Application construction method and device, equipment, storage medium and product

Abstract

The invention discloses an application construction method and device, equipment, a storage medium and a product, and relates to the technical field of application construction. According to input user demand description data, an industry identifier is identified, and a demand knowledge graph is constructed; searching a target industry knowledge base according to the industry identifier so as to arrange the demand knowledge graph to obtain a target application design scheme; searching a software architecture design pattern library according to the target application design scheme, and outputting a target technical architecture scheme; and searching a code template library according to the target technical architecture scheme, and outputting a target application. According to the method, richer user demand information can be extracted from the user demand description data in various modes, a target application design scheme better meeting actual application demands is designed based on industry knowledge, and the quality and efficiency of application construction are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of application construction, and in particular to an application construction method, device, equipment, storage medium and product. BACKGROUND

[0002] The prior art can only generate an application based on input data when automatically generating an application software. When a user only provides limited requirement description data, the prior art cannot expand the user requirements, resulting in that the generated application cannot meet the user requirements, and even the generated target application is not complete enough, reducing the user experience and prolonging the application development cycle. SUMMARY

[0003] The embodiments of the present application provide an application construction method, device, equipment, storage medium and product, which can make the generated target application more in line with the actual requirements of the user by relying on industry knowledge base for requirement expansion and knowledge arrangement, and improve the quality and construction efficiency of the application.

[0004] To achieve the above object, the embodiments of the present application provide an application construction method, comprising:

[0005] According to the input user requirement description data, an industry identifier and a requirement knowledge graph are identified;

[0006] According to the industry identifier, a target industry knowledge base is searched to arrange the requirement knowledge graph, and a target application design scheme is obtained;

[0007] According to the target application design scheme, a software architecture design mode library is searched to output a target technology architecture scheme;

[0008] According to the target technology architecture scheme, a code template library is searched to output a target application.

[0009] As an improvement of the above-mentioned scheme, the user requirement description data includes at least two modal data, and the requirement knowledge graph is constructed by the following way:

[0010] The user requirement description data of each modal is respectively subjected to feature extraction and local fusion to obtain intermediate features of each modal;

[0011] According to the weight of each modal, the intermediate features are fused to obtain a requirement knowledge graph.

[0012] As an improvement of the above-mentioned scheme, according to the industry identifier, a target industry knowledge base is searched to arrange the requirement knowledge graph, and a target application design scheme is obtained, which comprises:

[0013] Searching the target industry knowledge base to supplement the demand knowledge graph and obtain a supplemented demand knowledge graph;

[0014] According to the target industry knowledge base, the supplemented demand knowledge graph is arranged into industry knowledge to obtain a target application design solution.

[0015] As an improvement to the above solution, the industry knowledge arrangement is performed on the demand knowledge graph to obtain a target application design solution, including:

[0016] Determine the target business based on the demand knowledge graph;

[0017] For any of the target businesses, generate a business logic processing flow;

[0018] Calling a first reward function to calculate a first reward value of the business logic processing flow;

[0019] When the first reward value does not meet the first iteration end condition, adjusting the business logic processing flow and returning to the step of calculating the first reward value to iterate;

[0020] When the first reward value meets the first iteration end condition, the current business logic processing flow is added to the target application design solution.

[0021] As an improvement to the above solution, searching a software architecture design pattern library according to the target application design solution and outputting a target technical architecture solution includes:

[0022] Select a reference architecture from the software architecture design pattern library based on the industry identifier and target application scale;

[0023] Adjusting the reference architecture to obtain a target technical architecture;

[0024] Generate target functional modules and their data flows according to the target application design plan;

[0025] Summarize the target technical architecture, each target functional module and the data flow of each target functional module, and output the target technical architecture solution.

[0026] As an improvement to the above solution, generating a target functional module and its data flow according to the target application design solution includes:

[0027] Generate a functional module partitioning scheme based on the target application design scheme; wherein the functional module partitioning scheme includes initial functional modules and their data flows;

[0028] The second reward function is called to calculate a second reward value of the functional module division scheme, wherein the second reward value is in positive correlation with at least one of a module cohesion degree, a module coupling degree, a cost and a user demand satisfaction degree;

[0029] When the second reward value does not meet a second iteration end condition, the functional module division scheme is adjusted, and the step of calculating the second reward value is iterated;

[0030] When the second reward value meets the second iteration end condition, an initial functional module and a data flow direction of the initial functional module in the current functional module division scheme are output as a target functional module and a data flow direction of the target functional module.

[0031] As an improvement of the above scheme, the target technical architecture scheme includes a target technical architecture, a target functional module and a data flow direction of the target functional module, and the target application is output by searching a code template library according to the target technical architecture scheme, which includes the following steps:

[0032] A target code template is selected from the code template library according to the target technical architecture and the target functional module;

[0033] A target parameter value is obtained from the target application design scheme;

[0034] The target code template is parameter-filled by using the target parameter value to generate a target code;

[0035] The target application is generated according to the target code.

