Method and device for realizing intelligent quality valve in demand submission process, equipment and medium

Abstract

The invention discloses a method, a device, equipment and a medium for realizing an intelligent quality valve in a demand submission process, and relates to the field of demand process quality review scenes in financial science and technology, and the method comprises the steps: obtaining a demand description, uploading the demand description to a created large-model intelligent platform, carrying out document expansion based on comparison with existing similar demands, and obtaining a demand submission result; carrying out standardization processing to supplement missing elements, and generating a demand document; the big-model intelligent platform is used for acquiring demand review conference voice data, uploading the demand review conference voice data to the big-model intelligent platform, analyzing the demand review conference voice data to generate a demand review conference summary, and submitting the demand review conference summary to the DevOps platform; and creating a demand review model, and obtaining private domain data in real time by the demand review model through a retrieval enhancement technology of the retrieval enhancement generation model so as to review the current demand document to be reviewed. According to the invention, automatic review operation can be realized, the burden of manual review can be greatly reduced, and the accuracy and efficiency of review are improved.

Description

Technical Field

[0001] This application relates to the field of demand process quality review scenarios in financial technology, and specifically to a method, device, equipment and medium for implementing an intelligent quality valve in the demand submission process. Background Art

[0002] Currently, most software development teams rely on the following steps to control the quality of the requirements submission stage: (1) Requirements collection and analysis: user requirements are collected through meetings, interviews, etc., and preliminary analysis is performed by product managers or analysts; (2) Requirements document writing: the analyzed requirements are converted into detailed documents, including functional requirements, non-functional requirements, user stories, etc.; (3) Requirements review: relevant personnel (such as developers, testers, stakeholders) are organized to review the requirements documents to ensure the accuracy and feasibility of the requirements.

[0003] However, the existing method of submitting requirements has problems such as low review efficiency, frequent changes in requirements, and inconsistent understanding of requirements, which seriously affect the progress and quality of related projects. Summary of the Invention

[0004] The present application provides a method, device, equipment and medium for implementing an intelligent quality valve in the demand submission process, which can realize automated review operations, greatly reduce the burden of manual review, and improve the accuracy and efficiency of the review.

[0005] In a first aspect, an embodiment of the present application provides a method for implementing an intelligent quality valve in a demand submission process, the method comprising: Obtain the requirement description in text or image format and upload it to the created large-scale intelligent platform. Expand the document based on comparison with existing similar requirements, perform standardization to fill in missing elements, and generate the requirement document. Acquire the voice data of the requirements review meeting and upload it to the big model intelligent platform. The big model intelligent platform analyzes the voice data of the requirements review meeting to generate a requirements review meeting summary, and submits the requirements review meeting summary to the DevOps platform. Create a demand review model. The demand review model obtains private domain data in real time through the retrieval enhancement technology of the retrieval enhancement generation model to review the current demand document to be reviewed. The private domain data includes existing demand documents, demand review meeting voice data, and demand review meeting summary.

[0006] In conjunction with the first aspect, in one embodiment, the large model intelligent platform is specifically used to: Based on the agile access mechanism, it connects to various large network models, uses the multi-dimensional model evaluation system to dynamically obtain the advantages and characteristics of various large network models, and uses the intelligent matching mechanism to flexibly assemble the appropriate large network models and promote them to the current business scenarios.

[0007] In conjunction with the first aspect, in one embodiment, obtaining a demand description in text or image form and uploading it to the created large-scale model intelligent platform, expanding the document based on comparison with existing similar demands, and performing normalization processing to fill in missing elements to generate a demand document specifically includes: Obtain a demand description in text or image form, and upload the obtained demand description to the large model intelligent platform. When the demand description is in image form, the large model intelligent platform recognizes the image based on image recognition technology; The large-scale intelligent platform identifies key information in the requirement description and matches it with similar requirements in the requirement database. Based on the structure, content, and expression specifications of the requirements, and in combination with the contextual semantics of the requirement description, the large-scale intelligent platform classifies, structurally analyzes, and expands the requirement description. It also fills in missing elements and generates a standardized requirement document. Among them, the elements of the requirement document include requirement background and objectives, requirement scope, functional requirement details, non-functional requirements, change management mechanism, requirement tracking matrix, risk management, and related impacts.

[0008] In conjunction with the first aspect, in one embodiment, obtaining the voice data of the requirements review meeting and uploading it to the big model intelligent platform, the big model intelligent platform analyzing the voice data of the requirements review meeting to generate a requirements review meeting summary, and submitting the requirements review meeting summary to the DevOps platform specifically includes: Recording is performed during the demand review meeting to obtain voice data of the demand review meeting, converting the voice data of the demand review meeting into text content, and uploading the voice data of the demand review meeting to the big model intelligent platform; The large-model intelligent platform analyzes the voice data of the demand review meeting and generates a demand review meeting summary according to preset standards. The demand review meeting summary includes key discussion points and decision results.

[0009] Upload the generated requirements review meeting summary to the DevOps platform and archive it.

