An automatic detection method for software requirement document quality based on artificial intelligence and GJB438C

Abstract

This application discloses a method, apparatus, medium, and device for automatic quality detection of software requirements documents based on artificial intelligence and GJB438C. It uses natural language processing technology to parse documents, extract structural elements and text content units, perform multi-dimensional quality characteristic vectorization analysis, and generate a comprehensive quality score and a visual evaluation report, thereby realizing automated and intelligent detection of document quality.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence and data processing technology, specifically to a method, apparatus, electronic device and storage medium for automatic detection of software requirements document quality based on artificial intelligence and GJB438C. Background Technology

[0002] In the software development process, requirements documents (such as software requirements specifications) are the core basis for project initiation, design, development, testing, and acceptance. Their quality directly determines project cost, timeline, and final deliverables. Traditionally, the quality review of requirements documents relies on manual review, which has the following significant drawbacks: Highly subjective and inconsistent standards: The effectiveness of the review depends heavily on the experience and personal understanding of the reviewers, making it difficult to establish unified and objective evaluation standards.

[0003] Inefficient and labor-intensive: When faced with large and complex requirements documents, manual review of each sentence is time-consuming and labor-intensive, and it is difficult to cover all the content.

[0004] It is easy to overlook things and consistency is difficult to guarantee: it is difficult for humans to fully detect logical contradictions, inconsistencies, ambiguities and lack of completeness among requirements.

[0005] Difficulties in knowledge transfer: Excellent review experience is hard to solidify and reuse. In recent years, although some tools have provided simple spell checks or template compliance verification, they lack the ability to intelligently and automatically analyze the deep semantic quality of requirements documents (such as clarity, unambiguity, testability, consistency, and completeness). Therefore, there is an urgent need for an automatic detection method for software requirements document quality based on AI and GJB438C, capable of automatically analyzing software document quality and providing intelligent result analysis and optimization suggestions. Summary of the Invention

[0006] This application discloses an automatic detection method for software requirements documents based on artificial intelligence and GJB438C, which aims to solve problems such as low document quality, low review pass rate, and inability to guide development.

[0007] To achieve the above objectives, the first aspect of this application provides an automatic software requirements document quality detection method based on artificial intelligence and GJB438C, comprising: The system receives an input software requirements document and uses natural language processing technology to parse the document, extracting structural elements and text content units. The structural elements include chapter numbers and chapter names, and the text content units include functional descriptions, non-functional descriptions, interface descriptions, and bidirectional tracking relationships. Based on the obtained structural elements and text content units, a multi-dimensional quality characteristic vectorization analysis is performed on the software requirements document; wherein, the multi-dimensional quality characteristic vectorization analysis includes at least ambiguity analysis, consistency analysis, and testability analysis. Ambiguity analysis is used to calculate the ambiguity of the text content unit by using a pre-trained domain-specific word meaning disambiguation model and a fuzzy word detection rule base. Consistency analysis is used to establish a semantic relationship graph between functional items based on the functional description, and to detect whether there are logical contradictions or descriptive conflicts between different functional items through neural networks and semantic similarity calculation. Testability analysis is used to deeply refine the functional description, identify the input conditions, output results and system behavior based on the semantic understanding model, and determine whether the function is testable. The feature vectors obtained from the multi-dimensional quality characteristic vectorization analysis are input into a comprehensive evaluation model to generate a comprehensive quality score for the software requirements specification document; at the same time, based on the analysis results of each dimension, the text locations and types of problems are accurately marked and modification suggestions are provided. Generate a visualized quality assessment report, which includes an overall quality score, quality radar charts for each dimension, a detailed list of defects, and aggregated improvement suggestions; wherein, the detailed list of defects supports locating the original text.

[0008] Optionally, according to the software requirements document quality automatic detection method based on artificial intelligence and GJB438C described in the first aspect, the multi-dimensional quality characteristic vectorization analysis further includes: Based on the software documentation standardization library and the GJB438C standard, check whether the full text of the software requirements document meets the format requirements, layout requirements, and whether the chapters are consistent with the chapters required by the GJB438C standard.