[0036] To achieve the above object, an embodiment of the present application further provides an application construction device, which includes:

[0037] A demand knowledge graph construction module is configured to identify an industry identifier and construct a demand knowledge graph according to input user demand description data;

[0038] A target application design scheme generation module is configured to search a target industry knowledge base according to the industry identifier to arrange the demand knowledge graph and obtain a target application design scheme;

[0039] A target technical architecture scheme generation module is configured to search a software architecture design mode library according to the target application design scheme to output a target technical architecture scheme;

[0040] A target application generation module is configured to search a code template library according to the target technical architecture scheme to output a target application.

[0041] To achieve the above object, the embodiment of the present application further provides an application construction device, comprising a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the application construction method according to any one of the above embodiments when executing the computer program.

[0042] To achieve the above object, the embodiment of the present application further provides a computer readable storage medium, comprising a stored computer program, wherein the computer readable storage medium controls a device where the computer readable storage medium is located to execute the application construction method according to any one of the above embodiments when the computer program runs.

[0043] To achieve the above object, the embodiment of the present application further provides a computer program product, comprising computer programs / instructions, wherein the computer programs / instructions implement the application construction method according to any one of the above embodiments when executed by a processor.

[0044] Compared with the prior art, the application construction method, device, equipment, storage medium and product provided by the embodiment of the present application can extract user demand information from user demand description data of multiple modalities and construct a demand knowledge graph, so as to obtain more rich user demand information, then combine a target industry knowledge base to perform knowledge arrangement, can realize demand expansion, and by relying on industry knowledge, the target application design scheme can be more consistent with the actual application scene, finally, by searching a software architecture design mode library and a code template library to automatically generate a target application, the quality and efficiency of application construction are improved. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 is a flowchart of an application construction method provided by an embodiment of the present application;

[0046] Figure 2 is a flowchart of an application construction method provided by another embodiment of the present application;

[0047] Figure 3 is an architecture diagram of an application construction system provided by an embodiment of the present application;

[0048] Figure 4 is an interface schematic diagram of an application construction system provided by an embodiment of the present application;

[0049] Figure 5 is a structural schematic diagram of an application construction device provided by an embodiment of the present application;

[0050] Figure 6 is a structural schematic diagram of an application construction device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0051] With reference to the drawings and the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts are within the scope of the present application.

[0052] Referring to Figure 1 is a flowchart of an application construction method provided by an embodiment of the present application. Specifically, the application construction method comprises steps S1-S4.

[0053] S1, according to the input user requirement description data, identifying industry identification and constructing requirement knowledge graph;

[0054] S2, according to the industry identification, searching for a target industry knowledge base to arrange the requirement knowledge graph, and obtaining a target application design scheme;

[0055] S3, according to the target application design scheme, searching for a software architecture design pattern library to output a target technology architecture scheme;

[0056] S4, according to the target technology architecture scheme, searching for a code template library to output a target application.

[0057] In step S1, the user requirement description data refers to the data input by the user to describe the application to be constructed, which is used to indicate the user's requirements when constructing the target application. The user requirement description data can include various modal data, such as natural language text, document materials, pictures and videos. For example, in some embodiments, the natural language text is text data, such as a paragraph of text describing product functions, or one or more paragraphs of text explaining business process requirements, etc. The document materials can be PDF (Portable Document Format) or Word documents, etc. The document materials can record detailed information of the user's requirements, such as project specifications, technical documents or operation manuals, etc. which can contain the core points and key details of the user's requirements. The pictures can be product design drawings, scene diagrams or business process diagrams, etc. related to the user's requirements, which can contain product interface styles and layouts, etc. The video can be a product demonstration video or a business process demonstration video, etc. to show the requirement scenario or operation process.

[0058] Further, in step S1, the user demand description data is input into the demand knowledge graph construction module, which in the embodiment of the present application includes a multi-modal large model processing system and a demand analysis and industry determination module; wherein the multi-modal large model processing system includes a natural language processing module, a document parsing module, an image recognition module, a video analysis module, and an intelligent mapping and knowledge graph construction module. Further, the natural language processing module is used for processing the input natural language text, such as word segmentation, part-of-speech tagging, or semantic understanding, and identifying key sentences and semantic information; the document parsing module is used for extracting core points, key data, and logical structures in documents of various formats by using corresponding parsing algorithms; the image recognition module uses convolutional neural networks and other technologies to extract features and identify targets from input images to determine key elements and their attributes in the images; the video analysis module uses video key frame extraction technology and target tracking technology to analyze video content to identify key elements and dynamic information in the video. Further, the intelligent mapping and knowledge graph construction module uses intelligent mapping algorithms to map the information extracted by the above modules to a unified knowledge graph space to construct a demand knowledge graph.

[0059] It is worth noting that in the embodiment of the present application, the demand knowledge graph (also referred to as "demand system framework") is a demand system framework constructed by integrating, analyzing, and structuring fragmented information from different modalities. This framework is presented in the form of a knowledge graph, showing various key elements in user demand, the relationships between key elements, and the positions of key elements in the overall demand system, providing a basic framework for subsequent demand analysis, application design, and development.

