Development method and development platform of low-code application, electronic equipment and storage medium

Abstract

The invention provides a low-code application development method and platform, electronic equipment and a storage medium, in the low-code application development method, natural language data is utilized to train a meta-language programming model, and a low-code application development request submitted by adopting a natural language is responded through the meta-language programming model, so that the low-code application development efficiency is improved. Carrying out meta-language feature extraction processing on the low-code application development request to obtain a field specification language code described by adopting a meta-language, and extracting a functional component and / or a business component matched with the field specification language code from a code component library, or extracting a functional component and / or a business component matched with the field specification language code from a code component library; according to the method, functional components and / or business components matched with the domain specification language codes are extracted from a code component library, corresponding application software is generated through splicing, the extracted content is displayed in a visual mode, the function of quickly and accurately generating low-code applications through natural language description requirements is achieved, operation is easy, application is convenient, and the method is suitable for popularization and application. According to the method, the development efficiency is improved, particularly, the skill requirement for the programming technology of a user is reduced, and the method has wider applicability.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a low-code application development method, development platform, electronic device, and computer-readable storage medium. Background Art

[0002] Software development is a systematic project encompassing requirements capture, analysis, design, implementation, and testing. Application software is typically implemented using a programming language and typically developed using software development tools. However, conventional software development tools require specialized technical personnel with programming skills. The entire software development process consumes significant time and effort, resulting in low development efficiency, high development costs, and the potential for code clutter and maintenance difficulties.

[0003] Low-code development has become a new trend in application development in recent years. The current low-code development platform (LCDP) mainly uses specific framework specifications and visual operation interfaces to enable developers with different experience levels to quickly develop application functions through drag-and-drop and function configuration interfaces in the visual interface, lowering the development threshold and improving development efficiency.

[0004] However, as an ordinary developer, it is easy to be confused by the different interaction design concepts and specifications of different platforms. Existing low-code platforms still have some shortcomings. For example, when developers use low-code platforms for application development, they still need to have certain programming knowledge and operational skills to correctly select and combine components; moreover, when the developer's needs are more complex, the existing low-code platforms have difficulty in accurately understanding the user's intentions and cannot meet the user's personalized needs. Different low-code application development platforms have different design concepts. Their respective operations are completely dependent on the design of human-computer user interface interaction. The operations have no clear semantics, and the usage methods of different platforms are very different. It is still difficult to learn and master the development steps of a platform.

[0005] Furthermore, with the development of artificial intelligence (AI), natural language processing (NLP) technology has matured, leading to the emergence of large language models (LLMs). Among these models, GPT (Generative Pre-trained Transformers) code models, which are generated from text, can generate code in a specific language based on natural language requirements, such as using Java or Go to implement a specific function or algorithm. However, due to the uncertainty inherent in fuzzy logic, fuzzy reasoning, and probability theory in AI, the code generated by current code models cannot guarantee the correctness required by computer programs and is difficult to verify. Summary of the Invention

[0006] In view of the shortcomings of the related technologies described above, the purpose of the present disclosure is to provide a low-code application development method, development platform, electronic device, and computer-readable storage medium to solve various problems in the related technologies.

[0007] The present disclosure provides a method for developing a low-code application, comprising the following steps:

[0008] In response to a low-code application development request submitted in natural language, semantically parse the natural language in the low-code application development request using a metalanguage programming model and perform metalanguage feature extraction processing based on the results of the semantic parsing to obtain a corresponding domain specification language code; wherein the metalanguage programming model is obtained through training with conversation data, and the domain specification language code is described in a metalanguage, which is constructed based on a high-level computer language;

[0009] Based on the domain specification language code, functional components and / or business components that match the domain specification language code are extracted from a code component library, or functional components and / or business components that match the domain specification language code are extracted from a code component library and assembled to generate corresponding application software; the functional components and / or business components are written in a high-level computer language; and

[0010] The extracted functional components and / or business components, or the generated application software are displayed in a visual manner.

[0011] In certain examples of the first aspect, the low-code application development method also includes the step of creating a metalanguage, which adopts the syntax of structured query language, is based on the grammatical structure of natural language and incorporates the logic requirements of computer high-level language programming, defines the sentence composition method, the combination rules between words, and the expression method of logical relationships, so that natural language can be converted into computer high-level language through metalanguage.

[0012] In certain examples of the first aspect, the low-code application development method also includes a step of writing domain specification language code through a meta-language, and the domain specification language code is associated with functional components and / or business components in the code component library.

[0013] In certain examples of the first aspect, the training of the metalanguage programming model includes the following steps: training a basic metalanguage programming model using training data, and fine-tuning the metalanguage programming model based on the training results to obtain a fine-tuned metalanguage programming model; testing and evaluating the fine-tuned metalanguage programming model using test data, and determining the fine-tuned metalanguage programming model that meets the evaluation requirements as a formed metalanguage programming model; wherein the training data and the test data are conversation data formed by filling data extracted from a material library into a conversation template, and the data in the material library includes structural data extracted from a business database, and / or structural data generated by a large language model, and / or structural data captured through the network, and / or structural data collected by document scanning.

[0014] In certain examples of the first aspect, a basic metalanguage programming model is trained using training data, and the metalanguage programming model is fine-tuned based on the training results to obtain a fine-tuned metalanguage programming model; the fine-tuned metalanguage programming model and the basic metalanguage programming model are fused to form a fused metalanguage programming model; the fused metalanguage programming model is tested and evaluated using test data, and the fused metalanguage programming model that meets the evaluation requirements is determined to be a formed metalanguage programming model; wherein the training data and the test data are conversation data formed by filling data extracted from a material library into a conversation template, and the data in the material library includes structural data extracted from a business database, and / or structural data generated by a large language model, and / or structural data captured through the network, and / or structural data collected by document scanning.

[0015] In certain examples of the first aspect, a basic metalanguage programming model is trained using training data, and the metalanguage programming model is fine-tuned based on the training results to obtain a fine-tuned metalanguage programming model, including the following steps: selecting at least one large language model as a basic metalanguage programming model and obtaining training data; calling a fine-tuning framework to input the training data into the basic metalanguage programming model for training and fine-tuning; during the fine-tuning process, observing indicators of the basic metalanguage programming model and test results of given examples, and fine-tuning parameters in the basic metalanguage programming model based on the observed results.

