Business requirement generation and verification method, device and equipment, and computer program product

Abstract

The application discloses a business demand generation and verification method, device and equipment, and a computer program product. The method comprises the following steps: receiving a business demand writing request of a user; generating a preliminary demand document by using a large language model according to the business demand writing request, a pre-constructed basic knowledge base and a pre-defined demand template; performing duplicate checking on the preliminary demand document by using the basic knowledge base and a demand duplicate checking strategy to obtain a demand duplicate checking result; and performing content checking on the preliminary demand document by using the basic knowledge base, the large language model and a demand content checking strategy according to the demand duplicate checking result to obtain a final demand document. The application deeply integrates AI technology around the whole process of demand writing, constructs a complete link from demand intention generation, content intelligent creation, demand duplicate checking to quality automatic control, improves the quality and efficiency of demand writing, reduces the labor cost, and promotes the development of bank business demand management in the intelligent and automatic direction.

Description

Technical Field

[0001] This application relates to the field of business requirements management technology, and in particular to a method, apparatus and equipment, and computer program product for generating and verifying business requirements. Background Technology

[0002] Since the introduction of articles related to "digital finance," "digitalization and intelligentization" transformation has become a key engine for the reform and development of the banking industry. Especially given the current situation of persistently low interest rates in the banking sector, seizing the opportunities presented by digitalization and intelligentization, and leveraging technology to empower the digital transformation and development of banks, is imperative. In the process of digital transformation, system construction is crucial, and business requirements are the starting point. Their completeness, accuracy, detail, and integration will greatly affect the efficiency, quality, and cost of application software development. Only by properly writing business requirements and controlling quality can a solid foundation be laid for subsequent system development, propelling the digital transformation forward smoothly.

[0003] Analysis of historical business requirements revealed several issues during the requirements gathering process, including low-quality requirements, high learning costs, and poor communication between business and technical teams. These problems negatively impacted product development efficiency and quality, potentially leading to project delays and significantly diminishing the product's market competitiveness. For instance, unclear relationships between requirements and a lack of effective tools or methods to visually represent these relationships made it difficult for the team to fully and accurately understand the requirements. Information asymmetry easily occurred between the business, project, and technical teams during communication and collaboration, resulting in slow decision-making, poor execution, and ultimately affecting overall work efficiency. Summary of the Invention

[0004] This application provides a method, apparatus, and computer program product for generating and verifying business requirements, so as to improve the accuracy and efficiency of business requirement generation and reduce the cost of business requirement generation.

[0005] The embodiments of this application adopt the following technical solutions:

[0006] In a first aspect, embodiments of this application provide a method for generating and verifying business requirements, the method comprising:

[0007] Receive user requests to write business requirements;

[0008] Based on the business requirements, write requests and pre-built basic knowledge bases and predefined requirement templates, and use a large language model to generate preliminary requirement documents;

[0009] The preliminary requirement document is checked for plagiarism using the pre-built basic knowledge base and the preset requirement plagiarism detection strategy to obtain the requirement plagiarism detection results;

[0010] Based on the results of the requirement deduplication check, the preliminary requirement document is checked using the pre-built basic knowledge base, the large language model, and the preset requirement content checking strategy to obtain the final requirement document.

[0011] Optionally, the step of writing requests and pre-built basic knowledge bases and predefined requirement templates based on the business requirements, and generating preliminary requirement documents using a large language model, includes:

[0012] Based on the business requirements, draft requests and reference materials, and combine them with the requirements development specifications to generate a basic requirements draft.

[0013] Based on the requirement type in the request written according to the business requirements, the corresponding modeling strategy is determined, and the corresponding modeling strategy is used to perform business modeling to obtain a large business model;

[0014] Based on the business requirements, the draft request, the reference materials, the basic requirements draft, and the predefined requirements template, the preliminary requirements document is generated using the pre-built basic knowledge base and the business big model.

[0015] Optionally, the business requirement writing request includes basic requirement description information, and the step of generating a preliminary requirement document using a large language model based on the business requirement writing request, a pre-built basic knowledge base, and a predefined requirement template includes:

[0016] Based on the basic requirement description information, the requirements are written using a pre-built basic knowledge base and a large language model to obtain the preliminary requirement document.

[0017] The aforementioned writing assistance includes at least one of rewriting, expansion, abbreviation, polishing, and continuation.

[0018] Optionally, the preset demand retrieval strategy includes at least one of a precise search strategy, an AI search strategy, and an advanced search strategy, and the method for generating and validating business demands further includes:

[0019] Based on the business requirements, the keywords carried in the request are written, and the precise search strategy is used to search the pre-built basic knowledge base to obtain the first search result and display it to the user.

[0020] Based on the business requirements, the request is written with keywords, and the AI ​​search strategy is used to call the large language model to search in the pre-built basic knowledge base to obtain the second search result and display it to the user.

[0021] Based on the business requirements, the request is written with keywords and multi-dimensional filtering conditions. The advanced search strategy is used to search the pre-built basic knowledge base to obtain the third search result and display it to the user.

[0022] Optionally, the step of using the pre-built basic knowledge base and preset requirement deduplication strategy to perform deduplication on the preliminary requirement document to obtain the requirement deduplication results includes:

[0023] The preliminary requirements document is parsed to obtain the parsing result of the preliminary requirements document;

[0024] Based on the parsing results of the preliminary requirement document and the requirement deduplication prevention and optimization library, requirement deduplication is checked to obtain the requirement deduplication results. The requirement deduplication prevention and optimization library is built based on the pre-built basic knowledge base.

[0025] Optionally, the requirement deduplication result includes requirement documents that have passed the requirement deduplication check. The final requirement document is obtained by performing content checks on the preliminary requirement document based on the requirement deduplication result, using the pre-built basic knowledge base, the large language model, and a preset requirement content checking strategy.

[0026] Using the pre-built basic knowledge base and the large language model, the requirement document that has passed the requirement deduplication is subjected to multi-dimensional content inspection to obtain multi-dimensional content inspection results.

[0027] Based on the multi-dimensional content inspection results, the requirement document that passed the requirement deduplication check is corrected using an intelligent error correction and prompting mechanism to obtain the final requirement document.

[0028] Optionally, the method for generating and validating the business requirements further includes:

[0029] Acquire multi-source data within the domain;

[0030] A basic knowledge base is constructed based on multi-source data in the aforementioned field. The basic knowledge base includes at least one of the following: a business requirements knowledge base, a policy base, a salable product base, and a sensitive word base.

[0031] A scenario optimization library is constructed based on the aforementioned basic knowledge base. The scenario optimization library includes a requirement generation optimization library and a requirement deduplication prevention optimization library.

[0032] Secondly, embodiments of this application also provide a business requirement generation and verification apparatus, the business requirement generation and verification apparatus comprising:

[0033] The receiving unit is used to receive users' requests to write business requirements.

[0034] The requirement generation unit is used to write requests and pre-built basic knowledge bases and predefined requirement templates based on the business requirements, and to generate preliminary requirement documents using a large language model.

[0035] The requirement deduplication unit is used to use the pre-built basic knowledge base and preset requirement deduplication strategy to deduplicate the preliminary requirement document and obtain the requirement deduplication result.

[0036] The requirement checking unit is used to check the content of the preliminary requirement document based on the requirement deduplication results, using the pre-built basic knowledge base, the large language model, and the preset requirement content checking strategy, to obtain the final requirement document.

[0037] Thirdly, embodiments of this application also provide an apparatus, comprising:

[0038] A processor; and a memory arranged to store computer-executable instructions, which, when executed, cause the processor to perform the generation and verification method for any of the aforementioned business requirements.

[0039] Fourthly, embodiments of this application also provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements the generation and verification method for any of the aforementioned business requirements.

