CN117193738A

CN117193738A - Application building method, device, equipment and storage medium

Info

Publication number: CN117193738A
Application number: CN202311031245.6A
Authority: CN
Inventors: 查洪钰; 崔陶; 凌军; 段鑫; 罗威; 石亚飞; 裴东辉; 程然; 王士进
Original assignee: iFlytek Co Ltd
Current assignee: iFlytek Co Ltd
Priority date: 2023-08-14
Filing date: 2023-08-14
Publication date: 2023-12-08

Abstract

The application discloses an application building method, an application building device, application building equipment and a storage medium. The application can also pre-establish the mapping relation between the standard protocol and the target protocol which can be identified by each low code platform, convert the generated executable code of the standard protocol into the executable code of the target protocol by referring to the mapping relation, and render the executable code of the target protocol through the low code platform, thus obtaining the built target application. By adopting the scheme of the application, a user only needs to describe the application development requirement by adopting natural language, and does not need to write any code, thereby improving the code development efficiency. And development logics of different low-code platforms are not required to be learned by developers, so that the capability requirement of the application development process on the developers is reduced.

Description

Application building method, device, equipment and storage medium

Technical Field

The present application relates to the field of low-code application building technologies, and in particular, to an application building method, apparatus, device, and storage medium.

Background

In the present digital age, the development and deployment of applications has become an important component of enterprise digital transformation. However, these processes tend to be bottlenecks in enterprise digital transformation due to the significant amount of manpower and time required for application development and deployment.

To solve this problem, low-code development platforms have appeared in recent years, which can help developers to quickly build applications, thereby improving development efficiency. However, these platforms still require manual code writing and thus still require a significant amount of manpower and time. In addition, the concepts of platform construction, concepts on the platform, operation modes, application construction paths and the like of various manufacturers are different, so that the learning threshold and the threshold of platform migration are high, the informatization literacy requirement on users is high, and the areas with relatively underdeveloped informatization level cannot be covered.

Disclosure of Invention

In view of the above problems, the present application is provided to provide an application building method, apparatus, device, and storage medium, so as to reduce workload of developers in an application development process, improve application development efficiency, and reduce requirements on the developers.

The specific scheme is as follows:

in a first aspect, an application building method is provided, including:

acquiring application development requirements input by a user;

invoking a preconfigured artificial intelligence model to parse the application development requirements to generate executable codes of a standard protocol;

according to the mapping relation between the standard protocol and the target protocol which can be identified by the low code platform, converting the executable code of the standard protocol into the executable code of the target protocol;

and rendering the executable code of the target protocol through the low-code platform to obtain the built target application.

Preferably, the method further comprises:

and after the target application is obtained, displaying a function overview and detail page of the target application so as to enable a user to browse the function page and detail of the application.

Preferably, the method further comprises:

acquiring an application function query instruction input by a user, wherein the query instruction is used for querying whether the target application can realize a target function;

analyzing the application function query instruction, judging whether the target application has the target function, and outputting a judging result.

Preferably, the method further comprises:

acquiring an application editing requirement input by a user, wherein the application editing requirement is used for indicating the editing of the target application;

Identifying an editing intention of a user based on the application editing requirement, wherein the editing intention comprises an operation type and an operation object, and the operation type comprises addition, deletion and/or modification;

and editing the target application based on the editing intention of the user to obtain the edited target application.

Preferably, the method further comprises:

and displaying an application editing page, wherein the application editing page contains application information related to the current editing in the target application.

Preferably, the operation object includes a function page and contents or components in the function page, wherein the function page includes any one of a form page, a statistics page and a flow page.

Preferably, before said invoking the preconfigured artificial intelligence model parses the application development requirements to generate executable code of a standard protocol, further comprising:

processing the application development requirements into a structured text;

the process of invoking the preconfigured artificial intelligence model to parse the application development requirements to generate executable code of a standard protocol includes:

and calling the artificial intelligence model to convert the structured text into executable codes of standard protocols.

Preferably, the process of invoking the artificial intelligence model and converting the structured text into executable code of a standard protocol comprises:

invoking the artificial intelligent model to generate application basic description information corresponding to the application development requirement;

invoking the artificial intelligent model, and generating an application function description list according to the application basic description information;

invoking the artificial intelligent model, and generating a data model list according to the application function description list;

and calling the artificial intelligent model, and generating roles and authority lists related to the application according to the data model list and the application function description list.

Preferably, if the editing intention of the user is more than one, identifying the editing intention of the user based on the application editing requirement comprises:

determining a target dialogue scene when a user inputs the application editing requirement, and acquiring a preconfigured intention list corresponding to the target dialogue scene, wherein the intention list comprises all editing intents which can be processed in the target dialogue scene;

the artificial intelligence model is invoked to identify respective editing intents of the user in the application editing requirements with reference to the list of intents.