[0016] A second aspect of the present disclosure provides a low-code application development platform, including:

[0017] A component packaging module is configured to package codes written in a high-level computer language into assembleable functional components and / or business components, and store the functional components and / or business components in a code component library;

[0018] A human-computer interaction module is configured to implement human-computer interaction and is used to obtain low-code application development requests submitted in natural language through a conversation;

[0019] A metalanguage programming model is configured to perform semantic parsing on the natural language in the low-code application development request and perform metalanguage feature extraction processing based on the results of the semantic parsing to obtain corresponding domain specification language code; wherein the metalanguage programming model is trained using conversational data, the domain specification language code is described using a metalanguage, and the metalanguage is constructed based on a high-level computer language;

[0020] The low-code engine is configured to: extract functional components and / or business components that match the domain specification language code from a code component library based on the domain specification language code, or extract functional components and / or business components that match the domain specification language code from the code component library and generate corresponding application software by assembling them; wherein the functional components and / or business components are written in a high-level computer language; and

[0021] The visualization environment module is configured to provide a visualization interface, through which the functional components and / or business components extracted by the low-code engine, or the generated application software, are displayed in a visual manner.

[0022] In some examples of the second aspect, the low-code application development platform also includes a domain specification language code writing module, which is configured to write domain specification language code through a metalanguage.

[0023] A third aspect of the present disclosure provides an electronic device comprising a memory, a processor, and a computer program stored on the memory, wherein the processor executes the computer program to implement the steps of the low-code application development method as described above.

[0024] A fourth aspect of the present disclosure provides a computer-readable storage medium, which stores a computer program, and is characterized in that when the computer program is executed by a processor, it implements the steps of the low-code application development method as described above.

[0025] As described above, the present disclosure provides a development method, a development platform, an electronic device, and a computer-readable storage medium for low-code applications, wherein the development method of the low-code application uses natural language data to train a metalanguage programming model, and a metalanguage is used in the metalanguage programming model, and the metalanguage is constructed based on a computer high-level language. In this way, the metalanguage programming model responds to a low-code application development request submitted in natural language, performs semantic parsing on the natural language in the low-code application development request, and performs metalanguage feature extraction processing based on the result of the semantic parsing, obtains the domain specification language code described in the metalanguage, and then directly extracts the functional components and / or business components that match the domain specification language code from the code component library, or extracts the functional components and / or business components that match the domain specification language code from the code component library and generates the corresponding application software by assembling them, and then displays the extracted content in a visual manner. In this way, the function of quickly and accurately generating low-code applications by describing requirements in natural language is realized, the operation is simple, the application is convenient, the development efficiency is improved, and in particular, the skill requirements for the user's programming technology are reduced, and it has a wider applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 Shown is a schematic diagram of an application scenario in one embodiment of a low-code application development method provided by the present disclosure.

[0027] Figure 2 A flowchart illustrating a method for developing a low-code application provided by the present disclosure in one embodiment is shown.

[0028] Figure 3 Shown is a schematic diagram of the language structure of the metalanguage in an example according to an embodiment of the present disclosure.

[0029] Figure 4 A schematic diagram of a training process for the metalanguage programming model provided by the present disclosure in one embodiment is shown.

[0030] Figure 5 Shown as per Figure 4 Schematic diagram of training the metalanguage programming model shown.

[0031] Figure 6 A schematic diagram showing a flow chart of training the metalanguage programming model provided by the present disclosure in another embodiment.

[0032] Figure 7 Shown as per Figure 6 Schematic diagram of training the metalanguage programming model shown

[0033] Figure 8 Shown is a structural diagram of an embodiment of a low-code application development platform provided by the present disclosure.

[0034] Figure 9 Shown is a functional block diagram of an electronic device provided by the present disclosure in one embodiment. DETAILED DESCRIPTION

[0035] The following describes the embodiments of the present disclosure through specific examples. Those skilled in the art can easily understand the other advantages and effects of the present disclosure from the information disclosed in this disclosure. The present disclosure can also be implemented or applied through different specific embodiments. The details of the present disclosure can also be modified or changed according to different viewpoints and application modules without departing from the spirit of the present disclosure. It should be noted that the embodiments and features in the embodiments of the present disclosure can be combined with each other unless there is a conflict.

[0036] The following is a detailed description of the embodiments of the present disclosure with reference to the accompanying drawings so that those skilled in the art can easily implement the present disclosure. The present disclosure can be embodied in many different forms and is not limited to the embodiments described herein.

[0037] Throughout the present disclosure, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present disclosure. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or a group of embodiments or examples. Furthermore, those skilled in the art may combine and integrate different embodiments or examples, and features of different embodiments or examples, as described in the present disclosure, without conflicting requirements.

[0038] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the context of this disclosure, "a group" means two or more, unless otherwise specifically defined.

[0039] In order to clearly describe the present disclosure, components not related to the description are omitted, and the same or similar components throughout the specification are denoted by the same reference numerals.

[0040] Throughout this specification, when a device is said to be "connected" to another device, this includes not only "direct connection" but also "indirect connection" with other elements interposed therebetween. Furthermore, when a device is said to "include" a certain component, unless otherwise stated, this does not exclude the inclusion of other components but rather implies that the device may include other components.

[0041] Although the terms first, second, etc. are used in this document to represent various elements in some examples, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, the first interface and the second interface, etc. are represented. Furthermore, as used in this document, the singular forms "one," "an," and "the" are intended to also include the plural forms, unless there is a contrary indication in the context. It should be further understood that the terms "comprise" and "include" indicate the presence of the described features, steps, operations, elements, modules, projects, types, and / or groups, but do not exclude the presence, occurrence, or addition of one or a group of other features, steps, operations, elements, modules, projects, types, and / or groups. The terms "or" and "and / or" used herein are interpreted as inclusive, or mean any one or any combination. Therefore, "A, B, or C" or "A, B, and / or C" means "any of the following: A; B; C; A and B; A and C; B and C; A, B, and C." Exceptions to this definition only occur when the combination of elements, functions, steps, or operations is inherently mutually exclusive in some way.

[0042] The technical terms used herein are intended only to refer to specific embodiments and are not intended to limit the present disclosure. The singular form used herein also includes the plural form unless the statement explicitly indicates otherwise. The term "comprising" as used in this specification is intended to specify specific features, regions, integers, steps, operations, elements, and / or components and does not exclude the presence or addition of other features, regions, integers, steps, operations, elements, and / or components.

