A knowledge base and deep learning-based software requirement analysis system and method

Abstract

The application relates to a knowledge base and deep learning-based software requirement analysis system and method, wherein the system comprises: a function list generation module configured to search a requirement analysis knowledge base based on received software function requirement information, and generate a software function list of requirements by a preset large language model under the guidance of a first prompt word template, wherein the software function requirement information represents software functions desired by a user; a function detail generation module configured to search each item of software functions in the function list based on the function list by calling the requirement analysis knowledge base, and generate software function details by refining and assembling the function list by the preset large language model under the guidance of a second prompt word template; and a function requirement generation module configured to generate a function requirement specification by the preset large language model under the guidance of the software function details and a third prompt word template. The application solves the problem of how to improve the generation quality of requirement documents in software requirement engineering.

Description

Technical Field

[0001] This application relates to the field of software development technology, and in particular to a software requirements analysis system and method based on knowledge base and deep learning. Background Technology

[0002] In software engineering and product development processes, the Functional Requirements Document (FRD) is one of the core deliverables in the project initiation phase, serving as a crucial bridge between customer business requirements and executable technical solutions. A high-quality FRD directly impacts the accuracy of subsequent system design, development, and testing, and also determines the quality, timeline, and customer satisfaction of the project delivery. Traditionally, the FRD development process is typically led by the product manager and generally involves three progressive stages: system module list compilation → system function detail definition → functional requirements document writing. While this process is widely adopted, it still faces numerous efficiency and quality challenges in practice.

[0003] In recent years, with the development of artificial intelligence technology, especially the breakthroughs in text generation and semantic understanding by natural language processing (NLP) and deep learning models, some research has attempted to apply them to assist software requirements engineering. However, existing technologies mostly focus on single-stage text polishing or simple question answering, lacking systematic modeling of the entire software requirements lifecycle, and failing to effectively integrate structured historical data from enterprise private knowledge bases, resulting in severely generalized generated content, insufficient professionalism, and poor feasibility.

[0004] Currently, no effective solution has been proposed for the problem of how to improve the quality of requirements documents generated in software requirements engineering. Summary of the Invention

[0005] This application provides a software requirements analysis system and method based on knowledge base and deep learning, which at least addresses the problem of how to improve the quality of requirements documents generated in software requirements engineering in related technologies.

[0006] In a first aspect, embodiments of this application provide a software requirements analysis system based on knowledge base and deep learning, the system including a function list generation module, a function details generation module, and a function requirements generation module; The function list generation module is used to retrieve the first search result by calling the requirement analysis knowledge base according to the received software function requirement information, and to process the first search result through a preset large language model under the guidance of the first prompt word template to generate a software function requirement list. The software function requirement information represents the software functions that the user wants. The function details generation module is used to retrieve the software functions in the list by calling the requirements analysis knowledge base according to the software function list, obtain the second search results, and, under the guidance of the second prompt word template, refine and assemble the second search results through the preset large language model to generate software function details. The functional requirements generation module is used to generate a complete functional requirements specification based on the software function details and guided by the third prompt word template, using the preset large language model.

[0007] In some embodiments, the function list generation module is used to put the received software function requirement information into a problem classifier to determine whether the software function requirement information is within the scope of the requirements analysis knowledge base; if it is within the scope of the requirements analysis knowledge base, the requirements analysis knowledge base is called to perform a search to obtain a first search result.

[0008] In some embodiments, the function list generation module is used to use the software function requirement information as search terms to perform hybrid knowledge retrieval in the requirement analysis knowledge base using the gte-rerank model to obtain a first search result, wherein the retrieval method of the gte-rerank model includes vector similarity retrieval or graph path matching retrieval.

[0009] In some embodiments, the function list generation module is used to process the first search result through a preset large language model to generate a required software function list under the guidance of the first prompt word template. The preset large language model is a DeepSeek-R1-Distill-Qwen-32B deep learning model. The content of the first prompt word template includes the analysis process and result requirements, referenced content, and security specifications.

[0010] In some embodiments, the function details generation module is used to call the requirements analysis knowledge base to retrieve each software function in the software function list one by one to obtain a second retrieval result, wherein the second retrieval result includes knowledge fragments that are retrieved from the knowledge base and correspond one-to-one with each software function.