[0040] The above-mentioned at least one technical solution adopted in the embodiments of this application can achieve the following beneficial effects: The business requirement generation and verification method of the embodiments of this application first receives a user's business requirement writing request; then, based on the business requirement writing request, a pre-built basic knowledge base, and a predefined requirement template, a preliminary requirement document is generated using a large language model; subsequently, the preliminary requirement document is checked for duplication using the pre-built basic knowledge base and a preset requirement duplication checking strategy to obtain the requirement duplication checking result; finally, based on the requirement duplication checking result, the preliminary requirement document is checked for content using the pre-built basic knowledge base, the large language model, and a preset requirement content checking strategy to obtain the final requirement document. The business requirement generation and verification method of the embodiments of this application deeply integrates AI technology around the entire requirement writing process, constructing a complete link from requirement intention generation, intelligent content creation to automated quality control. By leveraging pre-built basic knowledge bases and predefined templates, combined with large language models, intelligent generation of requirement documents was achieved, improving the efficiency and standardization of requirement writing. The requirement deduplication function effectively avoided requirement duplication, saved development resources, and enhanced the scientific nature of requirement management. Meanwhile, requirement content checking comprehensively ensured the quality of requirement documents, reduced the difficulty of manual review, and provided business personnel with a more efficient and convenient intelligent writing experience, powerfully promoting the development of banking business requirement management towards intelligence and automation. Attached Figure Description

[0041] 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:

[0042] Figure 1 This is a schematic diagram illustrating the overall process of generating and verifying a business requirement in an embodiment of this application.

[0043] Figure 2 This is a flowchart illustrating a method for generating and verifying business requirements in an embodiment of this application.

[0044] Figure 3 This is a flowchart illustrating the writing process for one of the requirements in an embodiment of this application;

[0045] Figure 4 This is a schematic diagram of a requirement-based plagiarism detection process in one embodiment of this application;

[0046] Figure 5 This is a schematic diagram of a requirement checking process in an embodiment of this application;

[0047] Figure 6 This is a schematic diagram of the structure of a business requirement generation and verification device according to an embodiment of this application;

[0048] Figure 7 This is a schematic diagram of the structure of a device according to an embodiment of this application. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0050] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0051] Currently, the main research areas related to the automatic generation and quality verification of business requirements in existing technologies are as follows:

[0052] (1) An intelligent enterprise-level demand management system for banks (patent publication number CN108089843A). This invention provides an intelligent and visualized management of the entire lifecycle of demand assessment, writing, review, and modification. It mainly focuses on the intelligent management solution for the entire lifecycle of demand (including demand assessment, review, modification, etc.) and how to manage the bank-wide demand of a multi-level corporate banking system.

[0053] (2) A method and system for generating software innovation requirements based on a large model (Patent Publication No. CN118607510B). This invention provides a method and system for generating software innovation requirements based on a large model. The generation method includes: hierarchical extraction and sampling, model fine-tuning training, text classification and combination, and creativity evaluation and verification. This application can automatically and stably generate understandable and creative requirement texts, solving the problem that existing software requirement engineering is difficult to automatically and stably generate innovative requirement texts, as well as the problems of innovation detection and understandability detection.

[0054] (3) A method, electronic device, and storage medium for checking the conformity of a requirement template (Patent Publication No. CN112733517B). This invention discloses a method, electronic device, and storage medium for checking the conformity of a requirement template. The method for checking the conformity of a requirement template in this invention includes: obtaining dependency structure information of a requirement statement, wherein the dependency structure information includes attribute information of each term in the requirement statement and syntactic dependencies between the terms; comparing the dependency structure information with dependency template structure information in a preset requirement template, and obtaining a comparison result; and using the comparison result as the result of checking the conformity of the requirement statement, thereby improving the quality of the requirement document.

[0055] The existing solutions have the following main drawbacks:

[0056] (1) Low level of automation in requirement generation. The current process of generating banking business requirements often relies on manual sorting and writing, with a low level of automation, especially in the insufficient application of large model technology. For example, in the existing solution, although intelligent management of requirements is achieved through keywords and fingerprint sequences, the advantages of large models in natural language processing and semantic understanding are not fully utilized, which may limit the depth and breadth of requirement understanding. In addition, since large models are not introduced for automated requirement analysis and generation, the system may be inadequate when dealing with complex and ever-changing requirements, requiring more manual intervention and adjustment, which increases management costs and error rate.

[0057] (2) Quality verification of requirements relies on manual labor, which limits accuracy and efficiency. Currently, the quality verification of banking business requirements mostly relies on manual review, which is not only time-consuming and labor-intensive, but also prone to inaccurate results due to human factors. For example, although existing solutions propose an automated method for checking the conformity of requirement templates, in practical applications, manual review and confirmation of the check results are still required. Especially when facing the key points of business requirement writing, such as content checking, business rule checking, sensitive word checking, and function point duplication checking, traditional methods are inadequate.

[0058] (3) Inadequate refined management of requirements. To rapidly advance business iterations, the banking industry often adopts agile development models, emphasizing implementation while neglecting requirements. This results in a lack of unified and standardized management of waterfall and agile requirements across the entire bank. Especially in the current context of vigorously promoting diverse requirement forms such as business modeling, modeling + functional points, functional points, and user stories, the lack of standardized definitions and classification control mechanisms for requirement elements under different models leads to fragmented requirement management, making it difficult to achieve cross-project and cross-team collaboration and reuse. Particularly in the absence of large-scale model-driven semantic-level analysis capabilities, the implicit relationships between requirements cannot be automatically uncovered, leading to incomplete risk assessments during requirement changes, which in turn affects deployment quality and delivery efficiency.

[0059] Therefore, the main technical objective of this application is:

[0060] (1) Construct a requirement capability map, knowledge base, etc., as the basis for requirement generation and deduplication.

[0061] To effectively support the entire requirements management process, this application constructs a requirements capability map and knowledge base system as crucial support for functions such as automatic requirements generation and intelligent deduplication. Through domain- and level-based structured storage and multi-dimensional intelligent retrieval, it can effectively avoid redundant requirements development and improve the efficiency and quality of requirements analysis. Specifically, the requirements capability map serves as an index to business requirements assets, supporting continuous updates and query displays, providing references for requirements writing, evidence for requirements review, and tools for requirements control, thus better leveraging the role of requirements assets. By integrating multi-source data within the industry to construct a basic business requirements knowledge base, it addresses issues such as requirements writing relying on experience, fragmented knowledge, and low adaptability to specific scenarios.

[0062] (2) Using demand management tools as a carrier, fully utilize large model technology, and take high-value business scenarios as the entry point, including demand proposal, demand generation, demand writing assistance, and demand retrieval.

[0063] By leveraging large-scale model technology, requirements can be automatically generated, changing the previous situation where requirements mainly relied on expert experience and incomplete knowledge systems. Specifically, on the one hand, the powerful natural language processing capabilities of large-scale models enable various processing methods for original requirements, including rewriting, expanding, abbreviating, polishing, and continuing; on the other hand, the learning and analysis capabilities of large-scale models on massive amounts of data generate enterprise-level requirements that conform to business scenarios and standards, reducing errors caused by insufficient experience or knowledge limitations when manually writing requirements.

[0064] (3) Coordinate the refined management of enterprise-level requirements across the entire bank, coordinate waterfall and agile requirements management, and classify and refine the management.

[0065] The system comprehensively manages the requirements of both waterfall and agile projects, achieving refined classification and management of enterprise-level requirements across the entire bank through the construction of a requirements writing template system. Utilizing large-scale modeling technology, templates are generated based on different needs such as business modeling, functional points, and user stories to meet the requirements writing requirements in various business scenarios. Simultaneously, users can select different chapters to generate based on their actual needs, such as functional descriptions, non-functional descriptions, and business models, ensuring the completeness and relevance of the requirements documents.

[0066] (4) Demand quality inspection and demand duplication check.

[0067] Utilizing large-scale model technology, we conduct a comprehensive review of user-written requirements and provide modification suggestions, including checking requirement document templates, grammar, compliance, spelling errors, sensitive words, and duplicate requirements. First, we embed anti-duplicate features at the requirement submission and review stages, using multi-dimensional intelligent comparison of new and existing business requirements to prevent redundant construction of functional modules from the outset. Second, we leverage AI capabilities to comprehensively review requirement documents, ensuring accuracy, completeness, compliance, and adherence to requirement templates. Intelligent error correction and suggestions reduce the workload and error rate of manual review, improving the quality and efficiency of requirement writing.

[0068] The key to the technical solution in this application lies in its comprehensive enterprise-level requirements management across the entire bank, based on a full-process requirements management platform. This platform integrates waterfall and agile requirements management, and categorizes and refines requirements management. It deeply integrates large-scale model technology to establish an enterprise-level requirements knowledge base, creating a requirements distribution map. The capabilities of the large-scale requirements model are applied to high-value business scenarios such as requirements expression and requirements writing, continuously improving the quality of business requirements within the bank and forming a virtuous cycle. High-quality requirements assets are used to feed back into the large-scale requirements model, achieving a dual improvement in both requirements and the model.