[0043] Although not defined differently, all terms used herein, including technical and scientific terms, have the same meanings as those generally understood by those skilled in the art to which this disclosure belongs. Terms defined in commonly used dictionaries are additionally interpreted as having meanings consistent with relevant technical literature and the current message. Unless otherwise defined, they should not be overly interpreted as ideal or highly formalized meanings.

[0044] Among the related technologies involved in low-code application development, the low-code development platform is a visual development tool that allows developers to use a visual operation interface to drag and drop and configure functions instead of writing code to build application software. Therefore, low-code application development has become a new trend. However, when developers use low-code platforms for application development, they still need to have certain programming knowledge and operation skills to correctly select and combine components. In addition, when the developer's needs are more complex, the existing low-code platform is difficult to accurately understand the user's intentions and cannot meet the user's personalized needs.

[0045] The embodiments of the present disclosure provide a low-code application development method and development platform, an electronic device, and a computer-readable storage medium, which can respond to low-code application development requests submitted by users in natural language, and convert the natural language low-code application development requests into domain specification language codes described by a set meta-language. The meta-language can be converted with a computer high-level language. In this way, functional components and / or business components that match the domain specification language codes can be directly driven and extracted, or functional components and / or business components that match the domain specification language codes can be extracted from a code component library and the corresponding application software can be generated by assembling them, thereby realizing the function of quickly and accurately generating low-code applications by describing requirements in natural language.

[0046] The present disclosure will be described in more detail through the following examples.

[0047] Figure 1 Shown is a schematic diagram of an application scenario in one embodiment of a low-code application development method provided by the present disclosure.

[0048] In such Figure 1 The illustrated application scenario includes at least one terminal device 100 and a server 200 communicatively connected to the at least one terminal device 100 via a network. The network can be a wired or wireless network, using standard communication technologies and / or protocols. The network is typically the Internet, but can also be any network, including but not limited to a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a mobile communication network (such as a 3G communication network, an LTE communication network, or a 5G communication network), a private network, a virtual private network, and any combination thereof.

[0049] Users can use terminal devices 100 to interact with server 200 over the network to receive or send data. Various low-code applications can be installed on the terminal devices, including, but not limited to, enterprise management applications (e.g., ERP systems, CRM systems, HR systems), business process management applications (e.g., logistics management systems, sales management systems, financial management systems), data analysis applications, mobile applications (e.g., shopping applications, web browser applications, search applications, accounting applications, logging applications, instant messaging tools, email clients, social platform software), web applications, IoT applications (e.g., smart home applications), and customized forms (e.g., questionnaires, product demonstrations). The terminal devices can be any electronic device with a display and web browsing support, and the type of terminal device is not specifically limited. For example, the terminal devices include, but are not limited to, electronic devices such as desktop computers, tablet computers, laptop computers, smartphones, and smart TVs. The terminal devices can also be electronic devices consisting of a host computer with multiple virtual machines and human-computer interaction devices (e.g., a touchscreen display, keyboard, and mouse) corresponding to each virtual machine.

[0050] The server 200 can provide various services. Specifically, the server 200 is used to provide background services for the applications installed in the terminal device 100. For example, the server 200 is used to provide support for low-code application development, or to provide a reused application parsing system for low-code applications. The server 200 can analyze and process the received user requests (low-code application development requests, operations in low-code applications, etc.), and return the processing results (extracting functional components and / or business components corresponding to the low-code application development requests, or generating application software corresponding to the low-code application development requests, feedback corresponding to operations in low-code applications, etc.) to the terminal device 100. The server includes but is not limited to one or more physical servers, server clusters, cloud servers based on cloud computing architecture, or distributed servers, etc.

[0051] It should be understood that Figure 1 The application scenarios shown do not constitute any limitation on the application scenarios to which the low-code application development method provided by this disclosure is applicable. For example, Figure 1 The application scenario shown may also include a greater number or more types of terminal devices or servers.

[0052] See also Figure 2 , showing a flow chart of the low-code application development method provided by the present disclosure in one embodiment.

[0053] For example, Figure 2 The low-code application development method described in the above can be applied to, but is not limited to, Figure 1 The application scenario shown. Figure 2 The described method is applied to the above Figure 1 The application scenario shown is Figure 1 The server 200 shown may execute Figure 2 The low-code application development approach shown.

[0054] like Figure 2 As shown, the low-code application development method provided by the embodiment of the present disclosure includes the following steps:

[0055] Step S201, respond to the low-code application development request submitted in natural language, perform semantic analysis on the natural language in the low-code application development request through the metalanguage programming model and perform metalanguage feature extraction processing based on the results of the semantic analysis to obtain the corresponding coding instruction domain standard language code.

[0056] The low-code application development request is in the form of natural language, that is, the low-code application development request is natural language data, which can be obtained through Figure 1 The terminal device 100 is implemented in the embodiment. For example, a human-computer interaction module is provided in the terminal device, and the human-computer interaction module is configured to implement human-computer interaction to obtain a low-code application development request submitted in a natural language.

[0057] In certain embodiments, the human-computer interaction module may be, for example, a text interaction module that obtains natural language data in the form of text information. In certain examples, a human-computer interaction interface is provided on the display device of the terminal device, and a text input box is provided in the human-computer interaction interface. Exemplarily, a user directly enters text information into the text input box using a text input device (e.g., a keyboard, stylus, etc.). Exemplarily, a user uploads an image or video containing text information, and an optical character recognition device (i.e., OCR recognition device) recognizes the image or video to obtain the text information and display it in the text input box.

[0058] In some embodiments, the human-computer interaction module may be, for example, a voice interaction module, which obtains natural language data in the form of voice information. In some examples, the voice interaction module provided on the terminal device may be, for example, a voice assistant, a smart speaker, etc., and illustratively, the user obtains the user's voice information and extracts the voice information through the voice assistant or the smart speaker. In some examples, the voice interaction module on the terminal device includes a voice-to-text device and a human-computer dialogue interface, wherein the voice-to-text device may be, for example, a voice assistant, a smart speaker, etc., and the human-computer dialogue interface may provide a text display box. Exemplarily, the user converts the sound information into text information through the voice-to-text device and displays the text information in the text display box.

[0059] Subsequently, users can also modify or adjust the natural language data in the human-computer interaction module to make its expression more in line with user needs or more in line with the development requirements of low-code applications.