[0011] In some embodiments, the function detail generation module is used, guided by the second prompt word template, to extract and assemble knowledge fragments of each software function from different dimensions using the preset large language model, so as to generate software function details with a unified language style and complete logic. The preset large language model is a DeepSeek-R1-Distill-Qwen-32B deep learning model. The content of the second prompt word template includes knowledge extraction requirements, output format requirements, inspection requirements, referenced content, and security specifications. The different dimensions of information include function descriptions, business rules, and field lists.

[0012] In some embodiments, the functional requirement generation module is used to assemble prompt words for each software function in the software function details, and combine the prompt word assembly results with the complete prompt word specification to obtain a third prompt word template that corresponds one-to-one with each software function. The complete prompt word specification includes the prompt word assembly results, security specifications, generation template, software functional requirement generation rules, role-playing requirements, processing procedures, and output requirements.

[0013] In some embodiments, the functional requirements generation module is used to generate a complete functional requirements specification through the preset large language model under the guidance of the third prompt word template, wherein the preset large language model is a DeepSeek-R1-Distill-Qwen-32B deep learning model.

[0014] In some embodiments, the first prompt word template, the second prompt word template, and the complete prompt word specification are stored in the requirements analysis knowledge base.

[0015] Secondly, embodiments of this application provide a software requirements analysis method based on a knowledge base and deep learning. The method is executed based on the system described in the first aspect above, and the method includes: Based on the received software functional requirement information, the requirement analysis knowledge base is called to retrieve the first search result. Guided by the first prompt word template, the first search result is processed by a preset large language model to generate a list of software functions corresponding to the requirements. The software functional requirement information represents the software functions that the user wants. Based on the software function list, the requirements analysis knowledge base is invoked to perform a round-robin retrieval of each software function in the list to obtain a second retrieval result. Under the guidance of the second prompt word template, the second retrieval result is refined and assembled through the preset large language model to generate software function details. Each software function detail has the ability to be adjusted and regenerated through the user's natural language interaction. Based on the software function details, guided by the third prompt word template, a complete functional requirements specification is generated through the preset large language model. The software function list, software function details, and functional requirements specification all have the ability to be referenced with one click, edited, and exported as documents.

[0016] Compared to related technologies, this application provides a software requirements analysis system and method based on a knowledge base and deep learning. The system includes: a function list generation module, used to retrieve first search results from a requirements analysis knowledge base based on received software function requirements information, and then process these results using a preset large language model under the guidance of a first prompting word template to generate a list of required software functions, wherein the software function requirements information represents the software functions desired by the user; a function detail generation module, used to retrieve each software function in the list from the requirements analysis knowledge base based on the software function list, obtaining second search results, and then refining and assembling these results using a preset large language model under the guidance of a second prompting word template to generate software function details; and a function requirements generation module, used to generate a complete function requirements specification based on the software function details and guided by a third prompting word template using a preset large language model. This system achieves deep integration of a knowledge base and a deep learning model, effectively integrating the project knowledge assets accumulated by the enterprise, and significantly improving the generation efficiency, accuracy, and standardization of function requirements documents by combining advanced deep learning model capabilities, thus solving the problem of how to improve the quality of requirement documents generated in software requirements engineering. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a structural block diagram of a software requirements analysis system according to an embodiment of this application; Figure 2 This is a schematic diagram of the workflow of the function list generation module according to an embodiment of this application; Figure 3 This is a schematic diagram of the workflow of the functional details generation module according to an embodiment of this application; Figure 4 This is a schematic diagram of the workflow of the functional requirement generation module according to the embodiments of this application; Figure 5 This is a schematic diagram of the internal structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0019] Obviously, the accompanying drawings described below are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, any changes to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0020] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0021] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects are in an "or" relationship. The terms "first," "second," and "third" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0022] This application provides a software requirements analysis system based on knowledge base and deep learning. Figure 1 This is a structural block diagram of a software requirements analysis system according to an embodiment of this application, such as... Figure 1 As shown, the system includes a function list generation module, a function details generation module, and a function requirement generation module; The function list generation module is used to retrieve the first search result by calling the requirements analysis knowledge base based on the received software function requirement information, and then, guided by the first prompt word template, process the first search result through a preset large language model to generate a software function requirement list. The software function requirement information represents the software functions that the user wants. It should be noted that, Figure 2 This is a schematic diagram of the workflow of the function list generation module according to an embodiment of this application, such as... Figure 2 As shown, the system automatically finds the most relevant historical software function combinations by performing vector similarity retrieval or graph path matching within a structured requirements analysis knowledge base. The system sorts these combinations based on their matching degree, generating a complete "software function list" that can be referenced and manually adjusted. For newly added or similar requirements, the system can reasonably expand the list through a knowledge reasoning mechanism to ensure its completeness. Furthermore, this software function list can be directly used for communication and confirmation with the client.