[0060] Further, the demand analysis and industry determination module is used to identify industry identifiers from user demand description data. Specifically, by analyzing the user demand description data, the industry identifiers can be identified, for example, the industry identifiers can be medical, financial, industrial, and transportation, etc. If the industry identifier is not clear, the content of the user demand description data is analyzed in depth by natural language processing technology and image recognition technology to determine the industry to which it belongs.

[0061] Further, before step S2, it also includes collecting and organizing knowledge such as standard business processes, product demand descriptions, technical specifications, industry standards, and regulatory policies of different industries to construct an industry knowledge base. For example, the industry knowledge base uses a highly structured data storage method, uses professional database management technology to store information in a classified and hierarchical manner, and is equipped with an efficient search engine, so that the required knowledge can be accurately searched and called.

[0062] Exemplarily, in step S2, the demand knowledge graph constructed in step S1 can also be supplemented in reliance on the target industry knowledge base. For example, when the industry identifier is finance, the demand knowledge graph can be supplemented in combination of the user demand description data and the professional knowledge in the target industry knowledge base, such as the demand points of “credit risk assessment model construction” and “compliance review process embedding”, and when the industry identifier is medical treatment, the demand points of “medical record structured management” and “medical data encryption transmission and storage” and other necessary demand points of medical data security guarantee can be supplemented.

[0063] Further, in step S2, the corresponding industry knowledge base, i.e., the target industry knowledge base, is called according to the industry identifier, and the industry knowledge is arranged automatically by using the reasoning capability of the multi-modal large model in reliance on the target industry knowledge base, so as to form the target application design scheme.

[0064] Exemplarily, after the target application design scheme is generated, a user interaction interface can be built to collect user feedback opinions. For example, the user can evaluate whether the target application design scheme meets the expectation and whether there are omissions and other problems to generate feedback information. After the feedback information is collected, the system can record and classify the feedback information in real time, and use the feedback information to drive the multi-modal large model to optimize the industry knowledge again. The multi-modal large model analyzes the common problems and special demands in the feedback information, and adjusts the reasoning strategy and knowledge retrieval path of itself.

[0065] Further, before step S3, a software architecture design mode base is also constructed, which can include multiple types of architectures such as layered architecture, micro-service architecture and event-driven architecture. In step S3, the multi-modal large model screens and optimally combines the target technical architecture scheme from the software architecture design mode base according to factors such as demand scale, performance requirement and industry characteristics. Exemplarily, for an e-commerce application with high concurrency and high real-time requirement, a micro-service architecture can be used in combination with technologies such as cache and message queue, and for an internal management system of a small enterprise, a simple layered architecture can be selected.

[0066] Further, before step S4, a code template library is also established, wherein the code template library includes code templates of different technical architectures and functional modules, and the templates are presented in a structured and parameterized form, facilitating filling and adjustment according to specific requirements. In step S4, the multi-modal large model maps the target technical architecture scheme to the code template library, and automatically generates structured code language by using natural language generation technology in combination with specific parameter requirements in the target application design scheme. Then, the generated structured code language is deployed to a corresponding development environment, and is compiled and linked by using an automatic building tool, to generate an executable software program. In the deployment process, the configuration requirements of the development environment should be strictly followed to ensure that the program can stably run. Further, the built software program is tested, including functional testing and performance testing, by simulating different user scenarios and loading different magnitudes of data to ensure that the software meets the requirements of specific business use scenarios. In the performance testing, indicators such as software response time and throughput are monitored, and if the indicators do not meet the requirements, the multi-modal large model is fed back for optimization and adjustment.

[0067] In order to facilitate understanding of the flow of the application construction method described in the present application, see Figure 2 is a flowchart of an application construction method provided by another embodiment of the present application, which comprises Figure 2 It can be seen that at the beginning, the user submits user demand description data, which may include natural language text, document materials, pictures and videos and various modalities of data; then, the execution subject (service side) identifies demand information from the user demand description data, for example, identifies key sentences in natural language text and document materials, and deeply mines key elements in pictures and videos; then, the demand information is mapped to a unified knowledge graph space, for example, a demand knowledge graph can be obtained by using an intelligent mapping algorithm; then, an industry identifier is identified from the user demand description data, for example, the user demand description data can be subjected to deep semantic analysis, feature matching and the like to determine the industry identifier; then, a target application design scheme is designed based on the demand knowledge graph and relying on a target industry knowledge base; then, the target application design scheme is sent to the user side; after the user confirms the target application design scheme, a target technical architecture scheme is matched, for example, the matching can be performed based on multi-dimensional factors such as demand scale estimation, performance indicators and industry characteristics; then, a front-end interface component is rendered based on business logic, a back-end service module is designed, and a database is created; then, a code template library is mapped, and structured target code is automatically generated; then, the target code is deployed to a corresponding development environment; then, the user side tests the built target application; and the process ends after the testing is completed.