[0009] Optionally, according to the software requirements document quality automatic detection method based on artificial intelligence and GJB438C described in the first aspect, the multi-dimensional quality characteristic vectorization analysis further includes: Based on the domain knowledge graph and software requirements reuse library, check whether there are any missing functional and non-functional descriptions; wherein, the non-functional descriptions include security, performance, and reliability requirements.

[0010] Optionally, according to the software requirements document quality automatic detection method based on artificial intelligence and GJB438C described in the first aspect, the multi-dimensional quality characteristic vectorization analysis further includes: The interface description is analyzed based on a domain knowledge graph to detect whether the interface elements include interface name, interface type, sender, receiver, interface interaction information, and interface priority.

[0011] Optionally, according to the software requirements document quality automatic detection method based on artificial intelligence and GJB438C described in the first aspect, the multi-dimensional quality characteristic vectorization analysis further includes: Natural language processing technology is used to analyze whether the format of the functional description statement conforms to a preset structured template; wherein, the structured template includes conditions, subjects, constraints, actions, and objects.

[0012] Optionally, according to the software requirements document quality automatic detection method based on artificial intelligence and GJB438C described in the first aspect, the multi-dimensional quality characteristic vectorization analysis further includes: Based on the extracted structural elements of the software development task book and the software requirements specification, the requirement traceability section in the software requirements specification is traversed to check whether all the section numbers and names in the software development task book appear in the requirement traceability section.

[0013] To achieve the above objectives, a second aspect of this application also provides an automatic software requirements document quality detection device based on artificial intelligence and GJB438C, comprising: The parsing module is used to receive the input software requirements document, parse the software requirements document using natural language processing technology, and extract the structural elements and text content units in the document. The vectorization analysis module is used to perform multi-dimensional quality characteristic vectorization analysis on the software requirements document based on the acquired structural elements and text content units; wherein, the multi-dimensional quality characteristic vectorization analysis includes at least ambiguity analysis, consistency analysis and testability analysis; The evaluation module is used to input the feature vectors obtained from the vectorization analysis of the multi-dimensional quality characteristics into a comprehensive evaluation model to generate a comprehensive quality score for the software requirements specification document; at the same time, based on the analysis results of each dimension, it accurately marks the text locations and types of problems and provides modification suggestions. The report generation module is used to generate visual quality assessment reports.

[0014] To achieve the above objectives, a third aspect of this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, when the processor executes the program, it implements the automatic detection method for software requirements document quality based on artificial intelligence and GJB438C as described in the first aspect.

[0015] To achieve the above objectives, the fourth aspect of this application also provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when executed by a processor, the program implements the automatic software requirements document quality detection method based on artificial intelligence and GJB438C as described in the first aspect.

[0016] To achieve the above objectives, the fifth aspect of this application also provides a software requirements document quality inspection system, including an automatic software requirements document quality inspection device based on artificial intelligence and GJB438C as described in the second aspect, and a display terminal communicatively connected to the inspection device. The display terminal is used to receive and display the visual quality assessment report generated by the report generation module, and allows users to directly jump to the corresponding position in the original text for modification by clicking the link in the defect details list.

[0017] This application discloses an automatic quality detection method for software requirements documents based on artificial intelligence and GJB438C. The method includes receiving an input software requirements document; parsing the document using natural language processing technology to extract structural elements and text content units; performing multi-dimensional quality characteristic vectorization analysis on the document based on the obtained structural elements and text content units; inputting the feature vectors obtained from the multi-dimensional quality characteristic vectorization analysis into a comprehensive evaluation model to generate a comprehensive quality score for the software requirements specification document; simultaneously, based on the analysis results of each dimension, accurately identifying the text locations and types of problems and providing modification suggestions; and generating a visualized quality assessment report. The technical advantages of this application are: it can automatically, intelligently, comprehensively, and objectively detect the quality of software requirements documents, improve review efficiency, reduce project risks, and realize the accumulation and reuse of quality assessment knowledge. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating an automatic software requirements document quality detection method based on artificial intelligence and GJB438C, provided as an embodiment of this application.