[0010] In conjunction with the first aspect, in one embodiment, the requirements review model is created, and the requirements review model uses the retrieval enhancement technology of the retrieval enhancement generation model to obtain private domain data in real time to review the current requirements document to be reviewed. The private domain data includes existing requirements documents, voice data from requirements review meetings, and requirements review meeting summaries, specifically including: Create a requirements review model. The requirements review model uses the retrieval enhancement technology of the retrieval enhancement generation model to obtain private domain data in the private domain retrieval database in real time. The private domain data includes existing requirements documents, voice data from the requirements review meeting, and requirements review meeting summaries. The requirements review model reviews the requirements document currently under review based on the existing requirements document, answers questions about the requirements document currently under review, and generates new private domain data based on the review results. Among them, when reviewing the current requirement document to be reviewed, the review is conducted based on the quality scoring method, specifically: ; in, Indicates the quality score of the current requirement document to be reviewed, and when If it is not less than the set threshold, it means the review is passed, otherwise it will suggest modifications. Indicates the The score of the evaluation factor, Indicates the total number of review factors, Indicates the The weight of the review factors, and .

[0011] In conjunction with the first aspect, in one embodiment, The overall architecture of the retrieval enhancement generation model includes the private domain retrieval database construction stage, the retrieval refinement stage, and the prompt generation stage; The private domain retrieval database construction phase is to construct a private domain retrieval database based on private domain data, so as to query data when the current requirement document to be reviewed is related to the existing historical requirement document and retrieval enhancement is required; The retrieval and ranking stage is a stage of searching and ranking data in the private domain retrieval database based on the current requirement document to be reviewed; The prompt generation stage is a stage in which the demand review model generates review results based on the retrieval and ranking stage.

[0012] In conjunction with the first aspect, in one embodiment, For the private domain search database construction phase, specifically: Collect existing private domain data, deconstruct and standardize the collected private domain data to achieve data cleansing, and introduce desensitization and semantic classification mechanisms during the cleaning process; The cleaned data is segmented into data blocks. A custom private domain-aware vector generation network is used to integrate contextual timestamps, business domain, and structural information for three-channel input to generate multi-dimensional dynamic knowledge vectors for the data blocks, thus achieving vector embedding of the data blocks. The custom private domain-aware vector generation network is constructed by introducing a domain attention mechanism based on the Transformer encoder. The data blocks after vector embedding are uniformly stored in an efficient vector database to build a private domain retrieval database, and an incremental monitoring and update mechanism is set up to monitor the addition and modification operations of private domain data; For the retrieval and sorting stage, specifically: Based on keyword matching or Embedding similarity recall mechanism, a multi-task intent recognition network is introduced to construct an intent-guided recall recognition network. For questions or requirements in the current requirements document to be reviewed, the constructed intent-guided recall recognition network determines whether it involves a recall intent for a specific requirement type. If so, a recall strategy is dynamically selected. Based on the question or requirement and combined with the cosine similarity calculation method, relevant data blocks are obtained in the private domain retrieval database to achieve data block recall. By introducing a semantic extension mechanism, the recall capability of hyponyms and hyponyms is enhanced based on the external general knowledge graph. The ranking score of the data block is calculated based on the ranking factor to finely sort the recalled data block to obtain the finely sorted data block. The calculation of the ranking score is specifically as follows: ; in, Indicates the The ranking score of the data blocks, Indicates the The semantic similarity between the data block and the query, Indicates the The timeliness score of each data block, Indicates the How often each data block is used or confirmed by experts, Indicates the The percentage of data blocks used in previous successful reviews, 、 、 、 represents the weight coefficient; For the prompt generation phase, specifically: For the sorted data blocks, a unified prompt template is constructed based on the issues in the current requirement document to be reviewed. The prompt template includes a requirement summary, referenced review rules, relevant historical review case snippets, and review focus points. Based on the carefully sorted data blocks, the context information of the demand review model is constructed. Based on the constructed context information, the demand review model responds to and reviews the questions or requirements in the current demand document to be reviewed according to the Prompt template, generates review results, and actively reviews and corrects the generated review results.

[0013] In a second aspect, an embodiment of the present application provides a device for implementing an intelligent quality valve in a demand submission process, the device comprising: The requirements document writing module is used to obtain the requirements description in text or image form and upload it to the created large-scale intelligent platform. It then expands the document based on the comparison with existing similar requirements, performs standardization processing to fill in missing elements, and generates the requirements document. A requirements review summary module is used to obtain voice data from the requirements review meeting and upload it to the large model intelligent platform. The large model intelligent platform analyzes the voice data from the requirements review meeting to generate a summary of the requirements review meeting, and submits the summary of the requirements review meeting to the DevOps platform. The review processing execution module is used to create a demand review model. The demand review model obtains private domain data in real time through the retrieval enhancement technology of the retrieval enhancement generation model to review the current demand document to be reviewed. The private domain data includes existing demand documents, demand review meeting voice data, and demand review meeting summary.

[0014] In the third aspect, an embodiment of the present application provides a device for implementing an intelligent quality valve for a demand submission process, wherein the device comprises a processor, a memory, and a program for implementing an intelligent quality valve for a demand submission process stored in the memory and executable by the processor, wherein when the program for implementing an intelligent quality valve for a demand submission process is executed by the processor, the steps of the method for implementing an intelligent quality valve for a demand submission process described above are implemented.

[0015] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a program for implementing an intelligent quality valve for a demand submission process is stored. When the program for implementing an intelligent quality valve for a demand submission process is executed by a processor, the steps of the above-mentioned method for implementing an intelligent quality valve for a demand submission process are implemented.