[0068] Compared with the prior art, the application construction method provided by the embodiment of the application can extract user demand information from user demand description data of multiple modalities and construct a demand knowledge graph, so that more rich user demand information is obtained, then knowledge arrangement is performed in combination with a target industry knowledge base, demand expansion can be performed, and by relying on industry knowledge, the target application design scheme can be made more in line with an actual application scene, finally, a target application is automatically generated by searching a software architecture design pattern library and a code template library, and the quality and efficiency of application construction are improved.

[0069] In an embodiment, the user demand description data includes data of at least two modalities, and the demand knowledge graph is constructed in the following manner:

[0070] Feature extraction and local fusion are respectively performed on the user demand description data of each modality to obtain intermediate features of each modality.

[0071] The intermediate features are fused according to the weights of the modalities to obtain a demand knowledge graph.

[0072] It should be noted that, in the joint representation of multi-modal fusion in the prior art, data of different modalities is simultaneously mapped to a unified feature space, although direct comparison and fusion of data are achieved, but this simple joint representation ignores the uniqueness of data of different modalities and the complex association between data, which is easy to cause information loss and poor fusion effect, and the distribution, dimension and noise level of data of different modalities are quite different, and direct unified representation is difficult to fully exploit the value of data of each modality.

[0073] Therefore, in the embodiment of the application, hierarchical feature fusion is performed on data of different modalities, that is, instead of simply mapping all modality data to the same feature space at one time, a hierarchical fusion manner is adopted, that is, preliminary feature extraction and local fusion are first performed on data of each modality to form intermediate features under each modality, then the intermediate features of different modalities are gradually fused into a unified high-level feature space in a hierarchical manner according to the characteristics of the intermediate features and the degree of association with user demand.

[0074] Compared with the prior art, the embodiment of the application can take into account the characteristics of data of different modalities when fusing data of different modalities, avoid loss of key information, and improve the quality of feature fusion.

[0075] Further, the embodiment of the application also introduces an attention mechanism, that is, in the process of multi-modal data fusion, the (attention) weights of each part of data of different modalities are calculated to highlight information that plays a key role in understanding and fusing demand.

[0076] Further, in some embodiments, the weight of each modality can also be dynamically adjusted according to actual conditions, that is, a dynamic weighting fusion strategy is adopted. For example, the weight can be adjusted in real time according to the reliability, information amount and relevance to the current user demand of the user demand description data of each modality. For example, if the information provided by the user in the natural language text is more detailed and accurate, and the image is relatively blurred or auxiliary, the weight of the text modality data can be increased. Furthermore, in the process of constructing the target application, some interactive links can be set to collect real-time feedback information of the user, so as to dynamically adjust the weight of the data of each modality according to the feedback information.

[0077] Compared with the prior art, the embodiments of the present application can highlight key information and avoid the influence of individual poor quality or insufficient correlation data on the fusion result by assigning appropriate weights to the user demand description data of different modalities and dynamically adjusting the weights when needed.

[0078] In one embodiment, the target industry knowledge base is searched according to the industry identifier to arrange the demand knowledge graph, and a target application design scheme is obtained.

[0079] The target industry knowledge base is searched to supplement the demand of the demand knowledge graph, and a supplemented demand knowledge graph is obtained.

[0080] According to the target industry knowledge base, the supplemented demand knowledge graph is arranged with industry knowledge, and a target application design scheme is obtained.

[0081] It is worth noting that the prior art has difficulty in accurately capturing subtle semantic differences and potential demands in complex development requests, and has limited understanding of ambiguous and implicit expressions. In the embodiments of the present application, in addition to using a multi-modal large model to understand and extract user demand description data, the demand knowledge graph is further supplemented by combining a target industry knowledge base, which can build a more comprehensive and complete user demand system by relying on industry knowledge and experience, thereby providing more reliable data support for the construction of a target application design scheme and improving the quality of application construction.

[0082] Further, the prior art often pre-establishes a knowledge base when using an application knowledge base, but the pre-established knowledge base lacks dynamic updating and flexible adaptation capabilities, and cannot track the latest industry knowledge.

[0083] Therefore, in some embodiments, instead of using a fixed industry knowledge base, a dynamic knowledge graph is constructed. After receiving a development request, the key fields and technical points in the data according to the user demand description are determined, and a deep learning-based graph matching algorithm is used, which not only considers the direct matching of knowledge graph nodes and edges, but also deeply excavates semantic similarity and logical association. For example, when processing a code development request related to a new artificial intelligence algorithm, through dynamic knowledge graph construction and matching, the latest research results and application cases of the algorithm can be quickly obtained and accurately matched with the request, instead of being limited to the outdated knowledge in the preset knowledge base.

[0084] Compared with the prior art, the embodiments of the present application can obtain the latest industry knowledge by dynamically updating or constructing an industry knowledge base, thereby improving the reliability and practicality of the target application design scheme and improving the application construction quality.