[0023] Specifically, such as Figure 2 As shown, the function list generation module is used to put the received software function requirement information into the problem classifier to determine whether the software function requirement information is within the scope of the requirements analysis knowledge base; if it is within the scope of the requirements analysis knowledge base, the requirements analysis knowledge base is called to retrieve the first search result.

[0024] It should be noted that, as Figure 2 As shown, upon entering the software requirements analysis system page, select the "Generate List" interface corresponding to the function list generation module. Enter the desired software functions in the input box. The server receives the front-end parameters and places the user-inputted software function requirements information as a question into the question classifier (preferably using the qwen2.5-coder-32b-instruct model, with a preferred temperature of 0.7). The model classifies the questions (two classification results). If the software function requirements information falls within the scope of the requirements analysis knowledge base, the knowledge base is used to retrieve the first search result. If the software function requirements information is not within the scope of the requirements analysis knowledge base, a software function list cannot be generated, and suggested information can be returned to the user, such as: "I now understand the project requirements related to the company's social security, employment, arbitration, and medical insurance industries. You can ask me, 'I need a function list for unemployment management software.'"

[0025] like Figure 2 As shown, the function list generation module is used to use software function requirement information as search terms to perform hybrid knowledge retrieval in the requirements analysis knowledge base through the gte-rerank model to obtain the first search result. The retrieval methods of the gte-rerank model include vector similarity retrieval or graph path matching retrieval.

[0026] It should be noted that if the software functional requirements information is within the scope of the requirements analysis knowledge base, then the software functional requirements information is used as the search term. The company's requirements analysis knowledge base is called, and a high-quality and accurate retrieval mode using the gte-rerank model (which preferably has a Score threshold of 0.2 and a TopK of 2) is used to perform hybrid knowledge retrieval (vector similarity retrieval or graph path matching retrieval). The retrieval results are stored in the result Array[Object] parameter, which is the first retrieval result.

[0027] like Figure 2As shown, the function list generation module is used to process the first search result to generate a list of required software functions under the guidance of the first prompt word template and through a preset large language model. The preset large language model is the DeepSeek-R1-Distill-Qwen-32B deep learning model. The content of the first prompt word template includes the analysis process and result requirements, referenced content and security specifications. The first prompt word template is stored in the requirements analysis knowledge base.

[0028] It should be noted that the DeepSeek-R1-Distill-Qwen-32B deep learning model then uses the first prompt word template in the database (this template is the main template for generating the list, which is pre-recorded and stored in the Oracle database) and combines it with the search results to perform intelligent analysis and generate a list of software functions to be returned.

[0029] Preferably, for the software functional requirements information, the first prompt word template, and the software function list, it is worth noting that Table 1 is an example table of software functional requirements information, Table 2 is an example table of the first prompt word template, and Table 3 is an example table of the software function list.

[0030] Table 1

[0031] Table 2

[0032] Table 3

[0033] It's worth noting that, based on user needs, the system can accurately retrieve relevant feature lists from the knowledge base and generate complete and standardized results. This capability is at the level of a senior product manager and can be used directly without adjustment. This result can be used for secondary confirmation with clients and can also serve as input for the next stage to generate feature details. In the demonstration test environment, using a single GPU card, the response time is only 1 minute. Compared to the time it takes for a product manager to manually search and extract information from the knowledge base, the intelligent requirement specification generation efficiency is far superior to manual writing, and it is less prone to information omissions.