[0069] like Figure 1 The diagram illustrates the overall process of generating and validating a business requirement in one embodiment of this application. First, to address issues such as fragmented entry points, lengthy operational chains, and the disconnect between AI capabilities and business processes, AI capabilities are integrated into the requirement management tool (i.e., the digital intelligence platform). This allows users to quickly access corresponding AI functions without switching between stages such as requirement writing, retrieval, checking, and anti-duplication, improving operational fluency and convenience. Second, to provide unified basic technical support and enable rapid expansion and reuse of scenarios, the requirement large-scale model intelligent engine currently primarily provides two capabilities:

[0070] 1) Connect the data links and integrate with digital platforms, content management systems, enterprise knowledge management systems, etc., to streamline the entire process of document acquisition, parsing, and data interaction;

[0071] 2) Encapsulate scenario capabilities, build engineering capabilities in accordance with industry standards, and realize customized development of business processing logic for each scenario based on the new core technology framework and Dify intelligent agent platform, and manage scenario interfaces in a unified manner.

[0072] Finally, a two-tiered dynamic knowledge base system of "foundation + scenario" is built to lay a solid knowledge foundation. This addresses issues such as reliance on experience in requirement-based code implementation, fragmented knowledge, and poor adaptability to specific scenarios.

[0073] 1) Mining and integrating multi-source data to build a basic knowledge base, including a business requirements knowledge base, a system base, a product database, a sensitive word database, etc., and deriving a customized scenario optimization database, such as comparing and verifying historical function points in anti-duplicate scenarios.

[0074] 2) Regularly update new business requirement data from the digital platform to form a data-to-knowledge closed loop of "acquisition-processing-storage-application-feedback" to ensure that the two-layer knowledge base remains fresh.

[0075] Specifically, embodiments of this application provide a method for generating and validating business requirements, such as... Figure 2 The diagram illustrates a flowchart of a method for generating and verifying business requirements according to an embodiment of this application. The method includes the following steps S210 to S240:

[0076] Step S210: Receive the user's request to write business requirements.

[0077] In banking and other business scenarios, users can trigger a business requirement writing request by inputting information such as function descriptions, basic requirements, or requirement keywords. This information serves as the foundational input for subsequent requirement generation and validation, providing crucial clues for the system to understand user needs. For example, a function description could be a user's specific idea for a particular banking business function, such as "implementing a fast mobile banking transfer function, supporting real-time arrival and multiple transfer methods"; basic requirements could be the original requirements provided by the user; and requirement-related keywords help the system quickly locate and associate relevant knowledge.

[0078] Step S220: Based on the business requirements, write the request and the pre-built basic knowledge base and predefined requirement template, and use the large language model to generate a preliminary requirement document.

[0079] Based on the received business requirements, a request is written, and combined with a pre-built basic knowledge base and predefined requirement templates, a preliminary requirement document is generated using a large language model.

[0080] The basic knowledge base integrates multi-source data within the bank, including business policies, historical requirement data, and user feedback, providing rich knowledge support for requirement generation. For example, when generating requirement documents for bank loan business, the basic knowledge base can provide relevant loan policies, descriptions of requirements for similar past loan products, and other information.

[0081] The requirement template system encompasses various types of requirement templates, enabling fine-grained classification and management of enterprise-level requirements across the entire bank. Based on different business scenario requirements, the large language model can automatically select appropriate templates to generate standardized requirement content, and supports users in selecting different chapters, such as functional descriptions, non-functional descriptions, and business models. For example, there are corresponding templates for agile development projects and traditional waterfall model projects to meet different development needs.

[0082] Step S230: Use the pre-built basic knowledge base and preset requirement deduplication strategy to perform deduplication on the preliminary requirement document to obtain the requirement deduplication result.

[0083] The pre-built basic knowledge base and preset requirement deduplication strategies are used to check for duplicates in the initially generated requirement documents. Anti-duplicate functions are embedded in key stages such as requirement proposal and review. By intelligently extracting key content from the requirement documents and combining vector semantic comparison and keyword weighting techniques, new requirements are compared with historical requirements in the basic knowledge base. For example, information such as business domains, components, and functionalities in new requirements are intelligently compared with existing requirements to effectively identify potential duplicate requirements, preventing redundant system functionality from the source and improving the standardization and scientific nature of requirement management.

[0084] Step S240: Based on the requirement deduplication results, the preliminary requirement document is checked for content using the pre-built basic knowledge base, the large language model, and the preset requirement content checking strategy to obtain the final requirement document.

[0085] Based on the results of the requirement plagiarism check, the preliminary requirement documents are further checked using a pre-built basic knowledge base, a large language model, and preset requirement content checking strategies. To address the issues of insufficient standardization and inconsistent quality in existing requirement content, a requirement content checking function is embedded in the requirement generation stage. Through the AI ​​capabilities of the large language model, requirement documents are checked for typos, grammar, and semantics; sensitive words are checked by comparing them with an industry-wide sensitive word database; requirement template compliance is checked based on different template categories; and requirement business logic is checked based on an understanding and analysis of the industry's enterprise knowledge base and various business rule bases. Simultaneously, intelligent error correction and prompting functions guide writers to correct non-standard or erroneous expressions, achieving automated checking of requirement documents from multiple sources (such as online writing, uploading, and large model generation), helping to improve the quality and efficiency of requirement writing, and reducing the difficulty of manual review of requirement content.

[0086] This application's embodiments deeply integrate AI technology throughout the entire requirements writing process, constructing a complete chain from requirements intent generation and intelligent content creation to automated quality control. By utilizing a pre-built basic knowledge base and predefined templates, combined with a large language model, it achieves intelligent generation of requirements documents, improving the efficiency and standardization of requirements writing; the requirements deduplication function effectively avoids requirement duplication, saves development resources, and enhances the scientific nature of requirements management; while requirements content checking comprehensively ensures the quality of requirements documents, reduces the difficulty of manual review, and provides business personnel with a more efficient and convenient intelligent writing experience, powerfully promoting the development of banking business requirements management towards intelligence and automation.

[0087] In some embodiments of this application, the step of generating a preliminary requirement document using a large language model based on the business requirements, a pre-built basic knowledge base, and a predefined requirement template includes: generating a basic requirement draft based on the business requirements and reference materials, combined with requirement development specifications; determining the corresponding modeling strategy based on the requirement type in the business requirements writing request, and performing business modeling using the corresponding modeling strategy to obtain a large business model; and generating the preliminary requirement document using the pre-built basic knowledge base and the large business model based on the business requirements writing request, the reference materials, the basic requirement draft, and the predefined requirement template.

[0088] like Figure 3 The diagram illustrates a requirement writing process in one embodiment of this application. Upon receiving a user's request to write business requirements, the system retrieves existing business knowledge and user experience issues collected from various channels based on the user's input of functional descriptions and research data. It then combines these with the requirement writing specifications to generate a basic requirement draft. Furthermore, for different requirement types, a model generation strategy is developed to form a pilot optimization program integrating business and technology.

[0089] After determining the modeling strategy, business modeling is carried out using this strategy. First, based on business modeling guidelines and past modeling experience, the modeling approach is streamlined. Following this logic, the overall business model's thought process is initialized, giving the model basic capabilities. Next, a 1-4 level framework for business modeling is formed, including: Level 1: Business Domain (e.g., retail banking, corporate banking); Level 2: Value Stream (in retail banking, this might include customer acquisition value stream, product sales value stream, etc.); Level 3: Activity Flow Diagram and Business Scenarios (taking customer acquisition value stream as an example, including marketing campaigns, customer registration campaigns, etc.); and Level 4: Task Definition List (specific tasks within each activity scenario, such as advertising placement tasks in a marketing campaign). After completing the first four levels of the framework, business parameters are further identified, and Level 5 modeling is completed.

[0090] After completing the business modeling framework, the main review rules are outlined according to the process model and related parameter modeling content. For example, review rules may include aspects such as the rationality of the business process and the accuracy of business parameters. Simultaneously, corresponding regulations, operation manuals, functional requirements, and modeling results are prepared; these will serve as input and output products for model training. For instance, existing business regulations are used as input for model training, allowing the model to learn the norms and requirements within those regulations; the business framework and task list generated by the modeling process are used as output products for verification and adjustment.

[0091] During the requirement expression phase, the large-scale model can analyze business data and combine it with market and business trends to provide forward-looking and feasible suggestions for requirement documents. The constructed requirement writing template system covers various types, and can automatically select appropriate templates to generate standardized requirement content based on different business scenario requirements. It also supports users in selecting different chapters to generate, such as functional descriptions, non-functional descriptions, and business models. For example, for the upgrade requirements of a bank's core business system, the large-scale model can generate preliminary requirement documents based on templates, including detailed content such as system performance improvements (non-functional requirements) and the addition of new business functions (functional requirements).