[0085] In one embodiment, the industry knowledge arrangement of the demand knowledge graph to obtain a target application design scheme comprises:

[0086] According to the demand knowledge graph, a target business is determined;

[0087] For any target business, a business logic processing flow is generated;

[0088] A first reward function is called to calculate a first reward value of the business logic processing flow;

[0089] When the first reward value does not meet a first iteration end condition, the business logic processing flow is adjusted, and the step of calculating the first reward value is iterated;

[0090] When the first reward value meets the first iteration end condition, the current business logic processing flow is added to the target application design scheme.

[0091] For example, when generating the business logic processing flow, a reinforcement learning mechanism is introduced, the generated business logic processing flow is regarded as a strategy, and a series of first reward functions related to the actual business scenario are set, such as the execution efficiency of the flow, the resource utilization rate, and the degree of fit with the business demand, etc. Let the large model constantly optimize the strategy according to the reward feedback (first reward value) in the process of generating the business logic processing flow, and gradually generate a target application design scheme that better meets the actual business demand. For example, the first iteration end condition can be that the first reward value is greater than a preset threshold, or it can be other, which is not limited here

[0092] Compared with the prior art which can only mechanically apply the existing business processing flow, the embodiment of the application can generate more diversified business logic processing flows and better adapt to the complex requirements of actual business scenarios, and has stronger universality and scalability, by setting the first reward function to guide the optimization and adjustment of the business logic processing flow.

[0093] In an implementation manner, the searching the software architecture design mode library according to the target application design scheme and outputting the target technology architecture scheme comprises:

[0094] selecting a reference architecture from the software architecture design mode library according to the industry identifier and the target application scale;

[0095] adjusting the reference architecture to obtain a target technology architecture;

[0096] generating target function modules and data flow directions of the target function modules according to the target application design scheme;

[0097] summarizing the target technology architecture, the target function modules and the data flow directions of the target function modules, and outputting a target technology architecture scheme.

[0098] It is worth noting that the target technology architecture scheme includes architecture composition, function responsibilities of the target function modules, interaction relationships between the target function modules, data flow directions and key technology selection, etc.

[0099] Specifically, based on the target application design scheme and the software architecture design mode library, the reasoning capability of the model and the understanding of various design modes are used to perform technology architecture design, and meanwhile, natural language processing technology, data analysis technology and pattern matching and other technologies are integrated to realize processing of the target application design scheme and generation of the target technology architecture.

[0100] In an implementation manner, the generating target function modules and data flow directions of the target function modules according to the target application design scheme comprises:

[0101] generating a function module division scheme according to the target application design scheme; wherein the function module division scheme includes initial function modules and data flow directions of the initial function modules;

[0102] calculating a second reward value of the function module division scheme by calling a second reward function; wherein the second reward value is in a positive correlation with at least one of a module cohesion degree, a module coupling degree, a cost and a user demand satisfaction degree;

[0103] when the second reward value does not meet a second iteration end condition, adjusting the function module division scheme and returning to the step of calculating the second reward value for iteration;

[0104] When the second reward value meets the second iteration end condition, an initial function module and a data flow direction thereof in a current function module division scheme are output as a target function module and a data flow direction thereof.

[0105] Exemplarily, a reinforcement learning framework can be introduced, and a module division agent can be constructed to implement function module division. Compared with the prior art, the embodiment of the application can realize optimal division of function modules by using a second reward function to guide function module division.

[0106] In an implementation, the target technical architecture scheme includes a target technical architecture, a target function module, and a data flow direction of the target function module, and the searching for a code template library according to the target technical architecture scheme and outputting a target application include:

[0107] selecting a target code template from the code template library according to the target technical architecture and the target function module;

[0108] obtaining a target parameter value from the target application design scheme;

[0109] performing parameter filling on the target code template by using the target parameter value to generate target code;

[0110] generating a target application according to the target code.

[0111] Exemplarily, the code template library can include a front-end interface component template, a business logic module template, and a database table structure template. The front-end interface component template includes code templates of common components such as buttons, text boxes, lists, and charts in a structured and parameterized form, and the button template can set parameters such as text content, color, and click events. The business logic module template is a code template corresponding to a function module obtained by dividing a complex business according to a modular design concept, and includes a specific business logic framework and interface definition, which can be filled and adjusted according to specific business requirements. The database table structure template is designed according to the theory of data relationship for different data association scenarios such as one-to-one, one-to-many, and many-to-many, and includes field definition and primary and foreign key settings.

[0112] Specifically, a target code template is selected from the code template library and filled with parameters according to specific parameter requirements of a target application design scheme and a target technical architecture scheme to generate complete and structured target code, and the target code includes front-end code, business logic code, and database code. Exemplarily, template matching technology, parameter analysis technology, and code synthesis technology can be integrated to implement complex code generation tasks.

[0113] Exemplarily, in some embodiments, template intelligent selection can be performed based on transfer learning. In the template library, each template is labeled with features, including applicable front-end components, back-end services, and database structures. Using transfer learning techniques, accumulated template selection experience in similar business scenarios and technical fields is transferred to new technical solutions. For new solution requirements, a subset of templates with high similarity is first selected from the template library, and then a deep learning model is used to match the technical solution and template features. After comprehensive evaluation, the most suitable template is selected.