[0034] The function details generation module is used to retrieve the software functions in the list by calling the requirements analysis knowledge base, obtain the second search results, and, guided by the second prompt word template, refine and assemble the second search results through a preset large language model to generate software function details. It should be noted that, Figure 3 This is a schematic diagram of the workflow of the functional details generation module according to an embodiment of this application, such as... Figure 3As shown, after the list is confirmed, the system enters the "Function Details Generation" stage. For each software function, if a highly matching historical instance exists in the knowledge base, its function description, business rules, field list, and other information are automatically extracted. If it is a new function or a combination of functions, a finely tuned deep learning model is invoked, combined with contextual information from the knowledge base and standard templates, to generate a function description, input / output, exception handling process, and other content that conforms to enterprise standards. All generated content undergoes consistency verification and terminology standardization to ensure a consistent language style and logical integrity. Furthermore, the generated results can be directly used for internal review or customer confirmation.

[0035] Specifically, the function details generation module is used to retrieve the second search results for each software function in the software function list by calling the requirements analysis knowledge base. The second search results include knowledge fragments that are retrieved from the knowledge base and correspond one-to-one with each software function.

[0036] It should be noted that, as Figure 3 As shown, upon entering the software requirements analysis system page, select the "Generate Function Details" interface corresponding to the function details generation module. Upload the generated software function list to the server using the template provided. The preferred template is an Excel spreadsheet with fields for [Serial Number, Function Name, Function Description]. The server receives the front-end parameters, interprets and understands the user-uploaded file, extracts a list of function module names, stores one data entry per function per row, and marks the generation progress as [Not Generated]. The server then calls the knowledge base one by one according to the module names in the function module name list to perform knowledge retrieval. The knowledge retrieval automatically extracts knowledge fragments that are consistent with or highly similar to the module names.

[0037] The function details generation module, guided by the second prompt word template, extracts and assembles information from different dimensions of knowledge fragments of various software functions using a preset large language model to generate software function details with a consistent language style and complete logic. The preset large language model is the DeepSeek-R1-Distill-Qwen-32B deep learning model. The content of the second prompt word template includes knowledge extraction requirements, output format requirements, inspection requirements, referenced content, and security specifications. The different dimensions of information include function descriptions, business rules, and field lists. The second prompt word template is stored in the requirements analysis knowledge base.

[0038] It should be noted that, as Figure 3As shown, based on the knowledge fragments retrieved above, the module descriptions / rules / table fields are extracted from the knowledge fragments according to the requirements of the second prompt word template using a large language model (preferably the DeepSeek-R1-Distill-Qwen-32B model), and then assembled into functional details as required and returned to the page.

[0039] Preferably, for the software functional requirements information, the first prompt word template, and the software functional list, it is worth noting that Table 4 is an example of the software functional list in Excel format, Table 5 is an example of the second prompt word template, and Table 6 is an example of the software functional details for each software function.

[0040] Table 4

[0041] Table 5

[0042] Table 6

[0043] It's important to note that the generated feature details for each software function are complete and highly usable. By the standards of a senior product manager, these results are ready to use without adjustment, for secondary confirmation with the client, or as input for the next stage of generating the functional requirements specification. In the demonstration test environment, using a single GPU card, the response time for generating a feature detail is only 1 minute. Compared to the time required for a product manager to manually search for and imitate features from a knowledge base, the efficiency of intelligent requirement specification generation is significantly higher than manual writing.

[0044] The functional requirements generation module is used to generate a complete functional requirements specification based on the software function details and guided by the third prompt word template, using a preset large language model.

[0045] It should be noted that, Figure 4 This is a schematic diagram of the workflow of the functional requirement generation module according to the embodiments of this application, such as... Figure 4 As shown, after the functional details of all software functions are confirmed, the system will automatically fill in the content of each functional module in a structured manner according to the enterprise's preset FRD document template, and generate a complete functional requirements specification. Each function in the document includes standard chapters such as business overview, business process, business rules, input and output, realizing "one-click generation and one-click export", which greatly reduces manual typesetting and copy and paste work.