[0092] This application's embodiments achieve refined classification management and intelligent processing of enterprise-level requirements across the entire process by deeply integrating large-scale modeling technology into the entire requirements writing process. From the generation of basic requirements drafts to business modeling, and then to the generation of preliminary requirements documents based on templates and large-scale models, the entire process fully utilizes pre-built basic knowledge bases and advanced modeling technologies. This not only improves the efficiency and standardization of requirements generation but also provides personalized requirements solutions based on different business scenarios and requirements types. Simultaneously, the analysis and suggestion functions of the large-scale model at the requirements intention stage provide forward-looking guidance for business development. Furthermore, the integration of intelligent modeling and requirements generation further enhances the level of automation and intelligence, effectively solving problems such as low automation and poor quality in requirements generation, as well as poor business-technical communication in existing technologies, thus providing strong support for the digital transformation of banking operations.

[0093] In some embodiments of this application, the business requirement writing request includes basic requirement description information. The step of generating a preliminary requirement document using a large language model based on the business requirement writing request, a pre-built basic knowledge base, and a predefined requirement template includes: using the pre-built basic knowledge base and a large language model to assist in requirement writing based on the basic requirement description information, thereby obtaining the preliminary requirement document; wherein, the requirement assistance includes at least one of rewriting, expansion, abbreviation, polishing, and continuation.

[0094] This application supports a requirement writing assistance function, enabling various processing methods for original requirements, including rewriting, expansion, abbreviation, polishing, and continuation. By connecting to an industry knowledge base, it makes requirement writing more aligned with business scenarios.

[0095] (1) Rewriting: The large language model re-expresses and rewrites the basic requirement description information based on the professional knowledge in the basic knowledge base. For example, if the basic requirement is described as "allowing users to easily check their account balance", the large language model will rewrite it as "providing users with a convenient and efficient account balance query function to ensure that users can obtain account fund information at any time and accurately", making the expression more professional and standardized.

[0096] (2) Expansion: The model delves deeper into the business details behind the basic requirement description and expands and supplements them. Taking the basic requirement "support user transfer" as an example, the expanded content may include "support real-time transfer between different accounts within the same bank, as well as fast transfer business across banks, while providing a transfer record query function to make it convenient for users to understand the transfer status at any time", making the requirement description more complete.

[0097] (3) Abbreviation: For lengthy requirement descriptions, the large language model extracts key information and simplifies and compresses it. For example, a complex requirement description, "When a user makes a purchase of a financial product, the system needs to verify the user's identity information, including username, password, verification code, etc. After successful verification, a list of available financial products is displayed. After the user selects a product, fills in the purchase amount and other information, and finally submits the purchase application," can be abbreviated to "When a user purchases a financial product, the system verifies the user's identity and displays a list of products. The user selects a product, fills in the information, and submits the application," highlighting the core points.

[0098] (4) Polishing: Optimize the text of the requirements description to improve readability and professionalism. For example, polish "This system should be able to process business quickly" to "The system should have efficient processing capabilities and be able to respond quickly and complete various business operations".

[0099] (5) Continue writing: Based on the existing requirements, continue logically to fully supplement the requirements. If the existing requirement is "develop a login function for a mobile banking APP", the continued content can be "the login function should support multiple login methods, such as fingerprint login, facial recognition login and password login, and also have a password retrieval function when forgotten".

[0100] Throughout the entire requirement writing process, the large language model is deeply integrated with the enterprise's internal knowledge base system, combined with professional knowledge such as business process specifications. For example, when expanding requirements, business process standards in the knowledge base are referenced to ensure that the added business details conform to the actual business operation process; when polishing the text, professional terminology and expression standards in the knowledge base are used to ensure that the generated requirement document accurately matches the actual business scenario requirements.

[0101] After the above-mentioned requirement writing process, the large language model integrates the processed content to generate a preliminary requirement document. This document, based on accurate language expression, has a high degree of business relevance and practicality.

[0102] This application's embodiments utilize a pre-built basic knowledge base and a large language model for requirement writing assistance, enabling diversified processing of basic requirement description information. Functions such as rewriting, expansion, abbreviation, polishing, and continuation can meet the optimization needs of different types of requirements, making requirement descriptions more professional, complete, concise, readable, and coherent. Simultaneously, deep integration with the enterprise's internal knowledge base system ensures that the generated requirement documents closely align with actual business scenarios, effectively improving the quality and practicality of requirement documents. This provides accurate and reliable data for subsequent business development, contributing to improved efficiency and quality in banking system development and promoting the digital transformation of banking operations.

[0103] In some embodiments of this application, the preset requirement retrieval strategy includes at least one of a precise search strategy, an AI search strategy, and an advanced search strategy. The method for generating and verifying business requirements further includes: using the precise search strategy to search the pre-built basic knowledge base based on the keywords carried in the request for writing the business requirements, obtaining a first search result and displaying it to the user; using the AI ​​search strategy to call the large language model to search the pre-built basic knowledge base based on the keywords carried in the request for writing the business requirements, obtaining a second search result and displaying it to the user; and using the advanced search strategy to search the pre-built basic knowledge base based on the keywords and multi-dimensional filtering conditions carried in the request for writing the business requirements, obtaining a third search result and displaying it to the user.

[0104] The preset demand retrieval strategies in this application embodiment may include precise search strategies, AI search strategies, and advanced search strategies, etc. The specific implementation methods of each strategy are as follows:

[0105] (1) Precise search strategy

[0106] Upon receiving keywords included in a business requirement writing request, a precise search strategy can be initiated. This strategy, based on keyword matching principles, performs a comprehensive search within a pre-built basic knowledge base. For example, if a user enters the keyword "user login failure processing," the system quickly scans all documents in the basic knowledge base, identifies all documents containing that keyword, and displays these documents as the first search result in a list format. Simultaneously, matching keyword content is highlighted within the documents, allowing users to directly locate the target information without reading the full text. Users can subsequently choose to view the complete original document or directly reference relevant content snippets, depending on their needs. When referencing content snippets, the system automatically indicates the source, such as "[Source: Payment System Business Requirements V2.1]," thereby preventing users from repeatedly writing similar content and improving content reuse rates.

[0107] (2) AI search strategy

[0108] Based on business requirements, keywords are written in the request to invoke an AI search strategy. This strategy enables a large language model to perform retrieval tasks within a pre-built basic knowledge base. The large language model doesn't simply match keywords; instead, it deeply understands the semantic intent of the user's question. For example, if a user asks a natural language question like "How to optimize the user login experience," the large language model analyzes the semantics behind the question, searches for relevant documents in the basic knowledge base, summarizes and refines the content of these documents, and forms an introduction to the user's desired needs or system functions, which is then presented to the user as a secondary search result. Users can also view the original materials or reference relevant content as needed, helping them quickly understand complex requirement logic or obtain a general description of system functions.

[0109] (3) Advanced search strategies

[0110] Based on business needs, keywords in the request and multi-dimensional filtering conditions (such as domain, department, system) set by the user are used to perform searches in a pre-built basic knowledge base using advanced search strategies. For example, in a cross-departmental collaboration scenario, the product manager sets conditions such as "domain = company," "department = investment banking," and "system = core system." Documents in the basic knowledge base are filtered according to these conditions, and the filtered documents are sorted in descending order based on keyword matching. Documents that meet the conditions are then presented to the user as a categorized document list as the third search result. During the project review phase, testers set combined conditions such as "domain = liquidation," "department = finance," and "system = liquidation system." The same process is used for searching and sorting, helping users accurately locate target documents, especially quickly finding key documents.

[0111] To facilitate understanding of the above embodiments, Table 1 below briefly summarizes the functions and application scenarios of different search strategies. In specific applications, the appropriate search strategy can be flexibly selected according to the needs of the scenario.