[0016] The beneficial effects of the technical solutions provided in the embodiments of the present application include: By automatically generating requirement documents and requirement review meeting summaries, introducing a large-model intelligent platform based on artificial intelligence technology, and optimizing the requirement review model through retrieval enhancement generation models, the requirement review model simulates the expert review process, conducts intelligent review of requirement documents, and realizes automated review operations, which can greatly reduce the burden of manual review and improve the accuracy and efficiency of the review. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A flowchart of the method for implementing the intelligent quality valve in the submission process is required for this application; Figure 2 Submit a functional module diagram of the intelligent quality valve implementation device for the process required by this application; Figure 3 Submit a hardware structure diagram of the intelligent quality valve implementation equipment for this application. DETAILED DESCRIPTION

[0018] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0019] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0020] In a first aspect, an embodiment of the present application provides a method for implementing an intelligent quality valve in a demand submission process.

[0021] In one embodiment, referring to Figure 1 , Figure 1 Submit a flowchart of the method for implementing the intelligent quality valve in the process for this application. Figure 1 As shown in the figure, the method for implementing the intelligent quality valve in the demand submission process includes: S1: Obtain the requirement description in text or image format and upload it to the created large-scale intelligent platform. Expand the document based on the comparison with existing similar requirements, perform standardization to fill in missing elements, and generate the requirement document. S2: Acquire the voice data of the requirements review meeting and upload it to the big model intelligent platform. The big model intelligent platform analyzes the voice data of the requirements review meeting to generate a requirements review meeting summary, and submits the requirements review meeting summary to the DevOps platform. S3: Create a demand review model. The demand review model obtains private domain data in real time through the retrieval enhancement technology of the retrieval enhancement generation model to review the current demand document to be reviewed. The private domain data includes existing demand documents, demand review meeting voice data, and demand review meeting summary.

[0022] This application not only enables the writing, tracking, and change management of requirement documents, but also tightly integrates with the DevOps platform. Through integration, quality gates can be set at the requirement stage to ensure that only requirements that have been fully reviewed and verified can enter the development stage.

[0023] It should be noted that the large model intelligent platform is specifically used to: access various large network models based on an agile access mechanism, dynamically obtain the advantageous features of various large network models using a multi-dimensional model evaluation system, and flexibly assemble suitable large network models through an intelligent matching mechanism and promote them to current business scenarios.

[0024] Furthermore, the large-scale model intelligent platform can be the Changjiang Lingxi Financial Large-Scale Model Intelligent Platform. Based on an agile access mechanism, it rapidly introduces various mainstream large-scale network models (such as Wenxin Yiyan, ChatGPT, and DeepSeek) from the market. It utilizes a multi-dimensional model evaluation system to dynamically capture the advantages and characteristics of each model. Through intelligent matching, it flexibly assembles and recommends suitable large-scale network models to applicable business scenarios. Various AI capabilities are provided to users as plug-ins, ready-to-use. Combined with data and system interaction plug-ins, creative inspiration can be realized through the construction of intelligent agents with a low threshold. The connection and coupling of large-scale network models allows intelligent agents to better meet business scenario requirements, forming a large-scale model application ecosystem that is shared and co-built by all employees.

[0025] Furthermore, in one embodiment, a requirement description in text or image format is obtained and uploaded to the created large-scale model intelligent platform. The document is expanded based on comparison with existing similar requirements, and normalized to fill in missing elements to generate a requirement document, specifically including: S101: Obtain a demand description in text or image form, and upload the obtained demand description to the large model intelligent platform. When the demand description is in image form, the large model intelligent platform recognizes the image based on image recognition technology; S102: The large model intelligent platform identifies the key information in the requirement description and matches it with similar requirements in the requirement database. Based on the structure, content and expression specifications of the requirements, and in combination with the contextual semantics of the requirement description, the large model intelligent platform classifies, structurally analyzes and expands the requirement description, and fills in missing elements to generate a standardized requirement document; wherein, the elements of the requirement document include the requirement background and objectives, requirement scope, functional requirement details, non-functional requirements, change management mechanism, requirement tracking matrix, risk management, and related impacts. It should be noted that for the generated requirement document, after being reviewed as a requirement document to be reviewed, it will be stored in the private domain data as an existing requirement document.

[0026] Specifically, AI is used to write requirements documents. When creating requirements, users can simply describe their requirements or upload an image to indicate the features they need to support. This content is then submitted to the Big Model Intelligent Platform, where a dedicated requirements agent performs image recognition and preliminary semantic extraction. The identified key information is then matched against similar requirements in the requirements database. Based on the structure, content, and expression specifications of the requirements, and in conjunction with the contextual semantics of the user input, the Big Model Intelligent Platform categorizes, analyzes, and expands the requirements description. The expansion logic is primarily based on a "semantic similarity comparison + structured element template filling" mechanism to ensure complete document content, clear boundaries, and clear acceptance criteria.