[0060] In this way, users can express their desired low-code application development requirements in natural language through the human-computer interaction module to form a low-code application development request. In some embodiments, the low-code application development request is formed at the terminal device and uploaded to the server. In some embodiments, the low-code application development request is formed at the server after natural language data is uploaded from the terminal device to the server.

[0061] Returning to step S201, in response to the low-code application development request submitted in natural language, the natural language in the low-code application development request is semantically parsed through the metalanguage programming model and metalanguage feature extraction is performed based on the results of the semantic parsing to obtain the corresponding domain specification language code.

[0062] In the disclosed embodiments, a metalanguage programming model is used to perform semantic parsing on the natural language in a low-code application development request, and metalanguage feature extraction is performed based on the results of the semantic parsing to obtain the corresponding domain specification language code. The metalanguage programming model is trained using conversational data, and the domain specification language code is described using a metalanguage that is constructed based on a high-level computer language.

[0063] In the disclosed embodiments, a "super language" is defined between natural languages and high-level computer languages (such as C++, Java, C#, Python, etc.) that hides the complexity of high-level languages and has precise semantics. This allows natural languages to be converted into a precise language that can be understood by code generation tools. We can call this "super language" a "metalanguage."

[0064] In low-code applications, code written in a high-level computer language is encapsulated into building-block components and stored in a code component library. These building-block components include functional components and / or business components that can be assembled. Using a meta-language, functional components and / or business components in the code component library can be directly extracted and assembled to generate the corresponding application software, enabling the development of low-code applications.

[0065] Metalanguage is positioned as a language at a higher level than high-level computer languages for building application software.

[0066] Therefore, in the embodiment of the present disclosure, the method for developing a low-code application also includes the step of creating a metalanguage.

[0067] In certain embodiments, the metalanguage employs a syntax similar to the SQL standard (Structured Query Language), which is easy to understand. Programming languages that employ SQL standard syntax offer significant advantages in terms of portability, learning curve, feature richness, community support, code quality, data standardization, support for modern features, and development efficiency, making them ideal for processing and managing relational data.

[0068] In this way, based on the grammatical structure of natural language and incorporating the logic requirements of computer high-level language programming, the composition method of metalanguage sentences, the combination rules between words, and the expression method of logical relationships are defined, so that natural language can be converted into computer high-level language through metalanguage.

[0069] As an example, see Figure 3 , which is a schematic diagram of the language structure of the metalanguage in an example in an embodiment of the present disclosure.

[0070] like Figure 3 As shown, in one example, the language structure of the metalanguage may include the following objects:

[0071] Application - consists of a series of forms, views, reports, panels, and roles.

[0072] Form - a page for filling in, modifying and viewing data. The form contains sub-objects such as field, toolbar, button, action, etc.

[0073] View view - view, search and analyze data lists. View contains subtable, button, action and other sub-objects.

[0074] Flow process - a process that contains one or more approval nodes, and the flow contains sub-objects such as node

[0075] Datasource - defines the data source for querying and analyzing data across form objects.

[0076] Dashboard panel - contains multiple user interface UI components.

[0077] role - definition of authorized access to an object.

[0078] In addition, the definition statement of the metalanguage is set:

[0079] create - create object statement;

[0080] alter - modify object statement;

[0081] drop - delete object statement;

[0082] show——visual interactive display;

[0083] active - activates or deactivates an object.

[0084] The authorization statement of the metalanguage is also set:

[0085] grant - object authorization statement;

[0086] revoke - revoke object authorization statement.

[0087] In an embodiment of the present disclosure, the low-code application development method also includes the step of writing domain specification language code through metalanguage.

[0088] According to the language structure, definition statements and authorization statements of the metalanguage defined above, domain specification language codes can be written using the metalanguage. The domain specification language codes are associated with functional components and / or business components in the code component library.

[0089] like Figure 3 As shown, for example, code examples using the metalanguage may include:

[0090] Create create application 'Zero Code Application';

[0091] Modify alter application 'Zero-code application' to 'Demo application';

[0092] delete drop application 'Demo Application';

[0093] Activate / deactivate active application 'demo application' to true / false;

[0094] apply application template 'Reimbursement Management Template Application' to 'Demo Application';

[0095] use application 'demo application';

[0096] show application 'Demo Application';

[0097] create form 'Reimbursement Form' {'Reimbursement Person' varchar, 'Reimbursement Date' datetime, 'Amount' decimal};

[0098] grant form 'expense report' to role 'employee';

[0099] CREATE VIEW 'REIMBURSEMENT TABLE LIST' FOR 'REIMBURSEMENT TABLE';

[0100] alter view add toolbar button'new' 'Expense Reimbursement';

[0101] alter view add inner button ANALYSIS.

[0102] By performing semantic parsing on the natural language in the low-code application development request through a metalanguage programming model and performing metalanguage feature extraction processing based on the results of the semantic parsing, the corresponding domain specification language code can be obtained. The domain specification language code is described in a metalanguage.

[0103] In order to improve the dialogue performance of the metalanguage programming model in application development requests raised in natural language, the metalanguage programming model needs to be trained. Therefore, in the embodiment of the present disclosure, the metalanguage programming model can be obtained and strengthened through dialogue data training.

[0104] In certain embodiments, the method for training a metalanguage programming model may include: first training a basic metalanguage programming model using training data to obtain a fine-tuned metalanguage programming model; then testing and evaluating the fine-tuned metalanguage programming model using test data; and determining a metalanguage programming model that meets the evaluation requirements as a finalized metalanguage programming model. The training data and the test data are dialog data generated by populating a dialog template with data extracted from a material library.

[0105] See also Figure 4 , which is a schematic diagram showing the process of training a metalanguage programming model in one embodiment.

[0106] like Figure 4 As shown, the training method of the metalanguage programming model includes the following steps:

[0107] Step S401 : training a basic metalanguage programming model using training data, and fine-tuning the metalanguage programming model according to the training results to obtain a fine-tuned metalanguage programming model.

[0108] In certain embodiments, the selected basic metalanguage programming model may be a general Large Language Model (LLM). For example, the LLM may be the Qwen-72b-Chat model, a Transformer-based LLM. Of course, the Qwen-72b-Chat model is merely an example; in actual applications, other LLMs may be used as the basic metalanguage programming model.

[0109] The training data is dialogue data formed by filling the dialogue template with data extracted from the material library.

