaborative verification of C code, models, detailed design documents, and architecture design documents. Meta-learning methods are used to perform semantic processing on the C code and models, detailed design documents, and architecture design documents. Four agents are used as the semantic layers for the detailed design document, architecture design document, C code, and models, respectively. In the detailed design document semantic layer, the agents segment sentences, transforming them into logically clear CFG (Directed Acyclic Graph) for embedding, and then chunking the entire detailed design document. In the architecture design document semantic layer, the agents embed embedding words into the Normal-XML format architecture diagrams in the architecture design document, and then... The architecture diagram is chunked, and the vectors transformed from detailed design and architecture design are clustered using the cosine similarity algorithm. The clustering requirements and architecture are then input into the large model to generate initial test cases. The C code semantic layer parses and transforms the code, uses CodeBERT to encode code segments, performs symbol encoding embedding, and then chunks to effectively avoid key information being split into other blocks, which could lead to semantic errors and incorrect test results. The model semantic layer uses a fuzzy type inference algorithm to combine the structure description file (XML) with the data to reconstruct the structure. The transformed data fragments are then concatenated to the corresponding positions in the structure description file (XML file) to become a new XML file that the large language model can understand.

[0042] Please see Figure 2 , Figure 2 This is a schematic block diagram of an automated software testing system according to an exemplary embodiment of the present invention. See also: Figure 2 The software automated testing system 100 includes: Pull module 101 is used to parse and extract comments from C code and models, and then generate detailed design documents based on the comments; obtain software architecture design documents; The initial test case module 102 is used to obtain initial test cases by performing correlation analysis with the detailed design document and software architecture design document, the C code and model. The compression processing module 103 is used to perform structured parsing and semantic compression on the C code and model to obtain a JSON file; the JSON file is an intermediate representation file that can be stably processed by a large language model; The large language module 104 is used to take a JSON file and initial test cases as input to the large language model, and the large language model outputs the target JSON file. The scoring module 105 is used to input the target JSON file into the credibility assessment module to obtain a test score; Module 106 generates a list of test results based on the target JSON file and the initial test cases if the test score is greater than or equal to the preset score.

[0043] It should be noted that the device provided in this embodiment is applicable to scenarios including but not limited to: software testing.

[0044] It is understood that the apparatus provided in this embodiment parses and extracts comments from C code and the model, generates a detailed design document based on the comments, and simultaneously obtains a software architecture design document. It then performs correlation analysis with the detailed design document and the software architecture design document, along with the C code and the model, to obtain initial test cases. After structured parsing and semantic compression of the C code and the model, a JSON file is obtained. The JSON file and the initial test cases are used as input to a large language model, which outputs a target JSON file. This target JSON file is then input to a credibility evaluation module to obtain a test score. If the test score is greater than or equal to a preset score, a test result list is obtained based on the target JSON file and the initial test cases. This invention combines a large language model during the testing process and uses test scores for filtering, which can effectively improve test coverage and testing efficiency.

[0045] Please see Figure 3 , Figure 3 This is a schematic diagram of an electronic device illustrated in an exemplary embodiment of the present invention. See also: Figure 3 The electronic device 200 includes: at least one processor 202; and Memory 201 is communicatively connected to at least one processor 202; wherein, The memory 201 stores instructions that can be executed by at least one processor 202, which enables the at least one processor 202 to perform any of the above-described software automation testing methods.

[0046] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0047] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0048] It should also be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0049] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0050] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0051] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method of software automated testing, characterized by, The method includes: After parsing and extracting comments from the C code and model, a detailed design document is generated based on the comments; software architecture design documents are obtained. Initial test cases were obtained by correlating the detailed design documents and software architecture design documents with the C code and model; The C code and model are structured and semantically compressed to obtain a JSON file; the JSON file is an intermediate representation file that can be stably processed by a large language model; The JSON file and initial test cases are used as input to the large language model, and the large language model outputs the target JSON file. Input the target JSON file into the credibility assessment module to obtain a test score; If the test score is greater than or equal to the preset score, a list of test results will be obtained based on the target JSON file and the initial test cases.

2. The method according to claim 1, characterized in that, After parsing and extracting comments from the C code and model, a detailed design document is generated based on the comments, including: The task launch script retrieves C code and models; The model code mapping module converts C code and natural language from the model into detailed design documents. The model code mapping module is a model code design document parsing module with a detailed design document database at its core.

3. The method according to claim 2, characterized in that, The initial test cases are obtained by correlating the detailed design documents and software architecture design documents with the C code and model, including: The detailed design document and software architecture design document are input into the deep learning network to obtain the requirement matrix vector and architecture matrix vector; the deep learning network is a deep learning network constructed through parsing mapping of a large language model and a coding large model; The clustering results are obtained by storing semantically related demand matrix vectors and architecture matrix vectors in correspondence using a similarity clustering model; The clustering results are input into a large language model to obtain initial test cases.

4. The method according to claim 1, characterized in that, The process of obtaining a JSON file by performing structured parsing and semantic compression on the C code and model includes: The C code and model are structured and broken down using a script to obtain a structure description file and a data file. The structure description file and data file are reconstructed to obtain an XML file; the XML file is a file that can be understood by a large language model. The XML file is semantically compressed to obtain a JSON file; the JSON file is a small XML-like semantic JSON package.

5. The method according to claim 1, characterized in that, The process involves taking a JSON file and initial test cases as input to a large language model, and the large language model outputting a target JSON file, including: The initial test cases are input into the large language model as prompts for the function to be tested. Use the JSON file as context input to the large language model for the function under test; After testing the function under test using a large language model, the target JSON file is output.

6. The method according to claim 5, characterized in that, The process of using a large language model to test the function under test and then outputting the target JSON file includes: The embedding module in the large language model is used to reconstruct the prompts and context of the function to be tested, thereby obtaining semantic vectors and architectural semantic vectors. The target JSON file is obtained by segmenting the data into blocks based on semantic vectors and architectural semantic vectors through similarity mapping.

7. The method according to claim 1, characterized in that, The process of inputting the target JSON file into the credibility evaluation module to obtain a test score includes: A Bayesian probabilistic graphical model is constructed using the credibility assessment module's incorrect responses, large model illusion phenomenon, and repetitive conversations. The reliability of the target JSON file is tested using the reliability formula of the Bayesian probabilistic graphical model to obtain a test score. The reliability formula is as follows: P(Tr|E)=P(E|Tr)×P(Tr)÷P(E); Where Tr represents the credibility state, E represents the evidence set, P() is the probability function, and P(Tr|E) is the fault tolerance weight coefficient.

8. The method according to claim 7, characterized in that, The test result list obtained based on the target JSON file and initial test cases includes: Extract test steps, expected results, and actual test results from the target JSON file; The test results list is obtained by associating and storing the test steps, expected results, and actual test results with the corresponding initial test cases.

9. A software automated testing system, characterized in that, The system includes: The pull module is used to parse and extract comments from C code and models, and then generate detailed design documents based on the comments; it also retrieves software architecture design documents. The initial test case module is used to obtain initial test cases by performing correlation analysis with the detailed design document and software architecture design document, the C code, and the model. The compression module is used to perform structured parsing and semantic compression on the C code and model to obtain a JSON file; the JSON file is an intermediate representation file that can be stably processed by a large language model; The large language module is used to take a JSON file and initial test cases as input to the large language model, and the large language model outputs the target JSON file. The scoring module is used to input the target JSON file into the credibility assessment module to obtain a test score; The generation module generates a list of test results based on the target JSON file and the initial test cases if the test score is greater than or equal to the preset score.

10. An electronic device, characterized in that, include: At least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-8.

Images