[0019] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0021] Explanation of key terms: Target data: refers to the software requirements document to be tested, such as the software development task book, software requirements specification, etc.

[0022] Training samples: refer to the labeled data used to train the domain-specific word sense disambiguation model, semantic understanding model, and comprehensive evaluation model, including document fragments and their corresponding quality labels.

[0023] Feature vector: refers to the numerical representation obtained through vectorized analysis of multi-dimensional quality characteristics, used to characterize the quality characteristics of a document in dimensions such as ambiguity, consistency, and testability.

[0024] Model input: refers to the text content units or feature vectors input into the pre-trained domain semantic disambiguation model, semantic understanding model, or comprehensive evaluation model.

[0025] Reasoning results: refer to the quality score, ambiguity value, logical contradiction detection results, or testability judgment results output by the model after processing the input.

[0026] The prompt message refers to the detailed list of defects, problem types, and modification suggestions displayed to the user when generating a visual quality assessment report.

[0027] Constraint information: refers to the format requirements, layout requirements, and chapter consistency requirements of the GJB438C standard for software requirements documents.

[0028] Multimodal data: This application mainly involves textual modality, but in a broader sense it may include fused data of document structure (chapter number) and text content (functional description).

[0029] Edge device / server / terminal: refers to the hardware carrier that executes the method of this application, such as a detection device deployed on the server or a display terminal used by the user.

[0030] Business object / control object: refers to the software system described in the software requirements document, such as a task planning system.

[0031] The common processing flow of existing technologies involves manually reviewing the requirements document sentence by sentence, or using only simple spell-checking tools. The problems with this approach are: strong subjectivity and inconsistent standards; low efficiency, making it difficult to cover large and complex documents; prone to omissions, unable to comprehensively detect logical contradictions, ambiguities, and incompleteness; and difficulty in knowledge transfer.

[0032] Example 1 refer to Figure 1 The first embodiment of this application provides an automatic software requirements document quality detection method based on artificial intelligence and GJB438C to solve the problems of low document quality, low review pass rate, and inability to guide development mentioned in the background art. This method can be executed by a processor, which can be set in a terminal or a server. The execution process of the automatic software requirements document quality detection method based on artificial intelligence and GJB438C can be as follows: Step S101: Receive the input software requirements document, parse the software requirements document using natural language processing technology, and extract the structural elements and text content units in the document; wherein, the structural elements include chapter numbers and chapter names, and the text content units include functional descriptions, non-functional descriptions, interface descriptions, and bidirectional tracking relationships.

[0033] In one embodiment of this application, an input requirements document is received and parsed using natural language processing technology to extract structural elements (chapter numbers, chapter names) and text content (including functional descriptions, non-functional descriptions, interface descriptions, and bidirectional tracking relationships) from the document. For example, if a software development task book and software requirements specification for a "task planning system" are input, a BERT-based model is used for sentence boundary recognition and key information extraction.

[0034] Step S102: Based on the obtained structural elements and text content units, perform multi-dimensional quality characteristic vectorization analysis on the software requirements document; wherein, the multi-dimensional quality characteristic vectorization analysis includes at least ambiguity analysis, consistency analysis and testability analysis.

[0035] In one embodiment of this application, a multi-dimensional quality characteristic vector analysis is performed on the software requirements document based on the obtained structural elements and text content units.

[0036] Furthermore, ambiguity analysis is used to calculate the ambiguity of text content units by using a pre-trained domain-specific word sense disambiguation model and a fuzzy word detection rule base.

[0037] In one embodiment of this application, a pre-trained domain semantic disambiguation model and a fuzzy word detection rule base are used to calculate the fuzziness of the text content units (functional description, non-functional description, and interface description) in step 1. Fuzzy words include, but are not limited to: 1) acceptable, competent; 2) minimum, minimum amount, no more than, not exceeding; 3) best, maximum, maximum; 4) between; 5) effectively; 6) rapid, fast, quick; 7) in most cases, usually, often, almost, always; 8) matching, equal, conforming, the same; 9) possibly, should, should not, will not; 10) several, some, many, few, multiple, large amount; 11) user-friendly, simple, easy. For example, if the words "should" and "as soon as possible" are detected in the functional description "The system should perform route planning as soon as possible after obtaining target information," they are marked as fuzzy words, and the fuzziness score of that sentence is deducted.