[0110] The data in the material library includes structural data extracted from a business database, and / or structural data generated by a large language model, and / or structural data captured through the network, and / or structural data collected through document scanning. That is, the data in the material library can be obtained from one or more of the structural data of a business database, the structural data generated by a large language model, the structural data captured through the network, and the structural data collected through document scanning. In this way, multi-source structural data can be obtained, achieving diversity in structural data, and helping to improve the generalization of the metalanguage programming model.

[0111] In certain embodiments, when forming dialogue data, structural data extracted from the material library may be filled into a plurality of preset dialogue templates, thereby achieving diversified dialogue training data and helping to improve the generalization of the metalanguage programming model.

[0112] In step S401, the basic metalanguage programming model is trained using training data, and the metalanguage programming model is fine-tuned according to the training results to obtain a fine-tuned metalanguage programming model. The following steps may be further included:

[0113] First, at least one large language model is selected as a basic metalanguage programming model and training data is obtained.

[0114] The acquiring of training data includes: extracting corresponding structural data from a material library and filling the extracted structural data into a plurality of preset dialogue templates to form dialogue data as training data.

[0115] Next, the fine-tuning framework is called to input the training data into the basic metalanguage programming model for training and fine-tuning.

[0116] Call the fine-tuning framework to provide the environment required for fine-tuning. For example, the fine-tuning framework may be XTuner, which is a lightweight tool library for fine-tuning large language models. It supports fine-tuning of multiple large language models such as Qwen, Llama2, and InternLM, and supports full fine-tuning and efficient fine-tuning algorithms such as Lora and QLora. The XTuner fine-tuning framework provides flexible configuration, efficient training process, and rich functional support. By fine-tuning large language models, the metalanguage programming model can be made more suitable for the development field of low-code applications.

[0117] During the fine-tuning process, the indicators of the basic metalanguage programming model and the test results of given examples are observed, and the parameters in the basic metalanguage programming model are fine-tuned according to the observed results.

[0118] There are usually two methods for fine-tuning the model: full fine-tuning and efficient fine-tuning. Among them, full fine-tuning involves fine-tuning all parameters of the entire model, which can achieve good results, but due to the huge computing resources required, the cost is relatively high, and it is suitable for scenarios that require the input of a large amount of domain knowledge. In contrast, efficient fine-tuning freezes the original parameters of the basic model and only fine-tunes some new parameters. This can achieve the purpose of fine-tuning while greatly reducing the demand for computing resources and lowering the cost. Therefore, in some embodiments, efficient fine-tuning can be selected. For example, the QLora efficient fine-tuning method can be selected. Compared with the full fine-tuning method, the QLora efficient fine-tuning method can save memory while maintaining speed, which not only greatly reduces the fine-tuning cost, but also achieves significant results. After fine-tuning the basic metalanguage programming model through the QLora efficient fine-tuning method, most of the accuracy of the original metalanguage programming model can be retained, while having a smaller size and faster speed.

[0119] Step S403 : Using the test data, the fine-tuned metalanguage programming model is tested and evaluated, and the fine-tuned metalanguage programming model that meets the evaluation requirements is determined as a finalized metalanguage programming model.

[0120] After fine-tuning the metalanguage programming model according to the training results to obtain a fine-tuned metalanguage programming model in step S401, the fine-tuned metalanguage programming model is tested and evaluated using a test data set. By comparing the performance of the metalanguage programming model on the test data before and after fine-tuning, an indicator evaluation is performed. The indicators include but are not limited to accuracy, precision, recall rate, etc. If the indicator evaluation in the fine-tuned metalanguage programming model does not meet the evaluation requirements, it is necessary to return to step S401 and perform training and fine-tuning again. If the indicator evaluation in the fine-tuned metalanguage programming model meets the evaluation requirements, the metalanguage programming model is determined to be a formed metalanguage programming model.

[0121] The test data is dialogue data formed by filling the dialogue template with data extracted from the material library.

[0122] In practical applications, the data in the material library is obtained from multiple data sources, including structural data from business databases, structural data generated by large language models, structural data captured from the web, and structural data collected through document scanning. Both the training data and the test data are conversational data generated by filling conversation templates with data extracted from the material library.

[0123] In this way, the formed metalanguage programming model can be trained and completed through the above steps S401 and S403, and the formed metalanguage programming model can be deployed and put into practical application.

[0124] Figure 5 Shown as per Figure 4 Schematic diagram of training the metalanguage programming model shown.

[0125] like Figure 5 As shown, a material library is formed by collecting structural data from various data sources, such as structural data from business data databases, structural data generated by large language models, structural data captured through the Internet, structural data collected through document scanning, etc.

[0126] The corresponding structural data is extracted from the material library and filled into various dialogue templates to form dialogue data, which can form training data sets and test data sets.

[0127] The basic metalanguage programming model is trained using the training data in the training data set, and the metalanguage programming model is fine-tuned according to the training results to obtain a fine-tuned metalanguage programming model.

[0128] The fine-tuned metalanguage programming model is tested and evaluated using the test data in the test dataset. If the fine-tuned metalanguage programming model does not meet the evaluation requirements (e.g. Figure 5 N in the figure), then return to training and fine-tuning; if the indicator evaluation in the fine-tuned metalanguage programming model meets the evaluation requirements (such as Figure 5 ), the fine-tuned metalanguage programming model is determined as a final metalanguage programming model, deployed and put into practical application.

[0129] See also Figure 6 , which is a schematic diagram of a process for training the metalanguage programming model in another embodiment.

[0130] like Figure 6 As shown, the training method of the metalanguage programming model includes the following steps:

[0131] Step S601 : training the basic metalanguage programming model using training data, and fine-tuning the metalanguage programming model according to the training results to obtain a fine-tuned metalanguage programming model.

[0132] In certain embodiments, the selected basic metalanguage programming model may be a general Large Language Model (LLM). For example, the LLM may be the Qwen-72b-Chat model, a Transformer-based LLM. Of course, the Qwen-72b-Chat model is merely an example; in actual applications, other LLMs may be used as the basic metalanguage programming model.

[0133] The training data is dialogue data formed by filling the dialogue template with data extracted from the material library.

[0134] The data in the material library includes structural data extracted from a business database, and / or structural data generated by a large language model, and / or structural data captured through the network, and / or structural data collected through document scanning. That is, the data in the material library can be obtained from one or more of the structural data of a business database, the structural data generated by a large language model, the structural data captured through the network, and the structural data collected through document scanning. In this way, multi-source structural data can be obtained, achieving diversity in structural data, and helping to improve the generalization of the metalanguage programming model.