[0046] Specifically, the functional requirements generation module is used to assemble prompts for each software function in the software function details, and combine the prompt assembly results with the complete prompt specification to obtain a third prompt template that corresponds one-to-one with each software function. The complete prompt specification includes the prompt assembly results, security specifications, generation template, software functional requirements generation rules, role-playing requirements, processing procedures, and output requirements. The complete prompt specification is stored in the requirements analysis knowledge base.

[0047] It should be noted that, as Figure 4 As shown, upon entering the software requirements analysis system page, select the "Generate Requirements and Specifications Function Description" interface corresponding to the function details generation module; upload the function requirements list file to the server according to the template in the interface operation area. The template is preferably an Excel spreadsheet with fields for [Function Name, Function Description, Business Rules, Table Fields]; the server receives the front-end parameters, interprets and understands the file uploaded by the user, extracts the list of function module names, stores one data for each function per row, and marks the generation progress as [Not Generated]; the server assembles the prompts one by one according to the module names in the function module name list. Preferably, for software functional requirements information, first prompt word templates, and software function lists, it is worth noting that Table 7 is an example table of prompt word assembly for each software function, and Table 8 is an example table of complete prompt word specifications.

[0048] Table 7

[0049] Table 8

[0050] The functional requirements generation module is used to generate a complete functional requirements specification through a preset large language model under the guidance of the third prompt word template. The preset large language model is the DeepSeek-R1-Distill-Qwen-32B deep learning model.

[0051] It should be noted that, as Figure 4As shown, the assembled prompt word indicator language model (preferably the DeepSeek-R1-Distill-Qwen-32B deep learning model) generates functional requirements specifications as required and returns to the page. Each time a function is generated on the server, it is marked as "Generated". The server checks whether all functions in the module name list have been generated; if all have been generated, it notifies the user that generation is complete. Furthermore, the details of each generated function can be "referenced" to the editing panel with one click, facilitating secondary fine-tuning by the user. It also supports one-click "exporting" of all content from the editing panel to a local Word file.

[0052] The software requirements analysis system based on knowledge base and deep learning provided in this application embodiment connects the three key stages of "functional list → functional details → functional requirements specification" through the functional list generation module, functional details generation module, and functional requirements generation module. This achieves fully automated generation from initial user requirements input to standardized FRD document output, which corresponds to the traditional method in the software industry where functional requirements specifications (FRDs) need to be manually written by product managers. Specifically: First, during the software feature list generation phase, product managers need to search for the functional components of similar systems from the company's past project data based on the client's initial requirements. However, most software companies store their historical project data in different document systems, collaboration platforms, or personal local files, lacking unified structured management and semantic retrieval capabilities. Product managers often need to spend a lot of time manually reviewing, comparing, and organizing this data, which is not only inefficient but also prone to information omissions or misjudgments, leading to inaccurate initial communication and affecting client trust and project progress.

[0053] In this embodiment, the function list generation module addresses the "knowledge fragmentation" challenge in software companies where historical project data is scattered across different departments, systems, and document formats. It designs and implements a structured domain knowledge base covering multi-dimensional information such as function lists, rule logic, database fields, and interaction flows from historical projects. This knowledge base supports cross-project and cross-system semantic retrieval and intelligent recommendation, ensuring accurate recall of similar function modules during the new project requirements analysis phase. This avoids information omissions and reinventing the wheel, improving the completeness and accuracy of requirements analysis. Simultaneously, it employs an architecture that integrates a deep learning model with the structured knowledge base. Leveraging the model's powerful natural language understanding and generation capabilities, and combining this with validated, high-quality historical data from the knowledge base, it ensures that the generated content is not only grammatically correct and logically clear but also conforms to industry standards and company practices. The generated system module list, function details, and requirements specifications are all at the level of a senior product manager, requiring minimal manual modification for direct customer confirmation, development reference, and testing, significantly improving document quality and team collaboration efficiency.

[0054] Secondly, during the detailed feature refinement phase, product managers need to supplement the module list with detailed information such as feature descriptions, business rules, inputs and outputs, data fields, and exception handling. For already implemented features, it is still necessary to repeatedly search through historical documents for copying and adjustments; while for new features, it heavily relies on the product manager's personal experience, abstraction ability, and information integration skills. Due to the lack of a standardized knowledge support system, the quality of content output by different personnel varies, making it difficult to guarantee consistency and completeness, thus increasing the cost of team collaboration and subsequent maintenance.