[0027] Then, standardization is performed, missing elements are filled in, and a standardized requirements document that meets the DOMM Level 3 standard is generated. This missing element filling operation refers to the large model intelligent platform filling in missing elements. Elements are specified in a standard-certified requirements document template, including: requirements background and objectives, requirements scope, functional requirements details, non-functional requirements, change management mechanism, requirements tracking matrix (establishing a correspondence between requirements and development and testing phases), risk management, and associated impacts. DOMM Level 3 refers to the "Capability Maturity Model for Integrated Development, Operation, and Maintenance of Systems and Software Engineering" (GB / T 42560-2023, referred to as the "DevOps National Standard") issued by the China Electronics Technology Standardization Institute.

[0028] The large-model intelligent platform of this application can assist in writing requirement documents and improve the efficiency and standardization of requirement document writing. Similarly, the large-model intelligent platform is also suitable for the verification and auxiliary writing of other project documents such as design documents, and comprehensively optimizes the document management process.

[0029] Furthermore, in one embodiment, voice data from the requirements review meeting is obtained and uploaded to the large model intelligent platform. The large model intelligent platform analyzes the voice data from the requirements review meeting to generate a requirements review meeting summary, and submits the requirements review meeting summary to the DevOps platform, specifically including: S201: recording the demand review meeting to obtain voice data of the demand review meeting, converting the voice data of the demand review meeting into text content, and uploading the voice data of the demand review meeting to the big model intelligent platform; S202: The large model intelligent platform analyzes the voice data of the demand review meeting and generates a demand review meeting summary according to preset standards. The demand review meeting summary includes key discussion points and decision results.

[0030] S203: Upload the generated requirements review meeting summary to the DevOps platform and archive it.

[0031] Specifically, AI is used to summarize the requirements review meeting. During the requirements review meeting, recording is performed to obtain voice data of the requirements review meeting, which is converted into text content. At the same time, the voice data of the requirements review meeting is uploaded to the big model intelligent platform. The big model intelligent platform analyzes the voice data of the requirements review meeting and automatically generates a summary of the requirements review meeting according to preset standards to facilitate the organization and archiving of meeting minutes. The summary of the requirements review meeting is then uploaded to the DevOps platform and retained as a requirements review node document for future tracing.

[0032] In this application, the term "DevOps platform" refers to an integrated development, operations, and maintenance management platform that has achieved Level 3 certification under the "System and Software Engineering Development, Operation, and Maintenance Integration Capability Maturity Model" (GB / T 42560-2023, referred to as the "DevOps National Standard") developed by the China Electronics Standardization Institute. The DevOps platform encompasses the entire R&D lifecycle, building an integrated architecture encompassing process management, development processes, quality control, continuous delivery, metrics analysis, outsourcing management, and multi-system integration. The DevOps platform integrates requirements, tasks, and versions into unified management, enabling online and visual development processes. It also integrates service systems, code projects, and outsourced human resources into a platform, effectively improving resource scheduling efficiency.

[0033] Furthermore, in one embodiment, a requirements review model is created. The requirements review model uses the retrieval enhancement technology of the retrieval enhancement generation model to obtain private domain data in real time to review the current requirements document to be reviewed. The private domain data includes existing requirements documents, voice data from requirements review meetings, and requirements review meeting summaries, specifically including: S301: Create a requirements review model. The requirements review model uses the retrieval enhancement technology of the retrieval enhancement generation model to obtain private domain data in the private domain retrieval database in real time. The private domain data includes existing requirements documents, voice data from the requirements review meeting, and requirements review meeting summaries. S302: The demand review model reviews the demand document to be reviewed based on the existing demand document, and answers questions about the demand document to be reviewed. At the same time, new private domain data is generated based on the review results.

[0034] Among them, when reviewing the current requirement document to be reviewed, the review is conducted based on the quality scoring method, specifically: ; in, Indicates the quality score of the current requirement document to be reviewed, and when If it is not less than the set threshold, it means the review is passed, otherwise it will suggest modifications. Indicates the The score of the evaluation factor, Indicates the total number of review factors, Indicates the The weight of the review factors, and Evaluation factors include background completeness, requirement clarity, and acceptance criteria clarity.

[0035] It should be noted that in the requirements review scenario, previously existing requirements documents, requirements review meeting voice data, requirements review meeting summaries, that is, historical data such as requirements and requirements reviews, are private domain data. The requirements review model needs to be trained and adapted for this type of private domain data before it can be used. Traditional model training methods require the preparation of expensive GPU machine clusters, which has high training costs and cannot be effective in real time, and cannot meet the needs of rapid knowledge updates. To solve this problem, this application uses the retrieval enhancement technology of the retrieval enhancement generation model (RAG) to optimize the requirements review model. Specifically, the requirements review model obtains private domain data from the private domain retrieval database in real time through the retrieval enhancement technology of the retrieval enhancement generation model. The requirements review model reviews and processes the current requirements document to be reviewed based on the existing requirements documents and the requirements document standards required by DOMM level 3, and answers questions about the current requirements document to be reviewed. At the same time, private domain data can be updated or effective immediately. Furthermore, the demand review model improves the recall rate of knowledge data through technologies such as multi-dimensional recall and refined knowledge segmentation, and selects the most relevant knowledge through precise sorting strategies and relevance screening strategies to introduce into the context of generation / question and answer, further enhancing the practicality and accuracy of the review.