[0135] In certain embodiments, when forming dialogue data, structural data extracted from the material library may be filled into a plurality of preset dialogue templates, thereby achieving diversified dialogue training data and helping to improve the generalization of the metalanguage programming model.

[0136] In step S601, the basic metalanguage programming model is trained using training data, and the metalanguage programming model is fine-tuned according to the training results to obtain a fine-tuned metalanguage programming model. The following steps may also be included:

[0137] First, at least one large language model is selected as a basic metalanguage programming model and training data is obtained.

[0138] The acquiring of training data includes: extracting corresponding structural data from a material library and filling the extracted structural data into a plurality of preset dialogue templates to form dialogue data as training data.

[0139] Next, the fine-tuning framework is called to input the training data into the basic metalanguage programming model for training and fine-tuning.

[0140] Call the fine-tuning framework to provide the environment required for fine-tuning. For example, the fine-tuning framework may be XTuner, which is a lightweight tool library for fine-tuning large language models. It supports fine-tuning of multiple large language models such as Qwen, Llama2, and InternLM, and supports full fine-tuning and efficient fine-tuning algorithms such as Lora and QLora. The XTuner fine-tuning framework provides flexible configuration, efficient training process, and rich functional support. By fine-tuning large language models, the metalanguage programming model can be made more suitable for the development field of low-code applications.

[0141] During the fine-tuning process, the indicators of the basic metalanguage programming model and the test results of given examples are observed, and the parameters in the basic metalanguage programming model are fine-tuned according to the observed results.

[0142] There are usually two methods for fine-tuning the model: full fine-tuning and efficient fine-tuning. Among them, full fine-tuning involves fine-tuning all parameters of the entire model, which can achieve good results, but due to the huge computing resources required, the cost is relatively high, and it is suitable for scenarios that require the input of a large amount of domain knowledge. In contrast, efficient fine-tuning freezes the original parameters of the basic model and only fine-tunes some new parameters. This can achieve the purpose of fine-tuning while greatly reducing the demand for computing resources and lowering the cost. Therefore, in some embodiments, efficient fine-tuning can be selected. For example, the QLora efficient fine-tuning method can be selected. Compared with the full fine-tuning method, the QLora efficient fine-tuning method can save memory while maintaining speed, which not only greatly reduces the fine-tuning cost, but also achieves significant results. After fine-tuning the basic metalanguage programming model through the QLora efficient fine-tuning method, most of the accuracy of the original metalanguage programming model can be retained, while having a smaller size and faster speed.

[0143] Step S603: Fusing the fine-tuned metalanguage programming model and the basic metalanguage programming model to form a fused metalanguage programming model; using test data to test and evaluate the fused metalanguage programming model, and determining the fused metalanguage programming model that meets the evaluation requirements as a finalized metalanguage programming model.

[0144] After fine-tuning the metalanguage programming model according to the training results to obtain a fine-tuned metalanguage programming model in step S601, the basic metalanguage programming model and the fine-tuned metalanguage programming model are fused to obtain a fused metalanguage programming model; thereafter, the fused metalanguage programming model is tested and evaluated using test data, and an indicator evaluation is performed by comparing the performance of the basic metalanguage programming model and the fused metalanguage programming model on the test data. The indicators include but are not limited to accuracy, precision, recall rate, etc. If the indicator evaluation in the fused metalanguage programming model does not meet the evaluation requirements, it is necessary to return to step S601 for further training and fine-tuning. If the indicator evaluation in the fused metalanguage programming model meets the evaluation requirements, the fused metalanguage programming model is determined to be a formed metalanguage programming model.

[0145] The test data is dialogue data formed by filling the dialogue template with data extracted from the material library.

[0146] In practical applications, the data in the material library is obtained from multiple data sources, including structural data from business databases, structural data generated by large language models, structural data captured from the web, and structural data collected through document scanning. Both the training data and the test data are conversational data generated by filling conversation templates with data extracted from the material library.

[0147] In this way, the formed metalanguage programming model can be trained and completed through the above steps S601 and S603, and the formed metalanguage programming model can be deployed and put into practical application.

[0148] Figure 7 Shown as per Figure 6 Schematic diagram of training the metalanguage programming model shown.

[0149] like Figure 7 As shown, a material library is formed by collecting structural data from various data sources, such as structural data from business data databases, structural data generated by large language models, structural data captured through the Internet, structural data collected through document scanning, etc.

[0150] The corresponding structural data is extracted from the material library and filled into various dialogue templates to form dialogue data, which can form training data sets and test data sets.

[0151] The basic metalanguage programming model is trained using the training data in the training data set, and the metalanguage programming model is fine-tuned according to the training results to obtain a fine-tuned metalanguage programming model.

[0152] The basic metalanguage programming model is integrated with the fine-tuned metalanguage programming model to form a fused metalanguage programming model.

[0153] The fused meta-language programming model is tested and evaluated using the test data in the test dataset. If the fused meta-language programming model does not meet the evaluation requirements (e.g. Figure 7 N in the figure), then return to training and fine-tuning; if the indicator evaluation in the fused metalanguage programming model meets the evaluation requirements (such as Figure 7 ), the fused metalanguage programming model is determined as a formed metalanguage programming model, deployed and put into practical application.

[0154] Back to Figure 2 In step S203, based on the domain specification language code, functional components and / or business components that match the domain specification language code are extracted from the code component library, or functional components and / or business components that match the domain specification language code are extracted from the code component library and corresponding application software is generated by assembling them; the functional components and / or business components are written in a high-level computer language.

[0155] In step S201, the natural language in the low-code application development request (conversation data described by the user in natural language) is semantically parsed through the metalanguage programming model and metalanguage feature extraction is performed based on the results of the semantic parsing to obtain the corresponding domain specification language code. The domain specification language code is described in the metalanguage, and the metalanguage is constructed based on a computer high-level language. Therefore, the domain specification language code can be converted into a computer high-level language.

[0156] In some embodiments, based on the domain specification language code, functional components and / or business components matching the domain specification language code are extracted from a code component library.

[0157] In some embodiments, based on the domain specification language code, functional components and / or business components matching the domain specification language code are extracted from a code component library and the corresponding application software is directly generated by assembling them.

[0158] Step S205 : Display the extracted functional components and / or business components, or the generated application software in a visual manner.

[0159] In certain embodiments, based on the domain specification language code, functional components and / or business components that match the domain specification language code are extracted from a code component library, and the extracted functional components and / or business components are displayed in a visual manner. Subsequently, the user can configure the displayed functional components and / or business components to generate the desired application software.