Preferably, when there is a hierarchical order among the editing intents in the intention list, the process of invoking the artificial intelligence model to identify the editing intents of the user in the application editing requirement by referring to the intention list includes:

acquiring a first prompt instruction prompt template, wherein the template comprises an intention bill slot and an application editing requirement slot and is used for indicating a model to analyze editing intention of a first layer contained in application editing requirements in the application editing requirement slot according to the layer sequence of all editing intention in the intention bill slot;

assembling the intention list into the intention list slot, assembling the application editing requirement into the application editing requirement slot to obtain a first prompt instruction prompt, and sending the first prompt instruction prompt into the artificial intelligent model to obtain a target editing intention output by the model;

judging whether the target editing intention is the last leaf editing intention in the intention list;

if not, selecting the editing intention of each layer after the target editing intention from the intention list according to the layer sequence as a new intention list, and returning to execute the steps of assembling the intention list into the intention list groove, assembling the application editing requirement into the application editing requirement groove, and obtaining a first prompt instruction prompt;

And if so, calling the artificial intelligent model to acquire the slot information required by the target editing intention, and forming each editing intention of an end user by the obtained target editing intention of each layer which is not the leaf editing intention, the target editing intention belonging to the leaf editing intention and the slot information thereof.

Preferably, the process of editing the target application based on the editing intention of the user to obtain the edited target application includes:

invoking the artificial intelligence model, and generating executable codes of the standard protocol based on the editing intention of the user as updated executable codes;

converting the updated executable code of the standard protocol into updated executable code of the target protocol;

and rendering the updated executable code of the target protocol through the low-code platform to obtain the edited target application.

Preferably, the application development requirement is multi-modal information, and the multi-modal information includes: application development requirements of any one of a text modality, an audio modality, and a visual modality.

Preferably, the built target application is published in an application market, and the method further comprises:

Acquiring application use requirements set forth by a user in the application market;

determining an application matched with the application use requirement in the application market;

extracting and filling business data on which form pages of the matched application depend by combining the application use requirement and the matched application information;

and starting the matched application, and filling the form page of the matched application by using the service data.

Preferably, the process of determining the application matched with the application use requirement in the application market includes:

determining a first vector representation corresponding to the application use requirement;

vector matching is carried out in a vector library based on the first vector representation, a matched vector representation is obtained, and an application corresponding to the matched vector representation is obtained and used as a matched application; the vector library comprises vector representations corresponding to each application constructed based on application information of each application in the application market.

In a second aspect, there is provided an application building apparatus comprising:

the development requirement acquisition unit is used for acquiring application development requirements input by a user;

the standard executable code generating unit is used for calling a preconfigured artificial intelligent model to analyze the application development requirements so as to generate executable codes of a standard protocol;

The protocol conversion unit is used for converting the executable code of the standard protocol into the executable code of the target protocol according to the mapping relation between the standard protocol and the target protocol which can be identified by the low code platform;

and the executable code rendering unit is used for rendering the executable code of the target protocol through the low-code platform to obtain the built target application.

In a third aspect, there is provided an application building apparatus comprising: a memory and a processor;

the memory is used for storing programs;

the processor is configured to execute the program to implement the steps of the application building method described above.

In a fourth aspect, there is provided a storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the application build method as described above.

According to the technical scheme, a user can directly input application development requirements of natural language description when the application needs to be built. Furthermore, the application can also pre-establish the mapping relation between the standard protocol and the target protocol which can be identified by each low code platform, thereby converting the generated executable code of the standard protocol into the executable code of the target protocol by referring to the mapping relation, and rendering the executable code of the target protocol through the low code platform, thus obtaining the built target application. Obviously, by adopting the scheme of the application, a user only needs to describe the application development requirement by adopting natural language, and does not need to write any code, thereby greatly reducing the workload of developers and improving the code development efficiency.

Meanwhile, the scheme of the application does not need a developer to learn the development logic of different low-code platforms, even the developer can be a user who does not understand codes, and the capability requirement of the application development process on the developer is reduced.

Drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the application. Also, like reference numerals are used to designate like parts throughout the figures. In the drawings:

FIG. 1 is a schematic flow chart of an application building method according to an embodiment of the present application;

FIG. 2 illustrates a schematic diagram of an application build process;

FIG. 3 illustrates an application build process interactive interface schematic;

FIG. 4 illustrates another application build process interactive interface schematic;

FIG. 5 illustrates yet another application build process interactive interface schematic;

FIG. 6 illustrates yet another application build process interactive interface schematic;

FIG. 7 illustrates yet another application build process interactive interface schematic;

FIG. 8 illustrates yet another application build process interactive interface schematic;

FIG. 9 illustrates yet another application build process interactive interface schematic;

FIG. 10 illustrates an application use process interactive interface schematic;

FIG. 11 illustrates another application use process interactive interface schematic;

FIG. 12 illustrates a process diagram identifying a plurality of editing intent contained in a user's application editing requirements;

fig. 13 is a schematic structural diagram of an application building device according to an embodiment of the present application;

fig. 14 is a schematic structural diagram of an application building device according to an embodiment of the present application.