[0036] In this application, the overall architecture of the retrieval enhancement generation model includes a private domain retrieval database construction stage, a retrieval refinement stage, and a prompt generation stage; the private domain retrieval database construction stage is to construct a private domain retrieval database based on private domain data, so as to query data when the current requirement document to be reviewed is related to the existing historical requirement document and retrieval enhancement is required; the retrieval refinement stage is to search and refine the data in the private domain retrieval database based on the current requirement document to be reviewed; the prompt generation stage is the stage in which the requirement review model generates the review results based on the retrieval refinement stage.

[0037] In this application, regarding the private domain search database construction phase, specifically: a1: Collect existing private domain data, deconstruct the data, and perform standardized annotation (based on DOMM Level 3 requirements) to cleanse the data to improve subsequent embedding accuracy. Desensitization and semantic classification mechanisms are introduced during the cleaning process to ensure data privacy and facilitate knowledge unitization. a2: The cleaned data is segmented into data blocks. A customized private domain-aware vector generation network is used to integrate the contextual timestamp, business domain (such as characteristic words in the financial and government sectors), and structural information (title, body, and conclusion segments). This three-channel input generates a multi-dimensional dynamic knowledge vector for the data block and implements vector embedding for the data block. The customized private domain-aware vector generation network is constructed by introducing a domain attention mechanism based on the Transformer encoder. This network focuses on industry terminology and review semantics, making the embedding more relevant to business scenarios. a3: Store the vector-embedded data blocks uniformly in an efficient vector database (such as FAISS or Weaviate) to build a private domain search database. Set up an incremental monitoring and update mechanism to monitor additions and modifications to private domain data. By monitoring these additions and modifications, data re-embedding and index updates are automatically triggered, ensuring the search database is synchronized with the latest enterprise knowledge in real time.

[0038] Specifically, during the construction phase of the private domain retrieval database, data collection and cleaning are first performed. Existing private domain data, such as requirements documents, voice data from requirements review meetings, and summaries of requirements review meetings, is collected and cleaned to ensure data quality. Knowledge segmentation and indexing are then performed, using multi-dimensional recall strategies such as keywords, topics, and time to segment the private domain data into data blocks. Vector embeddings for these data blocks are then generated and stored in an efficient vector database, completing the construction of the private domain retrieval database. Furthermore, a real-time update mechanism can be introduced, and automated processes designed to update private domain data regularly or in real time ensure that the knowledge in the private domain retrieval database is always up to date.

[0039] In this application, for the search and ranking stage, specifically: b1: Based on keyword matching or embedding similarity recall mechanism, a multi-task intent recognition network is introduced to build the intent-guided recall recognition network (PAEN). b2: For questions or requirements in the current requirements document to be reviewed, the constructed intent-guided recall recognition network determines whether it involves a recall intent for a specific requirement type. If so, a recall strategy is dynamically selected. Based on the question or requirement and combined with the cosine similarity calculation method, relevant data blocks are obtained from the private domain retrieval database to achieve data block recall. Furthermore, a semantic extension mechanism is introduced to enhance the recall capability of hyponyms and hyponyms based on the external general knowledge graph. Specifically, the constructed intent-guided recall recognition network determines whether the issues or requirements in the current requirement document to be reviewed involve specific requirement recall intentions, such as whether they are standardized or meet acceptance conditions. When the recall intention is clear, a recall strategy is dynamically selected, such as enabling "requirement number" retrieval, recalling associated meeting content, and other strategy combinations. Then, based on the similarity calculation method, relevant data blocks are obtained to realize the recall of data blocks. At the same time, a semantic extension mechanism is introduced to enhance the user's query and recall capabilities of synonyms and hyponyms based on external general knowledge graphs (such as the requirement glossary) to avoid missing key information. b3: Calculate the ranking score of the data block based on the ranking factor to finely sort the recalled data block to obtain the finely sorted data block. The calculation of the ranking score is specifically as follows: ; in, Indicates the The ranking score of the data blocks, Indicates the The semantic similarity between the data block and the query, Indicates the The timeliness score of each data block, Indicates the How often each data block is used or confirmed by experts, Indicates the The percentage of data blocks used in previous successful reviews, 、 、 、 Represents the weight coefficient; ranking factors include text similarity, document timeliness, expert review weight, historical review success rate, etc. Combined with the calculated ranking scores, the data blocks are sorted in descending order to achieve refined sorting.

[0040] When sorting data blocks based on ranking scores, this can be done through a related ranking model. This ranking model uses an improved weighted boosting model (introducing a temporal attention factor based on XGBoost) and combines it with a lightweight graph neural network (GNN) to model the relationships between data blocks, thereby improving the hierarchy and interpretability of the sorting. Furthermore, data blocks with similarity below a set threshold can be filtered out to ensure the quality of context in the subsequent generation stage.

[0041] In this application, for the prompt generation stage, specifically: c1: For the sorted data blocks, a unified prompt template is constructed based on the issues in the current requirement document to be reviewed. The prompt template includes a requirement summary, referenced review rules (such as Article X of the DOMM Level 3 standard), relevant historical review case snippets, and review focus points. Review focus points are enterprise-defined focus points used to guide the generation of key points, such as acceptance criteria and the rationality of the implementation path. c2: Based on the carefully sorted data blocks, the context information of the requirements review model is constructed. Based on the constructed context information, the requirements review model responds to and reviews the questions or requirements in the current requirements document to be reviewed according to the prompt template, generates review results, and actively reviews and corrects the generated review results.