[0112] Table 1

[0113] Dimension Precise Search AI Search Advanced Search Input method Keywords / phrases Natural Language Problems Keyword + Multi-condition Filtering Core competencies Exact match Semantic understanding and summary Multi-dimensional precise filtering Output format Document list + paragraph highlighting Generate summary + reference documents List of categorized documents (sorted by weight) Best scenario Find specific terms or codes Quickly understand complex logic Precise cross-system / department search

[0114] This application provides three different requirement retrieval strategies—precise search, AI search, and advanced search—to meet the diverse retrieval needs of users during the business requirement writing process. The precise search strategy, with its accuracy and efficiency, helps users quickly find specific terms, phrases, or code snippets, achieving precise searching, content reuse, and efficient collaboration. The AI ​​search strategy, leveraging the semantic understanding capabilities of large language models, enables users to quickly understand complex requirement logic and obtain a general description of system functions. The advanced search strategy, through multi-dimensional filtering and keyword matching ranking, allows users to accurately locate target documents when faced with a large number of documents, playing a particularly important role in complex scenarios involving cross-systems and cross-departmental collaboration. These three strategies work together to greatly improve the efficiency and accuracy of requirement retrieval, providing strong support for the generation and verification of business requirements, and promoting the intelligent and efficient development of banking business requirement management.

[0115] In some embodiments of this application, the step of using the pre-built basic knowledge base and the preset requirement deduplication strategy to check the preliminary requirement document for deduplication and obtain the requirement deduplication result includes: parsing the preliminary requirement document to obtain the parsing result of the preliminary requirement document; and performing requirement deduplication based on the parsing result of the preliminary requirement document and the requirement deduplication prevention and optimization library to obtain the requirement deduplication result, wherein the requirement deduplication prevention and optimization library is built based on the pre-built basic knowledge base.

[0116] like Figure 4 The diagram illustrates a requirement deduplication process in one embodiment of this application. The initially generated requirement document undergoes chapter parsing to obtain the parsing results. The requirement deduplication prevention optimization library is built upon a pre-constructed basic knowledge base. Functional point knowledge and key descriptive information are obtained from the full business requirement knowledge base (i.e., the basic knowledge base). This information covers various functional points from past business requirements within the industry, along with their detailed descriptions, providing rich reference material for requirement deduplication. By integrating this information, a dedicated optimization library for requirement deduplication prevention is constructed to improve fuzzy semantic matching capabilities and expand the coverage of deduplication prevention.

[0117] Based on the parsing results of the preliminary requirements document and the constructed requirements deduplication optimization library, requirements deduplication is checked. Unlike traditional comparison algorithms, this application's implementation has been upgraded. On one hand, it supports intelligent extraction of key content from online rich text or attachments, such as extracting core business areas, components, and functionalities from uploaded Word documents, PDF files, and other attachments. On the other hand, it combines vector semantic comparison and keyword weighting technology to calculate the similarity between the extracted content and the information in the requirements deduplication optimization library. In this way, it compares with historical assets to output Top N similar documents, while filtering out false positives for empty templates. For example, when calculating similarity, different weights are assigned to keywords in the documents to more accurately measure the degree of similarity between documents, thereby effectively identifying potential requirements duplication issues.

[0118] This application embodiment comprehensively analyzes the preliminary requirements document and, combined with a requirements deduplication optimization library built on a basic knowledge base, employs innovative methods such as intelligent content extraction, vector semantic comparison, and keyword weighting for requirements deduplication detection. Compared with traditional deduplication methods, it can not only extract key information more accurately from various document formats, but also effectively improve fuzzy semantic matching capabilities, expand the scope of deduplication prevention, avoid false positives for empty templates, identify and prevent requirement duplication issues from the source, prevent redundant construction of system functions, save development resources, and significantly improve the standardization and scientific nature of requirements management, providing a strong guarantee for the stable and efficient development of banking business systems.

[0119] In some embodiments of this application, the requirement deduplication result includes requirement documents that have passed the requirement deduplication check. The step of performing content checks on the preliminary requirement documents based on the requirement deduplication result using the pre-built basic knowledge base, the large language model, and a preset requirement content checking strategy to obtain the final requirement document includes: performing multi-dimensional content checks on the requirement documents that have passed the requirement deduplication check using the pre-built basic knowledge base and the large language model to obtain multi-dimensional content check results; and correcting the requirement documents that have passed the requirement deduplication check using an intelligent error correction and prompting mechanism based on the multi-dimensional content check results to obtain the final requirement document.

[0120] like Figure 5 The diagram illustrates a requirement checking process according to an embodiment of this application. The specific process for checking the content of requirement documents that have passed requirement deduplication is as follows:

[0121] (1) Multi-dimensional content inspection

[0122] Rich text preprocessing layer: When the required document (rich text input) enters the system, it first enters the rich text preprocessing layer. This layer performs format standardization on the document, converting it into a uniform format for easier subsequent processing; it also provides input support to ensure the document can smoothly enter the next processing stage; and it has an error tolerance mechanism to handle some formatting errors or abnormal situations that may exist in the document.

[0123] Slicing layer: The preprocessed document enters the slicing layer, where semantic units are sliced, breaking the document down into unit fragments with independent semantics, in preparation for subsequent in-depth analysis.

[0124] Engineering code processing layer and large model deep analysis layer:

[0125] Regular expressions and required section checks: At the engineering code processing layer, using regular expressions and other technologies, based on the different categories of templates built into the system, such as business modeling, modeling + function points, function points, and user stories, we check whether the document contains required section content, thus realizing the requirement template compliance check.

[0126] Sensitive word filtering: The document content is compared with the inline sensitive word database. Sensitive words in the document are identified through the sensitive word filtering operation, thus completing the sensitive word check.

[0127] Large-scale model in-depth analysis: Large language models (such as the DeepSeek large model) leverage their general capabilities to perform typos, grammar, and semantic checks on the sliced ​​semantic units. Simultaneously, based on understanding and analysis of industry-wide enterprise knowledge bases and business rule bases for various lines of business, they check the accuracy and completeness of the current requirement's business logic, achieving requirement business logic checks. Furthermore, the large model also uses prompt word engineering to accurately locate and highlight grammatical errors and typos in the document.

[0128] Results Fusion Layer: This layer merges the results obtained from the different inspection methods mentioned above, while also considering asynchronous response scenarios, to form a multi-dimensional content inspection result.

[0129] (2) Document correction

[0130] Based on the results of multi-dimensional content checks, intelligent error correction and prompting mechanisms are used to revise requirement documents that have passed the requirement plagiarism check. The intelligent error correction function clearly points out typos, grammatical errors, non-compliance with template requirements, and unreasonable business logic in the document, and provides corresponding correction suggestions. For example, for typos, it directly provides the correct words; for business logic problems, it provides optimization ideas and directions by combining industry knowledge bases and business rule bases. Based on these prompts, the writer can choose to accept revisions item by item or accept revisions with one click to modify and improve the requirement document, ultimately obtaining a high-quality requirement document that meets the requirements.

[0131] This application's embodiments embed a requirement content inspection function during the requirement generation stage, utilizing a multi-layered processing architecture and multi-dimensional inspection methods to conduct a comprehensive and detailed inspection of requirement documents. From rich text preprocessing to result fusion, it covers multiple inspection dimensions, including format processing, semantic segmentation, template compliance, sensitive words, and business logic, fully leveraging the versatility of the large language model and its ability to understand and analyze business knowledge. The intelligent error correction and prompting mechanism can guide writers to correct non-standard or erroneous expressions in a timely and accurate manner, achieving automated inspection of requirement documents from multiple sources. This effectively solves the problems of insufficient standardization and inconsistent quality of existing requirement content, significantly reducing the difficulty of manual review of requirement content, improving the quality and efficiency of requirement writing, and providing strong support for the stable operation and continuous optimization of banking business systems.

[0132] In some embodiments of this application, the method for generating and verifying business requirements further includes: acquiring multi-source data within the domain; constructing a basic knowledge base based on the multi-source data within the domain, the basic knowledge base including at least one of a business requirement knowledge base, a policy base, a salable product base, and a sensitive word base; and constructing a scenario optimization library based on the basic knowledge base, the scenario optimization library including a requirement generation optimization library and a requirement anti-duplicate optimization library.

[0133] This application embodiment constructs a two-layer dynamic knowledge base system, and the specific implementation steps are as follows:

[0134] First, we acquire multi-source data within the field from multiple channels. These data sources are extensive, including business requirement data generated by internal business systems, various policy documents, information on available products, as well as relevant industry trends and user feedback from external sources. For example, we obtain historical business requirement documents from the bank's core business systems, the latest financial regulatory policy documents from the legal department, detailed introductions of available products from the marketing department, and user feedback from customer service channels.

[0135] Based on the acquired multi-source data within the domain, a basic knowledge base containing various types of libraries is constructed:

[0136] 1) Business Requirements Knowledge Base: The collected business requirements data is organized, classified, and stored to form a business requirements knowledge base. This knowledge base covers detailed information on various past business requirements of the bank, providing rich reference cases for subsequent requirement generation and verification.