[0114] Further, after generating the target code, the target code is deployed to a development environment, and an automatic build tool is used to perform compilation and linking, etc., to generate an executable program. Then, test tasks such as function testing and performance testing are performed, different user operation scenarios and operation data are simulated according to the business process, and the software is tested in the whole process. Finally, according to the test conclusion, the function modules are optimized and adjusted, and then the function self-test is performed again, until the software runs normally on the platform, and the target application is obtained.

[0115] In order to more clearly understand the application construction method, the present application is described below by developing a medical health management application as an example, including the following steps one to step five:

[0116] Step one: the user uploads user demand description data including medical record document, patient rehabilitation video, doctor's natural language description of application function demand and various modalities. The multi-modal large model identifies and analyzes the text, objects, scenes, etc. in the picture through image recognition technology, such as identifying lesions in medical images, identifying patient health status in rehabilitation videos, etc. Finally, a demand knowledge graph including patient diagnosis and treatment process, rehabilitation points and doctor's expected function demand information may be constructed.

[0117] Step two: use knowledge graph technology to build the association between medical industry knowledge, and display the knowledge structure in a graphical way, so as to facilitate the system to understand the knowledge system, discover potential knowledge association, and assist decision analysis. Combine with user demand to identify industry sub-scene identification, add medical insurance reimbursement accounting and drug inventory management links, etc., to realize more accurate application system design.

[0118] Step three: system architecture analysis, design and management, based on intent analysis to understand user's demand and intent for system architecture, and design reasonable technical architecture scheme according to demand, store, call and optimize common architecture template, and then design target application design scheme suitable for application demand, including patient management, diagnosis and treatment service and data statistics, etc. Multiple micro-service modules.

[0119] Step four: based on template matching and parameter analysis, etc. technology, automatically generate the code that meets the needs. The system uses the built-in software architecture design pattern library to analyze and process the user input parameters, and automatically recommends and optimizes the combination of the most suitable code generation templates according to the demand scale, performance requirements and industry characteristics, etc. Use the code template library to generate structured code language covering the front-end operation interface, business logic processing and database table structure. The code fragments of each part are combined into complete and usable code.

[0120] Step five: code deployment and software testing, packaging and configuring the developed application program to make it run in the target environment, testing the deployed application in terms of function, performance and compatibility, etc. to ensure its stability and reliability, and optimizing and improving the function of the application according to the test results and user feedback, to get the application program that meets the medical health management business scenario.

[0121] Further, referring to Figure 3 is the architecture diagram of the application construction system provided by an embodiment of the present application. By Figure 3 It can be seen that the platform base containing digital base, AI (Artificial Intelligence, artificial intelligence) large model, intelligent agent and trusted security is built, and high-efficiency data processing and model operation are realized with the help of cloud network computing power support; It has the ability to interface with external systems, including ERP (Enterprise Resource Planning, enterprise resource planning) system, financial management system, inventory management system, image data platform and location analysis platform, etc. Through unified interface specification and data interaction protocol, smooth data transmission and business cooperation between multiple systems are ensured, and a complete ecological application system is built. Further, the industry application can be finance, industry and transportation, etc., and a dedicated industry knowledge base and knowledge graph are built. Further, through knowledge classification, knowledge update, knowledge graph and knowledge engine, precise knowledge services are provided for industry users; combined with the characteristics of each industry, core capability modules such as demand import, document processing and picture recognition are customized, realizing the deep customization and intelligent upgrading of industry application. Further, the software development scheme is based on architecture engine for system architecture design, understands user needs through intent analysis, and realizes rapid architecture building through template management. Further, code generation technology is adopted, and according to template matching, parameter analysis and code synthesis, code that meets business needs is automatically generated. In the application deployment stage, the deployment construction, application testing and function optimization process are covered to ensure the stable operation and continuous improvement of the software, and through RAG (Retrieval-Augmented Generation, retrieval augmented generation) enhancement and plug-in management, the flexible expansion of software function is supported.

[0122] Further, referring to Figure 4Fig. 1 is a schematic diagram of an interface of an application construction system according to an embodiment of the present application, wherein Figure 4 Fig. 1 shows a user interface in an application construction process, and divides the application construction process into requirement analysis, architecture design, component generation and application deployment, and the user can see the progress of the application construction from the interface.

[0123] Compared with the prior art, the application construction method provided by the embodiment of the present application can extract user requirement information from user requirement description data of multiple modalities and construct a requirement knowledge graph, so as to obtain more abundant user requirement information, then combine a target industry knowledge base to perform knowledge arrangement, can perform requirement expansion, and by relying on industry knowledge, the target application design scheme can be made more in line with the actual application scene, finally, the target application is automatically generated by searching a software architecture design mode library and a code template library, and the quality and efficiency of the application construction are improved.