[0042] Furthermore, customized rule templates (such as format standards and key field detectors) can be inserted as search guide tokens to enhance generation accuracy and enterprise consistency. Furthermore, combined with the "Common Review Logic Library," the system generates results that support both natural language and structured suggestions (for example, "Suggest adding acceptance criteria fields" + JSON structure field annotations).

[0043] Through the above methods, the demand review model can be enhanced, enabling it to acquire and apply existing private domain knowledge in real time, thereby improving the efficiency and accuracy of the review; at the same time, through technical means such as multi-dimensional recall, refined knowledge segmentation, precise ranking strategies and relevance screening, the practicality and accuracy of the review are further enhanced.

[0044] Compared with the traditional RAG that directly calls BERT or DPR to generate vectors, this application uses the self-developed PAEN network structure, introduces private domain perception and structure enhancement mechanisms, and is more adaptable to the dynamic and multidimensional nature of enterprise knowledge; it innovatively introduces a multi-factor fusion sorting mechanism, changing the traditional single similarity sorting to a refined sorting model that integrates semantics, timeliness, expert authority, and historical success rate, thereby improving accuracy and practicality; the intent-guided recall recognition network integrates sequence modeling and contextual situation judgment capabilities, and has stronger task matching capabilities than conventional classifiers, and is especially suitable for judging complex intents in demand review; it adopts a structural semantic collaborative embedding solution that integrates document structure, time, and business domain semantics, which is far superior to the existing general vector embedding method in terms of adaptability to private domain documents.

[0045] Furthermore, for the method for implementing the intelligent quality valve in the requirement submission process of this application, a comprehensive quality score may be defined in actual application, specifically: ; in, Represents the comprehensive quality score, and when When the quality is not less than the set minimum threshold, it means that the demand can enter the development stage. Indicates the compilation pass rate, Indicates the static code scanning score, Indicates the test coverage, Indicates the number of security vulnerabilities found. 、 、 、 Represents the weight parameter, which is customized by the quality policy.

[0046] The method for implementing an intelligent quality valve in the requirement submission process of the embodiment of the present application automatically generates requirement documents and requirement review meeting summaries, introduces a large model intelligent platform based on artificial intelligence technology, and optimizes the requirement review model through retrieval enhancement generation model, so that the requirement review model simulates the expert review process, performs intelligent review of requirement documents, and realizes automated review operations, which can greatly reduce the burden of manual review and improve the accuracy and efficiency of the review.

[0047] In a second aspect, an embodiment of the present application also provides a device for implementing an intelligent quality valve in a demand submission process.

[0048] In one embodiment, referring to Figure 2 , Figure 2 For the purpose of this application, please submit a functional module diagram of the intelligent quality valve implementation device. Figure 2 As shown, the intelligent quality valve implementation device for the requirement submission process includes: a requirement document writing module, a requirement review and summary module, and a review processing execution module.

[0049] The requirement document writing module is used to obtain the requirement description in text or picture form and upload it to the created large model intelligent platform, expand the document based on the comparison with existing similar requirements, and perform standardization processing to fill in the missing elements to generate the requirement document; the requirement review summary module is used to obtain the voice data of the requirement review meeting and upload it to the large model intelligent platform. The large model intelligent platform analyzes the voice data of the requirement review meeting to generate a requirement review meeting summary, and submits the requirement review meeting summary to the DevOps platform; the review processing execution module is used to create a requirement review model. The requirement review model obtains private domain data in real time through the retrieval enhancement technology of the retrieval enhancement generation model to review the current requirement document to be reviewed. The private domain data includes existing requirement documents, requirement review meeting voice data, and requirement review meeting summary.

[0050] In a third aspect, an embodiment of the present application provides a device for implementing an intelligent quality valve in a demand submission process. The device for implementing an intelligent quality valve in a demand submission process may be a personal computer (PC), a laptop computer, a server, or other device with data processing capabilities.

[0051] Reference Figure 3 , Figure 3 This is a hardware structure diagram of the intelligent quality valve implementation device for the demand submission process involved in the embodiment of the present application. In the embodiment of the present application, the intelligent quality valve implementation device for the demand submission process may include a processor, a memory, a communication interface, and a communication bus.

[0052] The communication bus may be of any type and is used to interconnect the processor, memory, and communication interface.

[0053] Communication interfaces include input / output (I / O), physical, and logical interfaces. These interfaces interconnect components within the device and other devices (such as other computing devices or user devices) used to implement the intelligent quality valve for demand submission. Physical interfaces can include Ethernet, fiber, or ATM interfaces; user devices can include displays and keyboards.

[0054] The memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0055] The processor may be a general-purpose processor that can invoke a program for implementing an intelligent quality valve for the demand submission process stored in a memory and execute the method for implementing an intelligent quality valve for the demand submission process provided in the embodiments of the present application. For example, the general-purpose processor may be a central processing unit (CPU). The method executed when the program for implementing an intelligent quality valve for the demand submission process is invoked can be referenced in the various embodiments of the method for implementing an intelligent quality valve for the demand submission process provided in the present application and will not be further described here.