[0137] 2) Policy Library: This library parses and inputs relevant internal and external policy documents to create a policy library. The content in the policy library provides a compliance basis for the development of business requirements, ensuring that requirements comply with the requirements of various rules and regulations.

[0138] 3) Available Products Database: Organizes and stores information on the bank's available products, including product features, functions, and applicable customer groups, providing accurate information support for product-related content in the writing of requirements.

[0139] 4) Sensitive Word Database: Collect and organize various sensitive words to form a sensitive word database. During the content review phase of requirements documents, the sensitive word database can be used to quickly screen for sensitive information in the documents, ensuring the compliance of the requirements documents.

[0140] Based on the established basic knowledge base, we will further build a scenario optimization library, including a requirement generation optimization library and a requirement anti-duplicate optimization library.

[0141] 1) Requirements Generation and Optimization Library: Combining different business scenarios and requirement types, this library utilizes information from a basic knowledge base to develop corresponding requirements generation strategies and templates. For example, different requirements generation strategies are developed for agile development projects and traditional waterfall model projects, and stored in the requirements generation and optimization library for use and optimization based on actual conditions during the requirements generation process.

[0142] 2) Demand Deduplication Prevention and Optimization Library: This library is constructed by extracting historical functional point information and key descriptions from the basic knowledge base. During demand deduplication checks, the information in this library is used to employ techniques such as intelligent key content extraction, vector semantic comparison, and keyword weighting to improve fuzzy semantic matching capabilities, expand the scope of deduplication prevention, and effectively identify potential demand duplication issues.

[0143] Regularly acquire new business requirement data and user feedback information from digital platforms (such as business requirement management platforms and user feedback platforms), process and store this new data according to the above construction process, and update it into the basic knowledge base and scenario optimization library, forming a data-to-knowledge closed loop of "acquisition-processing-storage-application-feedback" to ensure that the two-layer knowledge base remains fresh and can reflect the latest changes and needs in the business domain in a timely manner.

[0144] This application's embodiments effectively address issues such as reliance on experience in requirement writing, fragmented knowledge, and low adaptability to specific scenarios by constructing a two-tiered dynamic knowledge base system of "foundation + scenario." The foundational knowledge base integrates multi-source data, providing comprehensive knowledge support for requirement generation and verification; the scenario optimization library is customized and optimized according to different business scenarios, improving the quality of requirement generation and the accuracy of requirement deduplication. Simultaneously, by regularly updating and maintaining the knowledge base, a closed loop of data-to-knowledge transformation is formed, ensuring that the knowledge base remains constantly updated and keeps pace with changes in the business domain. This provides efficient, accurate, and sustainable support for the generation and verification of banking business requirements, powerfully promoting the digital transformation and intelligent development of banking operations.

[0145] In addition, some alternative implementation schemes in the aforementioned embodiments are described below:

[0146] (1) In terms of the standardization and accumulation of demand assets, in addition to building an enterprise-level knowledge base system, intelligent association and recommendation of demand assets can also be achieved through semantic tag system and graph technology.

[0147] (2) The DeepSeek large model technology used in this application can also be replaced by other AI models with the same semantic understanding and text generation capabilities, such as Tongyi Qianwen, Hunyuan large model or Wenxin Yiyan, as long as they have the same core functions of natural language processing, demand text generation and compliance verification, they can achieve the same technical effect.

[0148] (3) For the automated generation of requirement documents, the traditional method based on rule engine and template matching can also be used as an alternative. Although the level of intelligence is slightly lower, it can still achieve efficient output in the case of clear structured input.

[0149] (4) For the deduplication and association analysis of requirements, in addition to similarity calculation based on semantic vectors, keyword matching, fingerprint algorithm or knowledge reasoning method based on ontology can also be used.

[0150] In summary, the key points of this application are mainly as follows:

[0151] (1) Achieve comprehensive intelligent coverage of demand-driven content creation.

[0152] By deeply integrating AI technology capabilities around the key nodes of the entire process of writing requirements (before writing, during writing, and after writing), a complete link is built from the generation of requirements intentions and intelligent content creation to automated quality control, realizing intelligent coverage of the entire process of requirements content creation and providing business personnel with a more efficient and convenient intelligent writing experience.

[0153] (2) Establish a comprehensive enterprise-level knowledge base system and form a demand distribution map.

[0154] This key point integrates multi-source data within the industry, covering business policies, historical requirement data, and user feedback, to build a dynamically updated knowledge base. The requirement distribution map is a crucial presentation format of the knowledge base. Based on business requirements and specifications, it outlines a list and descriptive baseline of information system business functionalities. It serves as an index of business requirement assets, supporting real-time updates and query displays. It provides a reference for requirement writing, ensuring content aligns with reality, provides input for requirement analysis, uncovers business requirements, and addresses issues such as scattered requirement knowledge, low adaptability, and lack of management support, thereby improving the efficiency and quality of requirement management.

[0155] (3) Generate templates according to different requirements

[0156] Deeply integrating large-scale modeling technology into the entire requirements writing process enables refined classification management and intelligent modeling of enterprise-level requirements across the entire bank. In the requirements proposal stage, large-scale models analyze business data and, combined with market and business trends, provide forward-looking and feasible suggestions. During the writing phase, the language processing capabilities of large-scale models are leveraged to intelligently rewrite, expand, abbreviate, polish, and continue original requirements, ensuring compatibility with business scenarios through knowledge base integration. A requirements writing template system is constructed to provide refined classification management of enterprise-level requirements across the entire bank, covering various template types. Based on business scenario requirements, large-scale models automatically generate standardized requirements content, supporting user-selected chapter generation. Regarding intelligent modeling, based on standards and experience, the process involves streamlining the thought process, initializing the business modeling large-scale model's thought chain, forming a 1-4 level framework, identifying business parameters to complete 5-level modeling, refining review rules, preparing training products, and using fine-tuning techniques to train the large-scale model. This achieves the integration of intelligent modeling and requirements generation, improving automation and intelligence levels, and resolving existing technical issues.

[0157] (4) AI Search (Intelligent Semantic Search)

[0158] Semantic understanding: Based on the Large Language Model (LLM), it analyzes user intent, goes beyond keyword matching, and provides semantic association results.

[0159] Intelligent summary: Automatically summarizes relevant requirements or system functions and generates a general description.

[0160] Application scenarios: Quickly understand complex requirements (such as user inputting ambiguous questions and the model returning a structured explanation); obtain a high-level summary of system functions (avoiding reading the full text).

[0161] (5) Demand-based plagiarism check

[0162] 1) Embedding Phase. Anti-duplicate functionality is embedded during key phases such as requirement submission and review, relying on the industry's requirement capability map.

[0163] 2) Enhanced Content Extraction and Comparison: Compared to traditional comparison algorithms, this new algorithm supports intelligent extraction of key content from online rich text and attachments. It employs a combination of vector semantic comparison and keyword weighting to compare with historical assets, outputting Top N similar documents and filtering out false positives for empty templates, thereby improving the accuracy of plagiarism detection.

[0164] 3) Anti-duplicate scenario optimization mechanism. Establish an anti-duplicate scenario optimization library, obtain functional point knowledge and key description information from the full business requirement knowledge base, enhance fuzzy semantic matching capabilities, and expand the coverage of anti-duplicate scenarios.

[0165] (6) Requirements content check

[0166] 1) AI Capability Embedding Stage. A requirement content inspection function is embedded during the requirement generation stage. AI capabilities are used to assist users in inspecting completed requirement documents, aiming to address the issues of insufficient standardization and inconsistent quality of existing requirement content, and reduce the difficulty of manual review.

[0167] 2) Scope of automated checks. This includes automated checks on online requirements, uploaded requirement documents, and large model generation requirements.

[0168] 3) Multi-dimensional inspection methods. Typo, grammar, and semantic checks: Utilizing the general capabilities of the DeepSeek large model for accurate inspection; Sensitive word checks: Identifying sensitive words by comparing the content of the requirement document with an industry-wide sensitive word database; Requirement template compliance checks: Providing prompts for required fields based on different categories of templates built into the system, such as business modeling, modeling + function points, function points, and user stories; Business logic inspection mechanism: Utilizing the large model based on understanding and analysis of the industry's enterprise knowledge base and business rule bases for various lines of business to check the accuracy and completeness of the current requirement's business logic, reducing the workload and error rate of manual review.