[0055] The feature details generation module in this application can automatically extract and reuse mature solutions for features that have been implemented in the past, reducing repetitive work. For new feature requirements, the model can make reasonable inferences and generate content based on customer descriptions, industry practices and network knowledge, assisting product managers with less experience to complete high-quality output, reducing reliance on individual abilities and improving the team's overall delivery capabilities and consistency.

[0056] Finally, in the writing stage of the functional requirements specification, each function needs to be formatted according to the template specified by the company or project, including module descriptions, flowcharts, field tables, state machines, etc. This process is highly repetitive and standardized, essentially low-value-added manual labor. Although some document generation tools support template filling, most still require manual input item by item, resulting in low automation and a high risk of typos, formatting errors, or omissions, seriously affecting the professionalism and usability of the document.

[0057] The functional requirements generation module in this embodiment is prone to problems such as typos, inconsistent formatting, missing information, or contradictions when written manually. This invention, through a standardized template-driven and model consistency constraint mechanism, ensures that all generated content maintains a high degree of consistency in terminology, structural organization, and field completeness. This fundamentally reduces the risk of human error, improves the readability, maintainability, and traceability of the requirements document, and provides a reliable basis for subsequent design, development, and testing stages.

[0058] Therefore, this application proposes a software requirements analysis system based on the fusion of a knowledge base and a deep learning model. Addressing the core pain points in the current software industry's Functional Requirements Document (FRD) writing process—such as low efficiency, fragmented information, repetitive work, and unstable quality—it constructs an end-to-end intelligent requirements document generation system. Compared to the traditional model relying on human experience and manual compilation, it automates the entire requirements generation process, significantly improving document output efficiency; it supports the mixed processing of new and old functions, balancing reusability and innovation; and it reduces human error rates, improving the accuracy and consistency of requirements documents. In other words, through the innovative fusion of "knowledge base + deep learning model," an intelligent, automated, and highly available requirements document generation system is constructed. This not only significantly improves the work efficiency and document quality in the software requirements analysis phase but also promotes the transformation and upgrading of software engineering towards an intelligent, knowledge-driven model, possessing significant technical value and industrial application significance.

[0059] Furthermore, this system adopts a modular architecture design, allowing the knowledge base to be continuously iterated and updated as enterprise projects accumulate. The deep learning model also supports incremental training and domain adaptation, enabling continuous optimization of the generated results. The system can be integrated into an enterprise's existing AI application platform or requirement management system, serving as an intelligent auxiliary tool to empower product teams, and possesses strong engineering feasibility and commercialization prospects.

[0060] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0061] This application provides a software requirements analysis method based on knowledge base and deep learning. The method is executed based on the system provided in the above embodiments, and includes the following steps: Based on the received software functional requirement information, the requirement analysis knowledge base is called to retrieve the first search result. Guided by the first prompt word template, the first search result is processed by a preset large language model to generate a list of required software functions. The software functional requirement information represents the software functions that the user wants. Based on the software function list, the requirements analysis knowledge base is called to retrieve each software function in the list, and a second search result is obtained. Guided by the second prompt word template, the second search result is refined and assembled through a preset large language model to generate software function details. Based on the software function details, and guided by the third prompt word template, a complete functional requirements specification is generated through a preset large language model.

[0062] It should be noted that the steps shown in the above process or in the flowchart of the accompanying figures can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0063] This embodiment provides an electronic device including a memory and a processor. The memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0064] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0065] Optionally, the electronic device may further include a processor, memory, network interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a software requirements analysis method based on knowledge base and deep learning. The display screen may be an LCD screen or an e-ink screen. The input device may be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the device's casing, or an external keyboard, touchpad, or mouse.

[0066] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated here.

[0067] Furthermore, in conjunction with the knowledge base and deep learning-based software requirements analysis methods described in the above embodiments, this application embodiment can provide a storage medium for implementation. This storage medium stores a computer program; when executed by a processor, the computer program implements any of the knowledge base and deep learning-based software requirements analysis methods described in the above embodiments.