[0056] Those skilled in the art will understand that Figure 3 The hardware structure shown in the figure does not constitute a limitation to the present application and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0057] In a fourth aspect, an embodiment of the present application also provides a computer-readable storage medium.

[0058] The computer-readable storage medium of the present application stores a program for implementing an intelligent quality valve for a demand submission process, wherein when the program for implementing an intelligent quality valve for a demand submission process is executed by a processor, the steps of the method for implementing an intelligent quality valve for a demand submission process as described above are implemented.

[0059] Among them, the method implemented when the intelligent quality valve implementation program of the demand submission process is executed can refer to the various embodiments of the intelligent quality valve implementation method of the demand submission process of this application, and will not be repeated here.

[0060] The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices. The terms "first", "second" and "third" are used to distinguish different objects, etc., and do not represent a sequence, nor do they limit the "first", "second" and "third" to different types.

[0061] In the description of the embodiments of this application, the words "exemplary," "for example," or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary," "for example," or "for example" in the embodiments of this application should not be construed as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary," "for example," or "for example" is intended to present the relevant concepts in a concrete manner.

[0062] In the description of the embodiments of the present application, unless otherwise specified, “ / ” means or, for example, A / B can mean A or B; “and / or” in the text is merely a description of the association relationship of associated objects, indicating that three relationships may exist, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of the present application, “multiple” refers to two or more than two.

[0063] In some processes described in the embodiments of the present application, multiple operations or steps are included that appear in a specific order. However, it should be understood that these operations or steps may not be performed in the order in which they appear in the embodiments of the present application or may be performed in parallel. The sequence numbers of the operations are only used to distinguish between different operations, and the sequence numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations or steps may be performed in sequence or in parallel, and these operations or steps may be combined.

[0064] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, or the part that contributes to the existing technology, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above and includes a number of instructions for enabling a terminal device to execute the methods described in each embodiment of this application.

[0065] The above are only preferred embodiments of the present application and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for implementing an intelligent quality valve in a demand submission process, characterized in that: The method for implementing the intelligent quality valve in the demand submission process includes: Obtain the requirement description in text or image format and upload it to the created large-scale intelligent platform. Expand the document based on comparison with existing similar requirements, perform standardization to fill in missing elements, and generate the requirement document. Acquire the voice data of the requirements review meeting and upload it to the big model intelligent platform. The big model intelligent platform analyzes the voice data of the requirements review meeting to generate a requirements review meeting summary, and submits the requirements review meeting summary to the DevOps platform. Create a demand review model. The demand review model obtains private domain data in real time through the retrieval enhancement technology of the retrieval enhancement generation model to review the current demand document to be reviewed. The private domain data includes existing demand documents, demand review meeting voice data, and demand review meeting summary.

2. The method for implementing an intelligent quality valve in a demand submission process according to claim 1, characterized in that: The large model intelligent platform is specifically used for: Based on the agile access mechanism, it connects to various large network models, uses the multi-dimensional model evaluation system to dynamically obtain the advantages and characteristics of various large network models, and uses the intelligent matching mechanism to flexibly assemble the appropriate large network models and promote them to the current business scenarios.

3. The method for implementing an intelligent quality valve in a demand submission process according to claim 1, characterized in that: The requirement description in text or image form is obtained and uploaded to the created large-scale model intelligent platform. The document is expanded based on the comparison with existing similar requirements, and normalized to fill in the missing elements to generate the requirement document, which specifically includes: Obtain a demand description in text or image form, and upload the obtained demand description to the large model intelligent platform. When the demand description is in image form, the large model intelligent platform recognizes the image based on image recognition technology; The large-scale intelligent platform identifies key information in the requirement description and matches it with similar requirements in the requirement database. Based on the structure, content, and expression specifications of the requirements, and in combination with the contextual semantics of the requirement description, the large-scale intelligent platform classifies, structurally analyzes, and expands the requirement description. It also fills in missing elements and generates a standardized requirement document. Among them, the elements of the requirement document include requirement background and objectives, requirement scope, functional requirement details, non-functional requirements, change management mechanism, requirement tracking matrix, risk management, and related impacts.

4. The method for implementing an intelligent quality valve in a demand submission process according to claim 1, wherein: The acquisition of the demand review meeting voice data and uploading it to the large model intelligent platform, the large model intelligent platform analyzing the demand review meeting voice data to generate a demand review meeting summary, and submitting the demand review meeting summary to the DevOps platform, specifically includes: Recording is performed during the demand review meeting to obtain voice data of the demand review meeting, converting the voice data of the demand review meeting into text content, and uploading the voice data of the demand review meeting to the big model intelligent platform; The large-scale intelligent platform analyzes the voice data of the demand review meeting and generates a demand review meeting summary according to preset standards. The demand review meeting summary includes key discussion points and decision results. Upload the generated requirements review meeting summary to the DevOps platform and archive it.