Detailed Description

The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.

Before describing the inventive solution, the English language referred to in this text is explained first:

prompt: an instruction is indicated. When interacting with an AI (such as an artificial intelligence model), the instruction to be sent to the AI can be a text description, such as "please help me recommend a popular music" input when you interact with the AI, or a parameter description according to a certain format, such as making the AI draw according to a certain format, and describing related drawing parameters.

Artificial intelligence model: the model is an artificial intelligent model based on deep learning technology, which consists of hundreds of millions of parameters, and can realize complex tasks such as natural voice processing, image recognition, voice recognition and the like through learning and training of a large amount of data. The artificial intelligence model may include a large language model, a large scale pre-training model.

Large language model: (Large language model, LLM) generally refers to a language model with a large number of parameters and capabilities that learns statistical rules and semantic relationships of a language by pre-training on large-scale text data. These models typically use an unsupervised learning method to predict the next word or fill in missing words to capture the context and semantic information of the language. The large language model is capable of generating coherent sentences, answering questions, completing translation tasks, and the like. The advantage of large language models is that they can understand and generate natural language, thereby enabling a more intelligent way of interaction. Common large language models such as GT4 and other enterprise developed large language models.

A Large-scale Pre-Trained-Models (Large Pre-Trained-Models) refers to Models that are Pre-Trained on Large-scale data sets, which typically include Large amounts of multi-modal data, such as text, images, video, etc., on the internet. The pre-training model may be a language model, an image model, a speech model, etc., that obtains a generic knowledge representation by learning patterns and features in the data. The goal of these models is to learn a generic representation capability for fine tuning or migration learning in various downstream tasks.

A large language model is a specific type of large-scale pre-training model that focuses on processing language tasks such as natural language generation, machine translation, text summarization, and the like. Large-scale pre-training models more broadly refer to various models that are pre-trained on large-scale data, including multiple model types of language models.

The application provides an automatic application building scheme based on an artificial intelligence model, which can automatically build an application program according to natural language description of a user, improves development efficiency and reduces capability requirements on developers.

The application construction scheme of the application can be suitable for automatically constructing applications in various fields, such as the automatic construction of related applications in education fields, office fields and the like.

The scheme of the application can be realized based on the terminal with the data processing capability, and the terminal can be a mobile phone, a computer, a server, a cloud terminal and the like.

Next, as described in connection with fig. 1, the application building method of the present application may include the steps of:

step S100, acquiring application development requirements input by a user.

Specifically, when there is a need to develop an application, the user can describe the application development need through natural language. The application development requirement can be a basic function of an application to be developed described by a user, for example, "i want to build an equipment report repair application, mainly comprising functions of report repair application, report repair approval, report repair evaluation and the like", "please help me build an offline management application, parents can help students apply for leaving a school, and application information needs to pass through a work approval and approval", and the like.

In practical situations, the user may input application development requirements of multiple modes, such as application development requirements of any one or more of a text mode, an audio mode, and a visual mode.

In order to facilitate the invocation of the artificial intelligence model in the next step, the application development requirements may be further processed into structured text in this embodiment. For example, for application development requirements of text modalities, text normalization may be performed to obtain structured text.

For the application development requirement of the audio mode, the audio mode can be converted into a text mode through a voice recognition function and further regulated into a structured text.

The application development requirements of the visual mode can be converted into a text mode through an image recognition function and further regulated into a structured text.

Referring to FIG. 2, processing of multimodal input, all converted to structured text, is illustrated.

And step S110, calling a preconfigured artificial intelligent model to analyze the application development requirements so as to generate executable codes of a standard protocol.

It can be appreciated that the artificial intelligence model invoked in this embodiment may be a large language model or a model of other structures, which itself has super natural language semantic understanding capability and code generation capability, for example, multiple types of artificial intelligence models such as GPT3.5 and GPT4 that are currently or future developed. In this step, the artificial intelligence model may be invoked to analyze the application development requirements, understand the development requirements of the user, and generalize the content of the user development requirements to generate a more detailed requirement description, so as to generate executable codes of the standard protocol based on the more detailed requirement description.

The standard protocol can be a set existing protocol or an intermediate protocol customized by the application, and the artificial intelligent model can be ensured to identify and generate executable codes conforming to the standard protocol, and can be converted into the protocol identifiable by the low-code platform through protocol mapping in the subsequent steps.