[0124] Referring to Figure 5 The embodiment of the present application also provides an application construction device 10, comprising:

[0125] A requirement knowledge graph construction module 11 is configured to identify an industry identifier and construct a requirement knowledge graph according to input user requirement description data;

[0126] A target application design scheme generation module 12 is configured to search a target industry knowledge base according to the industry identifier, to arrange the requirement knowledge graph, and obtain a target application design scheme;

[0127] A target technology architecture scheme generation module 13 is configured to search a software architecture design mode library according to the target application design scheme, and output a target technology architecture scheme;

[0128] A target application generation module 14 is configured to search a code template library according to the target technology architecture scheme, and output a target application.

[0129] In one embodiment, the user requirement description data includes data of at least two modalities, and the requirement knowledge graph construction module 11 is specifically configured to:

[0130] Perform feature extraction and local fusion on the user requirement description data of each modality respectively, to obtain intermediate features of each modality;

[0131] Fuse the intermediate features according to the weights of the modalities, to obtain a requirement knowledge graph.

[0132] In one embodiment, the target application design scheme generation module 12 is specifically configured to:

[0133] search a target industry knowledge base to supplement the demand knowledge graph, and obtain a supplemented demand knowledge graph;

[0134] According to the target industry knowledge base, the supplemented demand knowledge graph is arranged with industry knowledge, and a target application design scheme is obtained.

[0135] In an implementation, the target application design scheme generation module 12 is further configured to:

[0136] According to the demand knowledge graph, a target business is determined;

[0137] For any target business, a business logic processing flow is generated;

[0138] A first reward function is called to calculate a first reward value of the business logic processing flow;

[0139] When the first reward value does not meet a first iteration end condition, the business logic processing flow is adjusted, and the step of calculating the first reward value is iterated;

[0140] When the first reward value meets the first iteration end condition, the current business logic processing flow is added to the target application design scheme.

[0141] In an implementation, the target technical architecture scheme generation module 13 is specifically configured to:

[0142] According to the industry identifier and the target application scale, a reference architecture is selected from a software architecture design mode library;

[0143] The reference architecture is adjusted to obtain a target technical architecture;

[0144] According to the target application design scheme, a target functional module and a data flow direction thereof are generated;

[0145] The target technical architecture, each target functional module, and the data flow direction of each target functional module are summarized, and a target technical architecture scheme is output.

[0146] In an implementation, the target technical architecture scheme generation module 13 is further configured to:

[0147] According to the target application design scheme, a functional module division scheme is generated; wherein the functional module division scheme includes an initial functional module and a data flow direction thereof;

[0148] A second reward function is called to calculate a second reward value of the functional module division scheme; wherein the second reward value is positively correlated with at least one of a module cohesion degree, a module coupling degree, a cost, and a user demand satisfaction degree;

[0149] when the second reward value does not meet the second iteration end condition, adjusting the function module division scheme, and returning to the step of calculating the second reward value for iteration;

[0150] when the second reward value meets the second iteration end condition, outputting the initial function module and the data flow direction thereof in the current function module division scheme as the target function module and the data flow direction thereof.

[0151] In an implementation, the target application generation module 14 is specifically configured to:

[0152] select a target code template from a code template library according to the target technical architecture and the target function module;

[0153] obtain a target parameter value from the target application design scheme;

[0154] perform parameter filling on the target code template by using the target parameter value, to generate target code;

[0155] generate a target application according to the target code.

[0156] The application construction apparatus provided by the embodiments of the present application can implement all process steps of the application construction method described in the above embodiments, the functions of each module and unit in the apparatus, and the technical effects achieved by the functions are respectively the same as the functions of the application construction method described in the above embodiments and the technical effects achieved by the functions, and thus the specific implementation manners are not described here.

[0157] Referring to Figure 6 , the embodiments of the present application further provide an application construction device 20, which includes a processor 21, a memory 22, and a computer program stored in the memory 22 and configured to be executed by the processor 21, and the processor 21 implements the steps in the above application construction method embodiments when executing the computer program, such as the steps S1-S4 described in Figure 1 , or the processor 21 implements the functions of each module in the above apparatus embodiments when executing the computer program.

[0158] The application construction device can be a desktop computer, a notebook computer, a palm computer, a cloud server, and the like. The application construction device can include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the schematic diagram is only an example of the application construction device, and does not constitute a limitation on the application construction device, and can include more or fewer components than the diagram, or combine certain components, or different components, for example, the application construction device can also include an input / output device, a network access device, a bus, and the like.

[0159] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The processor is a control center of the application construction device, and connects various parts of the application construction device through various interfaces and lines.

[0160] The memory can be used to store the computer programs and / or modules, and the processor realizes various functions of the application construction device by running or executing the computer programs and / or modules stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area. The program storage area can store an operating system, at least one application program required by a function, etc.; and the data storage area can store data created according to the use of the controller, etc. In addition, the memory can include a high-speed random access memory, and can also include a nonvolatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory device.