[0068] In one embodiment, Figure 5 This is a schematic diagram of the internal structure of an electronic device according to an embodiment of this application, such as... Figure 5 As shown, an electronic device is provided, which can be a server, and its internal structure diagram can be as follows. Figure 5As shown, the electronic device includes a processor, a network interface, internal memory, and non-volatile memory connected via an internal bus. The non-volatile memory stores the operating system, computer programs, and a database. The processor provides computing and control capabilities, the network interface communicates with external terminals via a network, the internal memory provides the environment for the operating system and computer programs to run, the computer programs are executed by the processor to implement a software requirements analysis method based on a knowledge base and deep learning, and the database stores data.

[0069] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. A specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0070] 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. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program 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 may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of 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.

[0071] Those skilled in the art should understand that 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 have been 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.

[0072] 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 software requirements analysis system based on knowledge base and deep learning, characterized in that, The system includes a function list generation module, a function details generation module, and a function requirement generation module; The function list generation module is used to put the received software function requirement information into a problem classifier to determine whether the software function requirement information is within the scope of the requirement analysis knowledge base. If it is within the scope of the requirement analysis knowledge base, the software function requirement information is used as a search term to perform a hybrid knowledge retrieval in the requirement analysis knowledge base using a GTE-Rerank model to obtain a first search result. The GTE-Rerank model's retrieval methods include vector similarity retrieval or graph path matching retrieval. Guided by a first prompt word template, the first search result is processed by a preset large language model to generate a list of required software functions. The software function requirement information represents the software functions that the user wants. The function details generation module is used to retrieve each software function in the software function list by calling the requirements analysis knowledge base to obtain a second search result. The second search result includes knowledge fragments that are matched from the knowledge base and correspond one-to-one with each software function. Guided by the second prompt word template, the module extracts and assembles the knowledge fragments of each software function from different dimensions using the preset large language model to generate software function details with a unified language style and complete logic. The different dimensions of information include function description, business rules, and field list. The functional requirements generation module is used to assemble prompt words for each software function in the software function details, and combine the prompt word assembly results with the complete prompt word specification to obtain a third prompt word template that corresponds one-to-one with each software function. The complete prompt word specification includes the prompt word assembly results, security specifications, generation template, software functional requirements generation rules, role-playing requirements, processing procedures, and output requirements. Guided by the third prompt word template, a complete functional requirements specification is generated through the preset large language model.

2. The system according to claim 1, characterized in that, The function list generation module is used to process the first search result under the guidance of the first prompt word template to generate a list of required software functions through a preset large language model. The preset large language model is a DeepSeek-R1-Distill-Qwen-32B deep learning model. The content of the first prompt word template includes the analysis process and result requirements, referenced content, and security specifications.

3. The system according to claim 1, characterized in that, The preset large language model is the DeepSeek-R1-Distill-Qwen-32B deep learning model, and the content of the second prompt word template includes knowledge extraction requirements, output format requirements, inspection requirements, cited content, and security specifications.

4. The system according to claim 1, characterized in that, The functional requirements generation module is used to generate a complete functional requirements specification through the preset large language model under the guidance of the third prompt word template, wherein the preset large language model is the DeepSeek-R1-Distill-Qwen-32B deep learning model.

5. The system according to claim 1, characterized in that, The first prompt word template, the second prompt word template, and the complete prompt word specification are stored in the requirements analysis knowledge base.

6. A software requirements analysis method based on knowledge base and deep learning, characterized in that, The method is performed based on the system according to any one of claims 1 to 5, and the method includes: Based on the received software functional requirement information, the requirement analysis knowledge base is called to retrieve the first search result. Guided by the first prompt word template, the first search result is processed by a preset large language model to generate a list of software functions corresponding to the requirements. The software functional requirement information represents the software functions that the user wants. Based on the software function list, the requirements analysis knowledge base is invoked to perform a round-robin retrieval of each software function in the list to obtain a second retrieval result. Under the guidance of the second prompt word template, the second retrieval result is refined and assembled through the preset large language model to generate software function details. Each software function detail has the ability to be adjusted and regenerated through the user's natural language interaction. Based on the software function details, guided by the third prompt word template, a complete functional requirements specification is generated through the preset large language model. The software function list, the software function details, and the functional requirements specification all have the ability to be referenced with one click, edited, and exported as documents.

Images