The executable code generated in this step may be domain specific language DSL. In addition, some script code, such as SQL, may be included to support data processing for certain functional modules in the developed application, such as statistics SQL relied upon in statistics pages, and the like.

With further reference to fig. 2, for the generated executable code of the standard protocol, the correctness and the integrity of each segment of code can be further checked, if a certain segment of code is found to have an error or not complete enough, the artificial intelligence model can be called again for the segment of code, and the executable code can be regenerated, so that the correctness and the integrity of the generated executable code of the standard protocol are improved.

And step 120, converting the executable code of the standard protocol into the executable code of the target protocol according to the mapping relation between the standard protocol and the target protocol which can be identified by the low code platform.

Specifically, in this embodiment, considering that the building concepts, the concepts on the platform, the operation modes, the application building paths, and the like of different low-code platforms are different, the learning threshold and the threshold of platform migration are higher. Therefore, the application can respectively establish the mapping relation between the standard protocol and the target protocol which can be identified by the low code platform aiming at different low code platforms. On the basis, the mapping relation can be consulted, and the executable code of the standard protocol can be converted into the executable code of the target protocol which can be identified by any low-code platform, so that the user does not need to learn the knowledge of different low-code platforms, and the requirement on the user capacity is greatly reduced. It will be appreciated that the converted executable code conforms to a target protocol that can be identified by the low code platform.

And step S130, rendering the executable code of the target protocol through the low-code platform to obtain the built target application.

In particular, executable code may be rendered by various types of engines on a low-code platform to generate a target application that is adapted to multiple terminals. As shown in fig. 2, when the low-code platform renders executable code, including but not limited to: corresponding code data in the executable code is rendered through a data modeling engine, a metadata engine, a form engine, a workflow engine and a chart engine.

The rendered target application can be adapted to various types of terminals, such as WEB type, H5 type and the like.

According to the application building method provided by the embodiment of the application, a user can directly input the application development requirement of natural language description when the application needs to be built. Furthermore, the application can also pre-establish the mapping relation between the standard protocol and the target protocol which can be identified by each low code platform, thereby converting the generated executable code of the standard protocol into the executable code of the target protocol by referring to the mapping relation, and rendering the executable code of the target protocol through the low code platform, thus obtaining the built target application. Obviously, by adopting the scheme of the application, a user only needs to describe the application development requirement by adopting natural language, and does not need to write any code, thereby greatly reducing the workload of developers and improving the code development efficiency.

In some embodiments of the present application, the above-mentioned step S110, a process of calling a preconfigured artificial intelligence model to parse the application development requirements to generate executable code of a standard protocol is described.

An alternative implementation scheme is provided in this embodiment, which may simulate a process of writing executable code by a developer, and may specifically include the following steps:

s1, calling the artificial intelligent model to generate application basic description information corresponding to the application development requirement.

It can be understood that the application development requirement input by the user can be a short description, and in this embodiment, in order to be able to build a target application meeting the user requirement, an artificial intelligent model can be called to generalize the content of the application development requirement of the user, so as to obtain more abundant application description information, namely application basic description information.

Specifically, the model can be invoked by a set prompt instruction, namely, a prompt instruction, namely, "help me expand the application development requirement input by the user, and generate more detailed application basic information". < application development requirements > ", wherein < application development requirements > are used in slots to fill specific application development requirements.

To facilitate understanding, application development requirements entered by a user: "help me build a student leave management application" is illustrated as an example. After the artificial intelligence model is called, the output application basic information intercepting part is as follows:

s2, calling the artificial intelligent model, and generating an application function description list PRD according to the application basic description information.

Specifically, the application basic description information may be utilized to generate a prompt instruction prompt, and further invoke an artificial intelligence model to generate an application function description manifest. The promt indicates, for example, "refer to the following application basic description information, and generate an application function description list. < application basic description information > ".

Still using the above example, the generated application function description list PRD cut-out is exemplified as follows:

s3, calling the artificial intelligent model, and generating a data model list according to the application function description list.

Specifically, the application function description manifest may be utilized to generate a hint instruction prompt, which in turn invokes the artificial intelligence model to generate a data model manifest. The prompt is, for example, "refer to the following list of functional descriptions, to generate a list of data models. < function description list > ".

Still referring to the above example, the generated data model manifest interception portion is exemplified as follows:

s4, calling the artificial intelligent model, and generating roles and authority lists related to the application according to the data model list and the application function description list.

Specifically, the data model manifest and the application function description manifest may be utilized to generate a prompt instruction prompt, and further invoke an artificial intelligence model to generate roles and permission manifests related to an application. The promtt is, for example, "refer to the following data model list and application function description list, and generate roles and authority list related to the application. < data model list >, < function description list > ".