[0161] The application construction device integrated module can be stored in a computer readable storage medium if it is implemented in the form of a software function unit and sold or used as an independent product. Based on this understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. The computer program can implement the steps of each method embodiment when executed by a processor. The computer program includes computer program code, which can be in the form of source code, object code, an executable file, or some intermediate form. The computer readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0162] To achieve the above-mentioned purposes, the embodiments of the present application also provide a computer program product, including computer programs / instructions, which are executed by a processor to realize the application construction method as described in any of the above-mentioned embodiments.

[0163] Compared with the prior art, the application construction device, equipment, storage medium and product provided by the embodiments of the present application can extract user demand information from user demand description data of multiple modalities and construct a demand knowledge graph, thereby obtaining more rich user demand information, then combining a target industry knowledge base to perform knowledge arrangement, which can expand the demand, and by relying on the industry knowledge, the target application design scheme can be more consistent with the actual application scene, finally, the target application is automatically generated by searching a software architecture design mode library and a code template library, thereby improving the quality and efficiency of application construction.

[0164] The above-mentioned is the preferred embodiment of the present application, and it should be pointed out that for ordinary skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which are also considered to be within the scope of protection of the present application.

Claims

1. An application construction method, characterized in that: include: Based on the input user demand description data, identify industry identifiers and build demand knowledge graphs; Searching the target industry knowledge base according to the industry identifier to compile the demand knowledge graph and obtain a target application design solution; Search the software architecture design pattern library according to the target application design solution and output the target technical architecture solution; Search the code template library according to the target technical architecture solution and output the target application.

2. The application construction method according to claim 1, wherein: The user demand description data includes data of at least two modalities, and the demand knowledge graph is constructed in the following manner: Performing feature extraction and local fusion on the user demand description data of each modality to obtain intermediate features of each modality; According to the weight of each modality, the intermediate features are fused to obtain the demand knowledge graph.

3. The application construction method according to claim 1, wherein: The step of searching the target industry knowledge base according to the industry identifier to compile the demand knowledge graph and obtain a target application design solution includes: Searching the target industry knowledge base to supplement the demand knowledge graph and obtain a supplemented demand knowledge graph; According to the target industry knowledge base, the supplemented demand knowledge graph is arranged into industry knowledge to obtain a target application design solution.

4. The application construction method according to claim 1, wherein: The industry knowledge arrangement of the demand knowledge graph to obtain a target application design solution includes: Determine the target business based on the demand knowledge graph; For any of the target businesses, generate a business logic processing flow; Calling a first reward function to calculate a first reward value of the business logic processing flow; When the first reward value does not meet the first iteration end condition, adjusting the business logic processing flow and returning to the step of calculating the first reward value to iterate; When the first reward value meets the first iteration end condition, the current business logic processing flow is added to the target application design solution.

5. The application construction method according to claim 1, wherein: The step of searching a software architecture design pattern library according to the target application design solution and outputting a target technical architecture solution includes: Select a reference architecture from the software architecture design pattern library based on the industry identifier and target application scale; Adjusting the reference architecture to obtain a target technical architecture; Generate target functional modules and their data flows according to the target application design plan; Summarize the target technical architecture, each target functional module and the data flow of each target functional module, and output the target technical architecture solution.

6. The application construction method according to claim 5, wherein: According to the target application design scheme, generating the target functional module and its data flow includes: Generate a functional module partitioning scheme based on the target application design scheme; wherein the functional module partitioning scheme includes initial functional modules and their data flows; Calling a second reward function to calculate a second reward value for the functional module division scheme; wherein the second reward value is positively correlated with at least one of the degree of module cohesion, the degree of inter-module coupling, the cost, and the degree of satisfaction of user needs; When the second reward value does not meet the second iteration end condition, adjusting the functional module division scheme, and returning to the step of calculating the second reward value to iterate; When the second reward value meets the second iteration end condition, the initial functional module and its data flow in the current functional module division scheme are output as the target functional module and its data flow.

7. The application construction method according to claim 1, wherein: The target technical architecture solution includes a target technical architecture, a target functional module, and a data flow of the target functional module. Then, searching a code template library according to the target technical architecture solution and outputting a target application includes: Selecting a target code template from a code template library according to the target technical architecture and the target functional module; Obtaining target parameter values ​​from the target application design solution; Filling the target code template with the target parameter value to generate the target code; A target application is generated according to the target code.

8. An application construction device, characterized in that: include: The demand knowledge graph construction module is used to identify industry identifiers and construct a demand knowledge graph based on the input user demand description data; A target application design solution generation module is used to search the target industry knowledge base according to the industry identifier to compile the demand knowledge graph and obtain a target application design solution; A target technical architecture solution generation module is used to search the software architecture design pattern library according to the target application design solution and output the target technical architecture solution; The target application generation module is used to search the code template library according to the target technical architecture solution and output the target application.

9. An application building device, characterized in that: The invention comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the application construction method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the application construction method according to any one of claims 1 to 7.

11. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the application construction method according to any one of claims 1 to 7 is implemented.

Images