[0169] 4) Intelligent error correction. Through intelligent error correction and prompts, the system guides writers to correct non-standard or erroneous expressions, and supports modes such as accepting revisions line by line and accepting revisions with one click.

[0170] This application has achieved at least the following technical effects:

[0171] (1) The construction of knowledge base and demand capability map significantly improves management efficiency through structured knowledge storage.

[0172] First, by categorizing knowledge into business domains (such as payment, finance, and risk control) through domain-based storage, the system not only facilitates accurate retrieval and avoids information clutter, but also visualizes the capability boundaries of each domain, allowing for quick location of functional modules and thus aiding in requirement planning. Second, leveraging intelligent reuse and deduplication mechanisms, the system can quickly retrieve historical requirements or solutions through domain tags, effectively avoiding repetitive work. It also automatically compares new requirements with existing knowledge, reducing redundant requirements and improving development efficiency. Furthermore, during requirement generation, the system automatically matches related domain knowledge (such as rules and processes) to ensure requirement completeness and analyzes the rationality of requirements based on the historical knowledge base, reducing subjective judgment bias and optimizing decision-making quality. Finally, by constructing a unified knowledge framework and standardizing terminology and processes, this system promotes cross-team collaboration and knowledge sharing, breaks down departmental information silos, and further reduces communication costs.

[0173] (2) Intelligent requirement generation: Combined with large model technology, high-quality requirement documents are automatically generated, reducing manual writing costs and improving the accuracy and completeness of requirement descriptions.

[0174] The large-scale model-enabled requirement writing feature intelligently recommends relevant content based on the context of user input, assisting users in quickly completing requirement documents. For example, when a user is writing requirements related to "user login," the system can automatically associate fragments such as "CAPTCHA generation rules" and "error message specifications" from similar historical requirements for the user to reference or directly insert, significantly improving writing efficiency. Simultaneously, the large-scale model can perform syntax validation, compliance checks, and sensitive word filtering on the generated requirement content, ensuring that the requirement documents conform to industry standards and company specifications, reducing subsequent review and rework costs. Furthermore, intelligent requirement generation supports multi-template adaptation and can automatically switch requirement structures according to different project types (such as agile development and traditional waterfall models), further enhancing the applicability and flexibility of requirement documents.

[0175] (3) Through refined classification and management of requirements, the standardization and flexibility of requirement writing have been achieved.

[0176] First, with the help of finely categorized templates, the system can generate structurally standardized requirement documents with one click according to different scenarios, ensuring output quality while flexibly adapting to various development modes such as intelligent modeling, functional points, and user stories. Second, users can choose to generate different chapters of the requirements, which greatly reduces the difficulty and time cost of writing professional documents, optimizes the user experience, and effectively improves work efficiency.

[0177] (4) The requirement content checking and deduplication checking functions can provide automated and comprehensive checks for written requirements, which significantly improves document quality and management efficiency.

[0178] It can automatically perform sensitive word and compliance checks, effectively preventing content risks and ensuring that requirements meet specifications and security standards. Simultaneously, by checking text grammar, spelling errors, and document templates, it ensures the accuracy and formatting of requirement expressions, thereby enhancing overall professionalism. Furthermore, its built-in requirement deduplication mechanism can quickly identify and prompt similar historical requirements, effectively promoting knowledge reuse and avoiding redundant development and resource waste.

[0179] This application also provides a business requirement generation and verification device 600, such as... Figure 6 The diagram shows a schematic representation of a business requirement generation and verification device according to an embodiment of this application. The business requirement generation and verification device 600 includes:

[0180] The receiving unit 610 is used to receive user requests for writing business requirements;

[0181] The requirement generation unit 620 is used to write requests and pre-built basic knowledge bases and predefined requirement templates according to the business requirements, and to generate preliminary requirement documents using a large language model.

[0182] The requirement deduplication unit 630 is used to deduplicatively check the preliminary requirement document using the pre-built basic knowledge base and the preset requirement deduplication strategy to obtain the requirement deduplication result;

[0183] The requirement checking unit 640 is used to check the content of the preliminary requirement document based on the requirement deduplication result, using the pre-built basic knowledge base, the large language model and the preset requirement content checking strategy, to obtain the final requirement document.

[0184] In some embodiments of this application, the requirement generation unit 620 is specifically used for: generating a basic requirement draft based on the business requirement writing request and reference materials, combined with the requirement compilation specification; determining the corresponding modeling strategy based on the requirement type in the business requirement writing request, and performing business modeling using the corresponding modeling strategy to obtain a large business model; and generating the preliminary requirement document based on the business requirement writing request, the reference materials, the basic requirement draft, and the predefined requirement template, using the pre-built basic knowledge base and the large business model.

[0185] In some embodiments of this application, the business requirement writing request includes basic requirement description information, and the requirement generation unit 620 is specifically used to: based on the basic requirement description information, use a pre-built basic knowledge base and a large language model to assist in requirement writing, and obtain the preliminary requirement document; wherein, the requirement assistance includes at least one of rewriting, expansion, abbreviation, polishing, and continuation.

[0186] In some embodiments of this application, the preset demand retrieval strategy includes at least one of a precise search strategy, an AI search strategy, and an advanced search strategy. The business demand generation and verification device 600 further includes: a demand retrieval unit, configured to: search the pre-built basic knowledge base using the precise search strategy based on the keywords carried in the business demand writing request, obtain a first search result and display it to the user; search the pre-built basic knowledge base using the AI ​​search strategy based on the keywords carried in the business demand writing request, call the large language model to search the pre-built basic knowledge base, obtain a second search result and display it to the user; and search the pre-built basic knowledge base using the advanced search strategy based on the keywords and multi-dimensional filtering conditions carried in the business demand writing request, obtain a third search result and display it to the user.

[0187] In some embodiments of this application, the requirement deduplication unit 630 is specifically used to: parse the preliminary requirement document to obtain the parsing result of the preliminary requirement document; perform requirement deduplication based on the parsing result of the preliminary requirement document and the requirement deduplication prevention and optimization library to obtain the requirement deduplication result, wherein the requirement deduplication prevention and optimization library is constructed based on the pre-built basic knowledge base.

[0188] In some embodiments of this application, the requirement deduplication result includes requirement documents that have passed the requirement deduplication check. The requirement checking unit 640 is specifically used to: use the pre-built basic knowledge base and the large language model to perform multi-dimensional content checks on the requirement documents that have passed the requirement deduplication check, and obtain multi-dimensional content check results; based on the multi-dimensional content check results, use an intelligent error correction and prompting mechanism to correct the requirement documents that have passed the requirement deduplication check, and obtain the final requirement document.

[0189] In some embodiments of this application, the business requirement generation and verification device 600 further includes: a construction unit, configured to acquire multi-source data within a domain; construct a basic knowledge base based on the multi-source data within the domain, the basic knowledge base including at least one of a business requirement knowledge base, a policy base, a salable product base, and a sensitive word base; and construct a scenario optimization library based on the basic knowledge base, the scenario optimization library including a requirement generation optimization library and a requirement anti-duplicate optimization library.

[0190] It is understood that the above-mentioned business requirement generation and verification device can implement each step of the business requirement generation and verification method provided in the foregoing embodiments. The relevant explanations of the business requirement generation and verification method are applicable to the business requirement generation and verification device, and will not be repeated here.

[0191] Figure 7 This is a schematic diagram of the structure of a device according to an embodiment of this application. For example... Figure 7 As shown, the device includes one or more processors (or processing units), and may also include one or more memories coupled to the processors, and may also include a communication module coupled to the processors.

[0192] A communication module can be used to communicate with other devices or apparatuses, such as sending or receiving data and / or signals. A communication module may have at least one communication module for communication. A communication module may include any interface necessary for communicating with other devices. Exemplarily, a communication module may be a transceiver, circuit, bus, module, or other type of communication module.

[0193] The processor may include, but is not limited to, one or more of the following: a general-purpose computer, a special-purpose computer, a microcontroller, a digital signal processor (DSP), or a controller-based multi-core controller architecture. The device may have multiple processors, such as application-specific integrated circuit (ASIC) chips, which are time-dependent on a clock synchronized with the main processor.

[0194] The memory may include one or more non-volatile memories and one or more volatile memories. Examples of non-volatile memories include, but are not limited to, at least one of the following: read-only memory (ROM), electrically programmable read-only memory (EPROM), flash memory, hard disk, compact disc (CD), digital video disc (DVD), or other magnetic and / or optical storage. Examples of volatile memories include, but are not limited to, at least one of the following: random access memory (RAM), or other volatile memories that do not persist during the duration of a power outage.