Still using the above example, the roles and permission list intercepting part generated are exemplified as follows:

in this embodiment, the generated application basic description information, application function description list, data model list, role and authority list may be executable codes of standard protocols. In this embodiment, the process of simulating the development of the executable code of the application by the developer gradually generates each type of code data by calling the code generation capability of the artificial intelligence model, and finally combines the code data into the executable code of the standard protocol.

In some embodiments of the present application, a man-machine interaction method using a building process is provided, and is described below with reference to fig. 3-9.

The application building scheme of the application can be displayed to the user through an interface in the form of an application development assistant, and as shown in fig. 3, an 'AI assistant creation' entry can be provided on the home page, and the user can enter a newly built application interface through the entry. At the same time, the home page may also be presented with applications that have been created, such as the "My applications" column illustrated in FIG. 3.

As shown in fig. 4, in the newly built application interface, an interactive dialog box with the application development assistant may be provided, and the user may input the application development requirement through a multi-modal form, and in fig. 4, only the case of inputting the application development requirement through text and voice forms is illustrated.

Further, in the newly created application interface, a plurality of input samples may be displayed for reference by the user.

After inputting a specific application development requirement in the dialog box, the user clicks the "send" icon, that is, enters an application building interface, as shown in fig. 5. At this time, the system is performing an application building process, which may be divided into a plurality of sub-steps, and the building process is sequentially performed according to the sequence of the sub-steps, and current building detail information of each sub-step is displayed. Fig. 5 illustrates a sub-step division manner of the application building process, which mainly includes four steps, namely: data model generation, application page generation, business process generation and role weight configuration. In addition, the progress bar can be displayed on the application building interface according to the current building progress. The application development requirement input by the user illustrated in fig. 5 is "please help me build an off-school management application, parents can help students apply off-school, and application information needs to be checked and approved by the executive main office".

After the target application is successfully built (i.e., the progress in fig. 5 reaches 100%), the system jumps to the build success interface, as shown in fig. 6. Two portals, namely 'unused' and 'desregulated', can be arranged on the successful interface of construction, and a user can jump to the application interface of the application through the 'unused' portals. The user can jump to the editing interface of the application through the 'adjustment removal' entry, and the user can describe own editing requirements, such as adding, deleting and modifying functions of the built target application.

Further, after the target application is successfully built, the function overview and detail page of the target application can be displayed, and each function page and detail information of the target application can be displayed on the page so that a user can browse and check whether the target application accords with expectations.

In addition, in order to further facilitate the use of the user, the present embodiment also provides the capability of supporting the user to query the application function. Specifically, the user may input an application function query instruction in the interactive dialog box for querying whether the target application is capable of implementing the target function. For example, the query instruction input by the user may be "whether to consider or not, and send a short message to the parents and the lesson teachers after the approval is finished". In this embodiment, after an application function query instruction of a user is obtained, the instruction is analyzed, whether the target application has the target function is determined, and then a determination result is output. For example, when the target application is judged to send a short message to the parent after the approval is finished, but not to the lesson teacher, the output judgment result may be "the current application will send a short message to the parent after the approval is finished, not to the lesson teacher".

In this embodiment, when the application function query instruction is analyzed, understanding of the user intention may be performed through the artificial intelligence model, and the result matched with the user intention may be output in combination with the application information of the created target application.

Obviously, according to the scheme provided by the embodiment, a user can directly provide a query instruction for the function of the target application in a natural language description mode, and whether the function of the created target application is complete or has deviation can be quickly known without browsing the target application page by page.

Further, when the user confirms that the created target application does not meet the expectations, an application editing requirement may be further input to instruct re-editing operation of the created target application.

In this embodiment, when receiving an application editing requirement input by a user, an editing intention of the user may be identified based on the application editing requirement, where the editing intention includes an operation type and an operation object.

Wherein the operation types include addition, deletion and/or modification. The operation object comprises a function page and contents or components in the function page, wherein the function page comprises any one of a form page, a statistics page and a flow page. That is, the user may add, delete, and change any one of the function pages, or add, delete, and change the content or components in the function page.

Referring to table 1 below, several different objects of operation are illustrated:

TABLE 1

Further, the target application is edited based on the editing intention of the user, and the edited target application is obtained.

Specifically, in this embodiment, an artificial intelligence model may be invoked, and executable code of the standard protocol may be generated based on the editing intention of the user as updated executable code. And further converting the updated executable code of the standard protocol into the updated executable code of the target protocol, and rendering the updated executable code of the target protocol through the low-code platform to obtain the edited target application.

As shown in fig. 7-8, which illustrate the case of adding a node in the flow based on the editing requirements of the user.

The user can input editing requirements again aiming at the created offline management application, such as 'I want to synchronize the information of the school student to the class card', so that the teacher can know the leave situation of the student conveniently. At this time, on the application editing interface, the operation object information corresponding to the current application editing requirement in the target application before editing can be displayed. Illustrated in fig. 7 is a "leave school application approval process".