[0160] In certain embodiments, based on the domain specification language code, functional components and / or business components that match the domain specification language code are extracted from a code component library and assembled to directly generate corresponding application software. The generated application software is then displayed in a visual manner. Subsequently, the user can continue to fine-tune the generated application software to obtain the desired application software.

[0161] The low-code application development method uses natural language data to train a metalanguage programming model, and uses a domain specification language code described by the metalanguage in the metalanguage programming model. The metalanguage is constructed based on a computer high-level language. In this way, the metalanguage programming model responds to a low-code application development request submitted in natural language, performs semantic parsing on the natural language in the low-code application development request, and performs metalanguage feature extraction processing based on the result of the semantic parsing to obtain the domain specification language code described in the metalanguage, and then directly extracts the functional components and / or business components that match the domain specification language code from the code component library, or extracts the functional components and / or business components that match the domain specification language code from the code component library and generates the corresponding application software by assembling them, and then displays the extracted content in a visual manner, thereby realizing the function of quickly and accurately generating low-code applications through natural language description requirements. The operation is simple, the application is convenient, and the development efficiency is improved. In particular, the skill requirements for users' programming technology are reduced, and the applicability is wider.

[0162] The disclosed embodiment further provides a low-code application development platform.

[0163] See also Figure 8 , showing a structural diagram of a low-code application development platform provided by the present invention in one embodiment.

[0164] like Figure 8 As shown, the development platform of the low-code application may include: a component encapsulation module 81, a human-computer interaction module 83, a metalanguage programming model 85, a low-code engine 87, and a visualization environment module 89.

[0165] The component encapsulation module 81 is configured to encapsulate codes written in a high-level computer language into assembleable functional components and / or business components, and store the functional components and / or business components in a code component library.

[0166] The human-computer interaction module 83 is configured to implement human-computer interaction and is used to obtain low-code application development requests submitted using natural language.

[0167] In certain embodiments, the human-computer interaction module may be, for example, a text interaction module that obtains natural language data in the form of text information. In certain examples, a human-computer interaction interface is provided on the display device of the terminal device, and a text input box is provided in the human-computer interaction interface. Exemplarily, a user directly enters text information into the text input box using a text input device (e.g., a keyboard, stylus, etc.). Exemplarily, a user uploads an image or video containing text information, and an optical character recognition device (i.e., OCR recognition device) recognizes the image or video to obtain the text information and display it in the text input box.

[0168] In some embodiments, the human-computer interaction module may be, for example, a voice interaction module, which obtains natural language data in the form of voice information. In some examples, the voice interaction module provided on the terminal device may be, for example, a voice assistant, a smart speaker, etc., and illustratively, the user obtains the user's voice information and extracts the voice information through the voice assistant or the smart speaker. In some examples, the voice interaction module on the terminal device includes a voice-to-text device and a human-computer dialogue interface, wherein the voice-to-text device may be, for example, a voice assistant, a smart speaker, etc., and the human-computer dialogue interface may provide a text display box. Exemplarily, the user converts the sound information into text information through the voice-to-text device and displays the text information in the text display box.

[0169] The metalanguage programming model 85 is configured to perform semantic analysis on the natural language in the low-code application development request and perform metalanguage feature extraction processing based on the results of the semantic analysis to obtain the corresponding domain specification language code.

[0170] In the embodiment of the present disclosure, the metalanguage programming model is obtained by training with conversation data described in natural language, the domain specification language code is described in the metalanguage, and the metalanguage is constructed based on a high-level computer language.

[0171] For details about the metalanguage programming model and its training method, please refer to the previous description and will not be repeated here.

[0172] In some embodiments, the low-code application development platform further includes a domain specification language code writing module 86 , which is configured to write domain specification language code using a metalanguage.

[0173] The low-code engine 87 is configured to: based on the domain specification language code, extract functional components and / or business components that match the domain specification language code from the code component library, or extract functional components and / or business components that match the domain specification language code from the code component library and generate corresponding application software by assembling them.

[0174] In the disclosed embodiment, a metalanguage interaction interface is provided between the metalanguage programming model 85 and the low-code engine 87 in the low-code application development platform. The metalanguage interaction interface can be denoted as MetaOpenAPI. The metalanguage interaction interface can adopt a standardized interaction protocol based on TCP / IP, which provides the possibility of realizing a natural language human-computer interaction mode.

[0175] In actual applications, the metalanguage programming model transmits the obtained domain specification language code to the low-code engine 87 through the metalanguage interaction interface. The low-code engine 87 extracts functional components and / or business components that match the domain specification language code from the code component library based on the domain specification language code, or extracts functional components and / or business components that match the domain specification language code from the code component library and generates corresponding application software by assembling them.

[0176] The visualization environment module 89 is configured to provide a visualization interface, through which the functional components and / or business components extracted by the low-code engine, or the generated application software, are displayed in a visual manner.

[0177] In some embodiments, the extracted functional components and / or business components are displayed in a visual manner through the visualization environment module 89. Subsequently, the user can configure the displayed functional components and / or business components to generate the required application software.

[0178] In some embodiments, the generated application software is displayed in a visual manner through the visualization environment module 89. Subsequently, the user can continue to fine-tune the generated application software to obtain the desired application software.

[0179] The present disclosure further provides an electronic device. Figure 9 , which shows a functional block diagram of an electronic device provided by the present disclosure in one embodiment.

[0180] like Figure 9 As shown, the electronic device 9 includes a processor 91 and a memory 93. The processor 91 and the memory 93 can communicate via a bus 92. The memory 93 can store a computer program. The processor 91 executes the computer program in the memory 93 to implement each step of the low-code application development method or each step of the metalanguage programming model training method.

[0181] Bus 92 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. Buses can be categorized as address buses, data buses, and control buses. For ease of illustration, although only one thick line is used in the figure, this does not imply that there is only one bus or only one type of bus.

[0182] In some embodiments, the processor 91 may be implemented as a central processing unit (CPU), a microcontroller unit (MCU), a system on chip (SoC), or a field programmable gate array (FPGA). The memory 93 may include volatile memory, such as random access memory (RAM), for temporarily storing data while running programs. The memory 93 may also include non-volatile memory (NVM) for data storage, such as read-only memory (ROM), flash memory, a hard disk drive (HDD), or a solid-state drive (SSD).