[0195] A computer program consists of computer-executable instructions that are executed by an associated processor. Programs can be stored in ROM. A processor can perform any appropriate action and processing by loading the program into RAM.

[0196] Possible implementations of this application can be achieved through a program, enabling the communication device to execute any of the processes discussed in the foregoing embodiments. Possible implementations of this application can also be achieved through hardware or a combination of software and hardware.

[0197] In some implementations, the program may be tangibly contained in a computer-readable storage medium, which may include in a device (such as in memory) or other storage device accessible by the device. The program may be loaded from the computer-readable storage medium into RAM for execution. The computer-readable storage medium may include any type of tangible non-volatile memory, such as ROM, EPROM, flash memory, hard disk, CD, DVD, etc.

[0198] This application also provides a computer-readable storage medium storing computer instructions or program code thereon, which, when executed by a processor, causes the processor to perform the methods and functions involved in any of the above embodiments. A computer-readable medium can be any tangible medium that contains or stores a program for or relating to an instruction execution system, apparatus, or device. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. More detailed examples of computer-readable storage media include electrical connections with one or more wires, magnetic media (e.g., disks, floppy disks, hard disks, magnetic tapes, magnetic storage devices), optical media (e.g., optical storage devices, DVDs), semiconductor media (e.g., solid-state drives), random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), or any suitable combination thereof.

[0199] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. Embodiments of this application also provide at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. This computer program product includes one or more computer-executable instructions, such as instructions included in a program module, which execute in a device on a target real or virtual processor to perform the processes, methods, and functions involved in any of the above embodiments. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0200] This application also proposes a computer program product, including a computer program or instructions that, when run on a computer, cause the computer to perform the processes, methods, and functions described in the above embodiments. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc., that perform specific tasks or implement specific abstract data types. In various embodiments, the functionality of program modules can be combined or divided as needed. The machine-executable instructions for the program modules can be executed locally or in a distributed device. In a distributed device, the program modules can reside in both local and remote storage media.

[0201] Generally, the various embodiments of this application can be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. Some aspects can be implemented in hardware, while others can be implemented in firmware or software, which can be executed by a controller, microprocessor, or other computing device. Although various aspects of the embodiments of this disclosure are shown and described as block diagrams, flowcharts, or represented using some other illustration, it should be understood that the blocks, apparatuses, systems, techniques, or methods described herein can be implemented as, as non-limiting examples, in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0202] It should be noted that although embodiments of this application have been described above with reference to the accompanying drawings, these embodiments are not independent of each other, and they can be combined to obtain other embodiments. The methods, situations, categories, and classifications of embodiments in this application are only for the convenience of description and should not constitute a special limitation. Various methods, categories, situations, and features in embodiments can be combined with each other if logically consistent. The various embodiments of this application can be arbitrarily combined to achieve different technical effects. The embodiments of this application will not list various combinations.

[0203] Furthermore, although the operation of the methods of this disclosure is described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all of the operations shown must be performed to achieve the desired result. Rather, the steps depicted in the flowcharts may be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps. It should also be noted that the features and functions of two or more devices according to this disclosure may be embodied in one device. Conversely, the features and functions of one device described above may be further divided and embodied by multiple devices.

[0204] It should also be noted that 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 limitation, 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.

[0205] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for generating and validating business requirements, characterized in that, The methods for generating and validating the business requirements include: Receive user requests to write business requirements; Based on the business requirements, write requests and pre-built basic knowledge bases and predefined requirement templates, and use a large language model to generate preliminary requirement documents; The preliminary requirement document is checked for plagiarism using the pre-built basic knowledge base and the preset requirement plagiarism detection strategy to obtain the requirement plagiarism detection results; Based on the results of the requirement deduplication check, the preliminary requirement document is checked using the pre-built basic knowledge base, the large language model, and the preset requirement content checking strategy to obtain the final requirement document.

2. The method for generating and validating business requirements according to claim 1, characterized in that, The step of writing requests based on the business requirements, using a pre-built basic knowledge base and predefined requirement templates, and generating preliminary requirement documents using a large language model includes: Based on the business requirements, draft requests and reference materials, and combine them with the requirements development specifications to generate a basic requirements draft. Based on the requirement type in the request written according to the business requirements, the corresponding modeling strategy is determined, and the corresponding modeling strategy is used to perform business modeling to obtain a large business model; Based on the business requirements, the draft request, the reference materials, the basic requirements draft, and the predefined requirements template, the preliminary requirements document is generated using the pre-built basic knowledge base and the business big model.

3. The method for generating and validating business requirements according to claim 2, characterized in that, The business requirement writing request includes basic requirement description information. The process of generating a preliminary requirement document using a large language model based on the business requirement writing request, a pre-built basic knowledge base, and a predefined requirement template includes: Based on the basic requirement description information, the requirements are written using a pre-built basic knowledge base and a large language model to obtain the preliminary requirement document. The aforementioned writing assistance includes at least one of rewriting, expansion, abbreviation, polishing, and continuation.

4. The method for generating and validating business requirements according to claim 1, characterized in that, The preset requirement retrieval strategy includes at least one of a precise search strategy, an AI search strategy, and an advanced search strategy. The method for generating and validating business requirements also includes: Based on the business requirements, the keywords carried in the request are written, and the precise search strategy is used to search the pre-built basic knowledge base to obtain the first search result and display it to the user. Based on the business requirements, the request is written with keywords, and the AI ​​search strategy is used to call the large language model to search in the pre-built basic knowledge base to obtain the second search result and display it to the user. Based on the business requirements, the request is written with keywords and multi-dimensional filtering conditions. The advanced search strategy is used to search the pre-built basic knowledge base to obtain the third search result and display it to the user.

5. The method for generating and validating business requirements according to claim 1, characterized in that, The step of using the pre-built basic knowledge base and preset requirement deduplication strategy to perform deduplication on the preliminary requirement document, and obtaining the requirement deduplication results, includes: The preliminary requirements document is parsed to obtain the parsing result of the preliminary requirements document; Based on the parsing results of the preliminary requirement document and the requirement deduplication prevention and optimization library, requirement deduplication is checked to obtain the requirement deduplication results. The requirement deduplication prevention and optimization library is built based on the pre-built basic knowledge base.

6. The method for generating and validating business requirements according to claim 1, characterized in that, The requirement deduplication result includes requirement documents that have passed the requirement deduplication. Based on the requirement deduplication result, the preliminary requirement documents are then subjected to content checks using the pre-built basic knowledge base, the large language model, and a preset requirement content checking strategy to obtain the final requirement documents, which include: Using the pre-built basic knowledge base and the large language model, the requirement document that has passed the requirement deduplication is subjected to multi-dimensional content inspection to obtain multi-dimensional content inspection results. Based on the multi-dimensional content inspection results, the requirement document that passed the requirement deduplication check is corrected using an intelligent error correction and prompting mechanism to obtain the final requirement document.

7. The method for generating and verifying business requirements according to any one of claims 1 to 6, characterized in that, The method for generating and validating the business requirements also includes: Acquire multi-source data within the domain; A basic knowledge base is constructed based on multi-source data in the aforementioned field. The basic knowledge base includes at least one of the following: a business requirements knowledge base, a policy base, a salable product base, and a sensitive word base. A scenario optimization library is constructed based on the aforementioned basic knowledge base. The scenario optimization library includes a requirement generation optimization library and a requirement deduplication prevention optimization library.

8. A device for generating and verifying business requirements, characterized in that, The device for generating and verifying the business requirements includes: The receiving unit is used to receive users' requests to write business requirements. The requirement generation unit is used to write requests and pre-built basic knowledge bases and predefined requirement templates based on the business requirements, and to generate preliminary requirement documents using a large language model. The requirement deduplication unit is used to use the pre-built basic knowledge base and preset requirement deduplication strategy to deduplicate the preliminary requirement document and obtain the requirement deduplication result. The requirement checking unit is used to check the content of the preliminary requirement document based on the requirement deduplication results, using the pre-built basic knowledge base, the large language model, and the preset requirement content checking strategy, to obtain the final requirement document.

9. An apparatus comprising: processor; And a memory configured to store computer-executable instructions, which, when executed, cause the processor to perform the method for generating and verifying the business requirements of any one of claims 1 to 7.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the method for generating and verifying the business requirements as described in any one of claims 1 to 7.

Images