When the system completes editing the target application and obtains the edited target application, the system can jump to an editing success interface, as shown in fig. 8.

Further, as shown in fig. 8, the displayed application editing page may include application information related to the current editing in the target application, that is, application information changed after editing compared with before editing. As shown in fig. 8, a flow node is added after editing: and sending the leave requesting data, wherein the data receiving object is an electronic ban card system. The application information for the changes may be highlighted, for example by marking with a dashed box or other form.

As shown in fig. 9, it exemplifies a case of adding a statistics page based on editing needs of a user.

The user can input editing requirements again aiming at the created offline management application, such as "I want a school student statistics function, and can display the change trend and the specific list of the school students in the month". When the system completes editing the target application and obtains the edited target application, the system can jump to an editing success interface, as shown in fig. 9. After editing, a statistical page of 'student off-school analysis' is added, wherein a change trend chart of the student off-school in the month and a list of the student off-school in the month are displayed.

The embodiment of the present application is further described with respect to the process of identifying the editing intention of the user based on the application editing requirement described in the above embodiment.

An artificial intelligence model may be invoked in this embodiment to identify the user's editing intent based on the application editing requirements.

Further, considering that the user may describe a plurality of different editing intents at the same time when describing the application editing requirement, an example is as follows, "i want to add a teacher role at educational administration and create a statistical analysis page for it, implement the leave-free times of more than 7 days, 14 days, 30 days, the statistical data of more than 7 days at each age and details", the application editing requirement includes a plurality of different editing intents as follows:

(1) Teacher role in creating educational administration department

(2) Creating a statistical analysis page

(3) Five statistical graphs are added, wherein three statistical graphs are as follows: leave the school for more than 7 days, 14 days, 30 days and leave the statistical chart of the false man, the other two statistical charts are: statistical data graphs and detail graphs for more than 7 days of each annual leave.

In order to be able to recognize a plurality of editing intents contained in an application editing requirement at the same time, a solution is provided in the present embodiment.

As described with reference to fig. 12, the application development assistant can recognize a plurality of editing intents included in the application editing requirements at the same time.

It will be appreciated that in different dialog scenarios (which may be understood as different functional pages), the user operations that the system can handle are a limited set, such as in the platform front page, which generally only allows creation operations of applications; and on a workflow page, intelligently performing adding, deleting and modifying operations of the flow nodes, and the like. Therefore, the application can pre-configure and maintain the corresponding relation between a dialogue scene and the processable editing intention list, namely, each dialogue scene in the scene library in fig. 12 has the corresponding relation with the editing intention in the intention library.

On this basis, a target dialog scene, such as an ID, name, description information, etc., of the target dialog scene when the user inputs an application editing requirement, may be determined. Furthermore, the intention list corresponding to the target dialogue scene can be obtained by referring to the preconfigured corresponding relation, and the intention list contains all editing intents which can be processed in the target dialogue scene.

To identify the user's editing intent, an artificial intelligence model may be invoked in this embodiment to identify each editing intent of the user in the application editing requirements with reference to the intent list.

An alternative embodiment may include:

s11, acquiring a first prompt instruction prompt template, wherein the template comprises an intention bill slot and an application editing requirement slot and is used for indicating a model to refer to each editing intention in the intention bill slot and analyzing each editing intention contained in the application editing requirement slot.

S12, assembling the intention list into the intention list groove, assembling the application editing requirement into the application editing requirement groove to obtain a first prompt instruction prompt, and sending the first prompt instruction prompt into the artificial intelligent model to obtain a target editing intention output by the model.

Examples are as follows:

aiming at the developed off-school management application, a user puts forward an application editing requirement: "I want to add a teacher role at educational administration and create a statistical analysis page for it, realize leave school more than 7 days, 14 days, 30 days of leave, statistics data and details of leave more than 7 days of each age. The corresponding intention list under the dialogue scene comprises: { create roles, create pages, add statistics charts }.

The assembled first hint instruction prompt may be:

"you are an NLP expert, requiring you to analyze the user's literal description intent.

The intents supported in the current scenario are known to include: { create roles, create pages, add statistics charts }.

Please edit the demand for the user's application < i want to add a teacher role at the educational administration department and create a statistical analysis page for it, realize leaving the school more than 7 days, 14 days, 30 days of leave, statistics data and detail > analysis of the first layer edit intention in the user demand of each age of leave more than 7 days. "

By adopting the scheme provided by the embodiment, the edit intention list which can be processed under each dialogue scene is arranged in advance, so that the artificial intelligent model can refer to the list to identify each edit intention of the user in the application edit requirement, and the accuracy of multi-intention identification is improved.