5. The method for implementing an intelligent quality valve in a demand submission process according to claim 1, characterized in that: The requirement review model is created. The requirement review model uses the retrieval enhancement technology of the retrieval enhancement generation model to obtain private domain data in real time to review the current requirement document to be reviewed. The private domain data includes existing requirement documents, voice data of the requirement review meeting, and summary of the requirement review meeting, specifically including: Create a requirements review model. The requirements review model uses the retrieval enhancement technology of the retrieval enhancement generation model to obtain private domain data in the private domain retrieval database in real time. The private domain data includes existing requirements documents, voice data from the requirements review meeting, and requirements review meeting summaries. The requirements review model reviews the requirements document currently under review based on the existing requirements document, answers questions about the requirements document currently under review, and generates new private domain data based on the review results. Among them, when reviewing the current requirement document to be reviewed, the review is conducted based on the quality scoring method, specifically: ； in, Indicates the quality score of the current requirement document to be reviewed, and when If it is not less than the set threshold, it means the review is passed, otherwise it will suggest modifications. Indicates the The score of the evaluation factor, Indicates the total number of review factors, Indicates the The weight of the review factors, and .

6. The method for implementing an intelligent quality valve in a demand submission process according to claim 5, characterized in that: The overall architecture of the retrieval enhancement generation model includes the private domain retrieval database construction stage, the retrieval refinement stage, and the prompt generation stage; The private domain retrieval database construction phase is to construct a private domain retrieval database based on private domain data, so as to query data when the current requirement document to be reviewed is related to the existing historical requirement document and retrieval enhancement is required; The retrieval and sorting stage is a stage of searching and sorting data in the private domain retrieval database based on the current requirement document to be reviewed; The prompt generation stage is a stage in which the demand review model generates review results based on the retrieval and ranking stage.

7. The method for implementing an intelligent quality valve in a demand submission process according to claim 6, characterized in that: For the private domain search database construction phase, specifically: Collect existing private domain data, deconstruct and standardize the collected private domain data to achieve data cleansing, and introduce desensitization and semantic classification mechanisms during the cleaning process; The cleaned data is segmented into data blocks. A custom private domain-aware vector generation network is used to integrate contextual timestamps, business domain, and structural information for three-channel input to generate multi-dimensional dynamic knowledge vectors for the data blocks, thus achieving vector embedding of the data blocks. The custom private domain-aware vector generation network is constructed by introducing a domain attention mechanism based on the Transformer encoder. The data blocks after vector embedding are uniformly stored in an efficient vector database to build a private domain retrieval database, and an incremental monitoring and update mechanism is set up to monitor the addition and modification operations of private domain data; For the retrieval and sorting stage, specifically: Based on keyword matching or Embedding similarity recall mechanism, a multi-task intent recognition network is introduced to construct an intent-guided recall recognition network. For questions or requirements in the current requirements document to be reviewed, the constructed intent-guided recall recognition network determines whether it involves a recall intent for a specific requirement type. If so, a recall strategy is dynamically selected. Based on the question or requirement and combined with the cosine similarity calculation method, relevant data blocks are obtained in the private domain retrieval database to achieve data block recall. By introducing a semantic extension mechanism, the recall capability of hyponyms and hyponyms is enhanced based on the external general knowledge graph. The ranking score of the data block is calculated based on the ranking factor to finely sort the recalled data block to obtain the finely sorted data block. The calculation of the ranking score is specifically as follows: ； in, Indicates the The ranking score of the data blocks, Indicates the The semantic similarity between the data block and the query, Indicates the The timeliness score of each data block, Indicates the How often each data block is used or confirmed by experts, Indicates the The percentage of data blocks used in previous successful reviews, 、 、 、 represents the weight coefficient; For the prompt generation phase, specifically: For the sorted data blocks, a unified prompt template is constructed based on the issues in the current requirement document to be reviewed. The prompt template includes a requirement summary, referenced review rules, relevant historical review case snippets, and review focus points. Based on the carefully sorted data blocks, the context information of the demand review model is constructed. Based on the constructed context information, the demand review model responds to and reviews the questions or requirements in the current demand document to be reviewed according to the Prompt template, generates review results, and actively reviews and corrects the generated review results.

8. A device for realizing an intelligent quality valve in a demand submission process, characterized in that: The intelligent quality valve implementation device for the demand submission process includes: The requirements document writing module is used to obtain the requirements description in text or image form and upload it to the created large-scale intelligent platform. It then expands the document based on the comparison with existing similar requirements, performs standardization processing to fill in missing elements, and generates the requirements document. A requirements review summary module is used to obtain voice data from the requirements review meeting and upload it to the large model intelligent platform. The large model intelligent platform analyzes the voice data from the requirements review meeting to generate a summary of the requirements review meeting, and submits the summary of the requirements review meeting to the DevOps platform. The review processing execution module is used to create a demand review model. The demand review model obtains private domain data in real time through the retrieval enhancement technology of the retrieval enhancement generation model to review the current demand document to be reviewed. The private domain data includes existing demand documents, demand review meeting voice data, and demand review meeting summary.

9. An intelligent quality valve implementation device for the demand submission process, characterized in that: The intelligent quality valve implementation device for the demand submission process includes a processor, a memory, and an intelligent quality valve implementation program for the demand submission process stored on the memory and executable by the processor. When the intelligent quality valve implementation program for the demand submission process is executed by the processor, the steps of the intelligent quality valve implementation method for the demand submission process as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program for implementing an intelligent quality valve for a demand submission process, wherein when the program for implementing an intelligent quality valve for a demand submission process is executed by a processor, the steps of the method for implementing an intelligent quality valve for a demand submission process as described in any one of claims 1 to 7 are implemented.

Images