[0183] In an embodiment of the present disclosure, a computer-readable storage medium may be provided, storing a computer program or instruction, which implements the method flow or function of any of the previous embodiments when executed.

[0184] That is, the method steps in the above-described embodiments are implemented as software or computer code that can be stored in a recording medium (such as a CD ROM, RAM, floppy disk, hard disk or magneto-optical disk), or are implemented as computer code originally stored in a remote recording medium or a non-transitory machine-readable medium downloaded via a network and to be stored in a local recording medium, so that the methods represented herein can be processed by such software stored on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA).

[0185] The present disclosure may also provide a computer program product, which includes one or more computer programs or instructions that, when executed, fully or partially execute the processes or functions in the present disclosure. The computer program product includes one or more computer programs or instructions.

[0186] A computer program or instruction can be stored in a readable storage medium or transferred from one readable storage medium to another. For example, the computer program or instruction can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The readable storage medium can be any accessible medium or a data storage device such as a server or data center that integrates one or more accessible media. The accessible medium can be a magnetic medium such as a floppy disk, hard disk, or magnetic tape; an optical medium such as a digital video disk; or a semiconductor medium such as a solid-state drive. The computer-readable storage medium can be a volatile or non-volatile storage medium, or can include both volatile and non-volatile types of storage media.

[0187] The above embodiments are merely illustrative of the principles and effects of this disclosure and are not intended to limit this disclosure. Anyone skilled in the art may modify or alter the above embodiments without departing from the spirit and scope of this disclosure. Therefore, any equivalent modifications or alterations made by a person skilled in the art without departing from the spirit and technical concepts disclosed herein shall be encompassed by the scope of protection of this disclosure.

Claims

1. A method for developing low-code applications, characterized in that: The steps include: In response to a low-code application development request submitted in natural language, semantically parse the natural language in the low-code application development request using a metalanguage programming model and perform metalanguage feature extraction processing based on the results of the semantic parsing to obtain a corresponding domain specification language code; wherein the metalanguage programming model is obtained through training with conversation data, and the domain specification language code is described in a metalanguage, which is constructed based on a high-level computer language; Based on the domain specification language code, functional components and / or business components that match the domain specification language code are extracted from a code component library, or functional components and / or business components that match the domain specification language code are extracted from a code component library and assembled to generate corresponding application software; the functional components and / or business components are written in a high-level computer language; and The extracted functional components and / or business components, or the generated application software are displayed in a visual manner.

2. The low-code application development method according to claim 1, characterized in that: It also includes the step of creating a metalanguage, which adopts the syntax of structured query language, is based on the grammatical structure of natural language and incorporates the logic requirements of computer high-level language programming, defines the sentence composition method, the combination rules between words and the expression method of logical relationships, so that natural language can be converted into computer high-level language through metalanguage.

3. The low-code application development method according to claim 1, characterized in that: The method further includes the step of writing a domain specification language code by using a metalanguage, wherein the domain specification language code is associated with a functional component and / or a business component in a code component library.

4. The low-code application development method according to claim 1, characterized in that: The training of the metalanguage programming model includes the following steps: Using the training data to train the basic metalanguage programming model, and fine-tuning the metalanguage programming model according to the training results to obtain a fine-tuned metalanguage programming model; as well as Using the test data to test and evaluate the fine-tuned metalanguage programming model, and determining the fine-tuned metalanguage programming model that meets the evaluation requirements as a finalized metalanguage programming model; The training data and the test data are conversation data formed by filling data extracted from a material library into a conversation template. The data in the material library includes structural data extracted from a business database, and / or structural data generated by a large language model, and / or structural data captured from the network, and / or structural data collected by document scanning.

5. The low-code application development method according to claim 1, characterized in that: The training of the metalanguage programming model includes the following steps: Using the training data to train the basic metalanguage programming model, and fine-tuning the metalanguage programming model according to the training results to obtain a fine-tuned metalanguage programming model; as well as The fine-tuned metalanguage programming model and the basic metalanguage programming model are integrated to form an integrated metalanguage programming model; the integrated metalanguage programming model is tested and evaluated using test data, and the integrated metalanguage programming model that meets the evaluation requirements is determined as a finalized metalanguage programming model; The training data and the test data are conversation data formed by filling data extracted from a material library into a conversation template. The data in the material library includes structural data extracted from a business database, and / or structural data generated by a large language model, and / or structural data captured from the network, and / or structural data collected by document scanning.

6. The method for developing a low-code application according to claim 4 or 5, characterized in that: The basic metalanguage programming model is trained using the training data, and the metalanguage programming model is fine-tuned according to the training results to obtain a fine-tuned metalanguage programming model, including the following steps: Select at least one language model as the basic metalanguage programming model and obtain training data; as well as Calling a fine-tuning framework to input the training data into the basic metalanguage programming model for training and fine-tuning; During the fine-tuning process, the indicators of the basic metalanguage programming model and the test results of given examples are observed, and the parameters in the basic metalanguage programming model are fine-tuned according to the observed results.

7. A low-code application development platform, characterized by: include: A component packaging module is configured to package codes written in a high-level computer language into assembleable functional components and / or business components, and store the functional components and / or business components in a code component library; A human-computer interaction module is configured to implement human-computer interaction and is used to obtain low-code application development requests submitted in natural language through a conversation; A metalanguage programming model is configured to perform semantic parsing on the natural language in the low-code application development request and perform metalanguage feature extraction processing based on the results of the semantic parsing to obtain corresponding domain specification language code; wherein the metalanguage programming model is trained using conversational data, the domain specification language code is described using a metalanguage, and the metalanguage is constructed based on a high-level computer language; The low-code engine is configured to: extract functional components and / or business components that match the domain specification language code from a code component library based on the domain specification language code, or extract functional components and / or business components that match the domain specification language code from the code component library and generate corresponding application software by assembling them; wherein the functional components and / or business components are written in a high-level computer language; and The visualization environment module is configured to provide a visualization interface, through which the functional components and / or business components extracted by the low-code engine, or the generated application software, are displayed in a visual manner.

8. The low-code application development platform according to claim 7, characterized in that: Also includes: The domain specification language code writing module is configured to write domain specification language code using a metalanguage.

9. An electronic device, characterized in that: The electronic device includes a memory, a processor, and a computer program stored on the memory, and the processor executes the computer program to implement the steps of the low-code application development method described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the low-code application development method described in any one of claims 1 to 6.

Images