In another alternative embodiment, considering that there may be a dependency relationship between different editing intents, for example, under an application overview page, the hierarchical order of precedence between the editing intents is: create roles- > create pages- > add statistics charts. Accordingly, the respective editing intents included in the above-described intention list may be arranged in a hierarchical order. Based on this, the above-mentioned process of invoking artificial intelligence model to identify each editing intention of the user in the application editing requirement with reference to the intention list may specifically include:

S21, a first prompt instruction prompt template is obtained, wherein the template comprises an intention list groove and an application editing requirement groove, and the template is used for indicating a model to analyze the editing intention of a first layer contained in the application editing requirement groove according to the layer sequence of all editing intents in the intention list groove.

S22, assembling the intention list into the intention list groove, assembling the application editing requirement into the application editing requirement groove to obtain a first prompt instruction prompt, and sending the first prompt instruction prompt into the artificial intelligent model to obtain a target editing intention output by the model.

Examples are as follows:

aiming at the developed off-school management application, a user puts forward an application editing requirement: "I want to add a teacher role at educational administration and create a statistical analysis page for it, realize leave school more than 7 days, 14 days, 30 days of leave, statistics data and details of leave more than 7 days of each age. The corresponding intention list under the dialogue scene comprises: create roles- > create pages- > add statistics charts.

The assembled first hint instruction prompt may be:

The intents and hierarchical relationships supported in the current scene are known as follows:

create roles- > create pages- > add statistics charts.

And sending the first prompt instruction prompt to an artificial intelligent model, wherein the model outputs the editing intention which is ranked first and is contained in the application editing requirement of the user preferentially according to the ranking order of all editing intents in the intention list.

The final target editing intention of the model output is: a character is created.

S3, judging whether the target editing intention is a leaf editing intention at the tail end in the intention list; if not, step S4 is executed, and if yes, step S5 is executed.

S4, selecting editing intents of all layers after the target editing intention from the intention list according to the layer sequence as a new intention list, and returning to the execution step S2.

Specifically, the leaf editing intention in the above example is "newly added statistical chart". In step S3, whether or not to terminate the intention recognition is determined by determining whether or not the target edit intention identified by the model is a leaf edit intention, and if the identified target edit intention is not a leaf edit intention, it is explained that there is a possibility that other edit intents are not completely identified, and therefore, the intention of each layer after the identified target edit intention in the intention list may be regarded as a new intention list, and step S2 may be executed again until the identified target edit intention in a certain round is a leaf edit intention, and step S5 may be executed.

Still referring to the above example, the identified target editing intent is "create role", which does not belong to leaf editing intent, and editing intent of each level after "create role" is extracted from the intent list, and the resulting new intent list is: creating a page- > newly added statistical chart.

Returning to the execution of step S2 again, the first prompt instruction prompt is obtained as follows:

creating a page- > newly added statistical chart.

The target edit intention of the model output is "create page", which still does not belong to leaf edit intention, so the new intention list is extracted again as: and (3) adding a statistical chart, and returning to execute the step S2 again, wherein the obtained first prompt instruction promt is as follows:

a statistical chart is added.

The target editing intention output by the model is 'five newly added statistical charts', belongs to leaf editing intention, and executes step S5.

S5, calling the artificial intelligent model to acquire slot information required by the target editing intention, and forming each editing intention of an end user by the acquired target editing intention of each layer which is not the leaf editing intention, the target editing intention belonging to the leaf editing intention and the slot information thereof.

Specifically, after the target editing intention of each non-leaf editing intention and the target editing intention as the leaf editing intention are obtained through the above-mentioned multiple iterations, the model may be further invoked to obtain slot information required for the target editing intention as the leaf editing intention. In specific implementation, an artificial intelligence model can be called, and slot information required by the target editing intention is acquired from application editing requirements input by a user.

Taking an example of an application editing requirement input by a user, i want to send a short message to a parent after approval is passed, the short message only comprises an editing intention, namely an added flow node, and the slot information corresponding to the editing intention is as follows:

the steps S1-S5 can obtain the target editing intention (including leaf editing intention and required slot information) of each level identified by the artificial intelligent model, and each editing intention of the end user is composed of the target editing intention of each level.

By adopting the scheme of the embodiment, the hierarchical sequence of each editing intention in the intention list is set by configuring the intention list corresponding to each dialogue scene, and the model can refer to the hierarchical sequence of each editing intention in the intention list in each round by calling the artificial intelligent model for multiple rounds, so that the first-level editing intention in the application editing requirement is identified, the identification of a plurality of editing intents contained in the application editing requirement is realized, and the accuracy of the result of multi-intention identification is further improved.

On the basis of the above embodiment, the target application created for the user can be published into the application market. The user may use the published application in the manner of an existing access application.

In addition, an alternative implementation of accessing an application through the application using a helper is also provided in this embodiment.