[0038] Furthermore, consistency analysis is used to build a semantic relationship graph between functional items based on functional descriptions. Through neural networks and semantic similarity calculations, it detects whether there are logical contradictions or descriptive conflicts between different functional items.

[0039] In one embodiment of this application, a semantic relationship graph between functional items is established based on the functional description in step 1. Through neural networks and semantic similarity calculations, it is detected whether there are logical contradictions, inconsistencies in usage, or conflicts in functional descriptions between different functional items. For example, if a conflict is detected between "when the drone's altitude is below 800 meters, the display interface shows the data in red text" and "the drone's altitude is displayed in real-time in white text," the corresponding score for that requirement document is deducted.

[0040] Furthermore, testability analysis is used to deeply refine the functional description, identify the input conditions, output results, and system behavior based on a semantic understanding model, and determine whether the function is testable.

[0041] In one embodiment of this application, the functional description in step 1 is deeply refined, and the input conditions, output results, and system behavior are identified based on a semantic understanding model to determine whether the function is testable and verifiable. For example, if it is detected that "the system runs stably under high concurrency" without clearly specifying the high concurrency and performance indicators, it is marked as insufficient testability.

[0042] In an optional embodiment of this application, the multi-dimensional quality characteristic vectorization analysis further includes: checking whether the full text of the software requirements document meets the format requirements, layout requirements, and whether the chapters are consistent with the chapters required by the GJB438C standard, based on the software document standardization library and the GJB438C standard.

[0043] In one embodiment of this application, during the normative analysis, it was detected that the software requirements specification of the "Task Planning System" was missing the description of Chapter 3.9 "CSCI Environment Requirements", and the corresponding score for that requirements document was deducted.

[0044] In an optional embodiment of this application, the multi-dimensional quality characteristic vectorization analysis further includes: checking whether there are missing functional and non-functional descriptions based on the domain knowledge graph and the software requirements reuse library; wherein, the non-functional descriptions include security, performance, and reliability requirements.

[0045] In one embodiment of this application, for integrity analysis, if the "Task Planning System" document is found to lack the basic requirement of "Route Evaluation", the corresponding score of the document with that requirement will be deducted.

[0046] In an optional embodiment of this application, the multi-dimensional quality characteristic vectorization analysis further includes: analyzing the interface description based on the domain knowledge graph to detect whether the interface elements include the interface name, interface type, sender, receiver, interface interaction information, and interface priority.

[0047] In one embodiment of this application, for interface element integrity analysis, if the input interface of the threat area is found to lack sender information, the corresponding score of the interface chapter is deducted.

[0048] In an optional embodiment of this application, the multi-dimensional quality characteristic vectorization analysis further includes: using natural language processing technology to analyze whether the format of the function description statement conforms to a preset structured template; wherein, the structured template includes conditions, subjects, constraints, actions, and objects.

[0049] In one embodiment of this application, no erroneous statements were detected during the functional description structure analysis.

[0050] In an optional embodiment of this application, the multi-dimensional quality characteristic vectorization analysis further includes: based on the extracted structural elements of the software development task book and the software requirements specification, traversing the requirement traceability chapter in the software requirements specification, and detecting whether all chapter numbers and names in the software development task book appear in the requirement traceability chapter.

[0051] In one embodiment of this application, no erroneous statements were detected during the tracing analysis.

[0052] Step S103: Input the feature vectors obtained from the multi-dimensional quality characteristic vectorization analysis into a comprehensive evaluation model to generate a comprehensive quality score for the software requirements specification document; at the same time, based on the analysis results of each dimension, accurately mark the text locations and types of problems and provide modification suggestions.

[0053] In one embodiment of this application, the multi-dimensional feature vector obtained in step 2 is input into a comprehensive evaluation model to generate a comprehensive quality score for the software requirements specification document. Simultaneously, based on the analysis results of each dimension, the problematic text locations and problem types are precisely identified, and modification suggestions are provided. For example, based on the aforementioned multi-dimensional features, the comprehensive evaluation model assigns the overall quality of the document a "medium risk" score of 80 points and lists all specific problems.

[0054] Step S104: Generate a visualized quality assessment report. The visualized quality assessment report includes an overall quality score, quality radar charts for each dimension, a detailed list of defects, and aggregated improvement suggestions. The detailed list of defects allows users to navigate to the original text.

[0055] In one embodiment of this application, a visualized quality assessment report is generated. The report includes, but is not limited to: an overall quality score, quality radar charts for each dimension, a detailed list of defects (linked to the original text), a ranking of top risk items, and aggregated improvement suggestions. For example, an HTML report is generated, allowing users to directly access and modify the corresponding sections of the document by clicking links within the report.

[0056] Example 2 To address the aforementioned technical problems, the second embodiment of this application provides an automatic software requirements document quality detection device based on artificial intelligence and GJB438C to solve the same technical problems as the method embodiment. The device 1000 may include the following modules: a parsing module 1001, a vectorization analysis module 1002, an evaluation module 1003, and a report generation module 1004.

[0057] The parsing module 1001 is used to receive the input software requirements document, parse the software requirements document using natural language processing technology, and extract the structural elements and text content units in the document.

[0058] The vectorization analysis module 1002 is used to perform multi-dimensional quality characteristic vectorization analysis on the software requirements document based on the acquired structural elements and text content units; wherein, the multi-dimensional quality characteristic vectorization analysis includes at least ambiguity analysis, consistency analysis and testability analysis.

[0059] The evaluation module 1003 is used to input the feature vectors obtained from the vectorization analysis of multi-dimensional quality characteristics into a comprehensive evaluation model to generate a comprehensive quality score for the software requirements specification document; at the same time, based on the analysis results of each dimension, it accurately marks the text locations and types of problems and provides modification suggestions.

[0060] The report generation module 1004 is used to generate visual quality assessment reports.

[0061] It is understood that this embodiment is a device claim corresponding to the method claim, and therefore this embodiment has all the technical effects of the method claim, which will not be repeated here.

[0062] Example 3 To address the aforementioned technical problems, the third embodiment of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The device is characterized in that, when the processor executes the program, it implements the automatic software requirements document quality detection method based on artificial intelligence and GJB438C provided in the preceding embodiments.

[0063] To address the aforementioned technical problems, the fourth embodiment of this application provides a computer-readable storage medium storing a computer program thereon, characterized in that, when executed by a processor, the program implements the automatic software requirements document quality detection method based on artificial intelligence and GJB438C provided in the preceding embodiments.

[0064] To address the aforementioned technical problems, the fifth embodiment of this application provides a software requirements document quality inspection system, including an automatic software requirements document quality inspection device based on artificial intelligence and GJB438C as provided in the previous embodiments, and a display terminal communicatively connected to the inspection device. The display terminal is used to receive and display the visual quality assessment report generated by the report generation module, and allows users to directly jump to the corresponding position in the original text for modification by clicking the link in the defect details list.

[0065] It is understood that this embodiment is the system authority corresponding to the method authority, and therefore this embodiment has all the technical effects of the method embodiment, which will not be repeated here.

[0066] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for automatic quality detection of software requirements documents based on artificial intelligence and GJB438C, characterized in that, include: The system receives an input software requirements document and uses natural language processing technology to parse the document, extracting structural elements and text content units. The structural elements include chapter numbers and chapter names, and the text content units include functional descriptions, non-functional descriptions, interface descriptions, and bidirectional tracking relationships. Based on the obtained structural elements and text content units, a multi-dimensional quality characteristic vectorization analysis is performed on the software requirements document; the multi-dimensional quality characteristic vectorization analysis includes at least ambiguity analysis, consistency analysis, and testability analysis. Ambiguity analysis is used to calculate the ambiguity of the text content unit by using a pre-trained domain-specific word meaning disambiguation model and a fuzzy word detection rule base. Consistency analysis is used to establish a semantic relationship graph between functional items based on the functional description, and to detect whether there are logical contradictions or descriptive conflicts between different functional items through neural networks and semantic similarity calculation. Testability analysis is used to deeply refine the functional description, identify the input conditions, output results and system behavior based on the semantic understanding model, and determine whether the function is testable. The feature vectors obtained from the multi-dimensional quality characteristic vectorization analysis are input into a comprehensive evaluation model to generate a comprehensive quality score for the software requirements specification document; at the same time, based on the analysis results of each dimension, the text locations and types of problems are accurately marked and modification suggestions are provided. Generate a visualized quality assessment report, which includes an overall quality score, quality radar charts for each dimension, a detailed list of defects, and aggregated improvement suggestions; wherein, the detailed list of defects supports locating the original text.

2. The automatic software requirements document quality detection method based on artificial intelligence and GJB438C according to claim 1, characterized in that, The multidimensional quality characteristic vectorization analysis also includes: Based on the software documentation standardization library and the GJB438C standard, check whether the full text of the software requirements document meets the format requirements, layout requirements, and whether the chapters are consistent with the chapters required by the GJB438C standard.

3. The automatic software requirements document quality detection method based on artificial intelligence and GJB438C according to claim 1, characterized in that, The multidimensional quality characteristic vectorization analysis also includes: Based on the domain knowledge graph and software requirements reuse library, check whether there are any missing functional and non-functional descriptions; wherein, the non-functional descriptions include security, performance, and reliability requirements.

4. The automatic software requirements document quality detection method based on artificial intelligence and GJB438C according to claim 1, characterized in that, The multidimensional quality characteristic vectorization analysis also includes: The interface description is analyzed based on a domain knowledge graph to detect whether the interface elements include interface name, interface type, sender, receiver, interface interaction information, and interface priority.

5. The automatic software requirements document quality detection method based on artificial intelligence and GJB438C according to claim 1, characterized in that, The multidimensional quality characteristic vectorization analysis also includes: Natural language processing technology is used to analyze whether the format of the functional description statement conforms to a preset structured template; wherein, the structured template includes conditions, subjects, constraints, actions, and objects.

6. The automatic software requirements document quality detection method based on artificial intelligence and GJB438C according to claim 1, characterized in that, The multidimensional quality characteristic vectorization analysis also includes: Based on the extracted structural elements of the software development task book and the software requirements specification, the requirement traceability section in the software requirements specification is traversed to check whether all the section numbers and names in the software development task book appear in the requirement traceability section.

7. An automatic software requirements document quality detection device based on artificial intelligence and GJB438C, characterized in that, include: The parsing module is used to receive the input software requirements document, parse the software requirements document using natural language processing technology, and extract the structural elements and text content units in the document. The vectorization analysis module is used to perform multi-dimensional quality characteristic vectorization analysis on the software requirements document based on the acquired structural elements and text content units; wherein, the multi-dimensional quality characteristic vectorization analysis includes at least ambiguity analysis, consistency analysis and testability analysis; The evaluation module is used to input the feature vectors obtained from the vectorization analysis of the multi-dimensional quality characteristics into a comprehensive evaluation model to generate a comprehensive quality score for the software requirements specification document; at the same time, based on the analysis results of each dimension, it accurately marks the text locations and types of problems and provides modification suggestions. The report generation module is used to generate visual quality assessment reports.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the automatic software requirements document quality detection method based on artificial intelligence and GJB438C as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the automatic detection method for software requirements document quality based on artificial intelligence and GJB438C as described in any one of claims 1 to 6.

10. A software requirements document quality inspection system, characterized in that, Includes the software requirements document quality automatic detection device based on artificial intelligence and GJB438C as described in claim 7, and a display terminal that is communicatively connected to the detection device; The display terminal is used to receive and display the visual quality assessment report generated by the report generation module, and allows users to directly jump to the corresponding position in the original text for modification by clicking the link in the defect details list.

Images