As shown in connection with fig. 10-11, various types of applications that have been created may be displayed in the application marketplace interface. At the same time, the application may also wake up on the interface using a helper, as shown in the lower right hand corner of FIG. 10. The user can input the application use requirement of the user in the dialog box of the application use assistant, so that the application can determine the application matched with the application use requirement in the application market, and simultaneously, the service data on which the form page of the matched application depends is extracted and filled by combining the application use requirement and the application information of the matched application, the matched application is started, and the form page of the matched application is filled by the extracted service data. The application interface shown in fig. 11 is shown when the input application use requirement is "six-grade class 1 shift Zhang Xiaoming needs to leave a school, he just needs to go to a hospital with sudden belleville pain, and can come back before school in the afternoon". The application determines that the application matched with the requirement is the resident school generation and school management application based on the requirement, so that the application is started and jumps to the school application form interface, and the business data is extracted and filled by combining the business data relied when the form interface is filled, so that the result shown in fig. 11 is finally obtained.

Therefore, by adopting the scheme of the embodiment of the application, the application use requirement of the user can be automatically analyzed through the application use assistant, the application can be automatically matched, the business data on which the form page of the application is filled is determined and extracted, the automatic filling of the form page is carried out, the user is not required to search the required application in the application market by himself, the manual filling of information in the form page of the application is not required, and the use convenience of the user is greatly improved.

Of course, fig. 10-11 only illustrate one example of a web page application interface, and in addition, the application assistant of the present application may also be disposed at the mobile APP end, and the corresponding interaction interfaces are not illustrated one by one.

An alternative implementation of determining an application matching the application usage requirement entered by the user in the application marketplace is described in this embodiment.

For each application in the application market, the embodiment of the application can submit the application information to the vector library in advance for constructing the vector representation, namely, based on the application information (including but not limited to application basic information, data model, page information, workflow information, role authority information and the like) of each application, the vector representation corresponding to the application is generated, so that the vector representation corresponding to each application in the application market is determined.

On the basis, after the application use requirement of the user is acquired, a first vector representation corresponding to the application use requirement can be determined, vector matching is carried out in a vector library, a matched vector representation is obtained, and an application corresponding to the matched vector representation is used as a matched application.

Of course, other application matching strategies may be designed besides the implementation manner, for example, similarity matching is performed between application use requirements and nouns and description information of each application, and an application with the highest similarity is selected as a matching application, which is not exhaustive in this embodiment.

In the foregoing embodiments, some optional man-machine interaction manners in the application building and using process are illustrated. It can be understood that, during the application building and using process, the system can respond to the request input by the user and give a reply, when the system can identify the requirement of the user and complete the corresponding operation, a success prompt can be given, as shown in fig. 3-11, such as "application building success", "application modification success", etc., otherwise, when the system cannot identify the requirement of the user or cannot complete the corresponding operation, a failure prompt can be given. In practical situations, the failure may be caused by multiple reasons, and the failure prompts of different reasons may also be different, and the following table 2 illustrates several different contents of the failure prompt messages:

TABLE 2

The application building device provided by the embodiment of the application is described below, and the application building device described below and the application building method described above can be referred to correspondingly.

Referring to fig. 13, fig. 13 is a schematic structural diagram of an application building device according to an embodiment of the present application.

As shown in fig. 13, the apparatus may include:

a development requirement acquisition unit 11 for acquiring an application development requirement input by a user;

a standard executable code generating unit 12, configured to invoke a preconfigured artificial intelligence model to parse the application development requirements to generate executable code of a standard protocol;

a protocol conversion unit 13, configured to convert an executable code of the standard protocol into an executable code of the target protocol according to a mapping relationship between the standard protocol and the target protocol that can be identified by the low code platform;

and the executable code rendering unit 14 is used for rendering the executable code of the target protocol through the low-code platform to obtain the built target application.

Optionally, the apparatus of the present application may further include:

and the application overview display unit is used for displaying the functional overview and detail page of the target application after the target application is obtained, so that a user can browse the functional page and the detail of the application.

Optionally, the apparatus of the present application may further include:

the function query instruction acquisition unit is used for acquiring an application function query instruction input by a user, wherein the query instruction is used for querying whether the target application can realize a target function;

and the function query instruction response unit is used for analyzing the application function query instruction, judging whether the target application has the target function or not and outputting a judging result.

Optionally, the apparatus of the present application may further include:

the application editing requirement acquisition unit is used for acquiring application editing requirements input by a user, wherein the application editing requirements are used for indicating the editing of the target application;

an editing intention recognition unit for recognizing an editing intention of a user based on the application editing requirement, wherein the editing intention comprises an operation type and an operation object, and the operation type comprises addition, deletion and/or modification;

and the application editing unit is used for editing the target application based on the editing intention of the user to obtain the edited target application.

Optionally, the apparatus of the present application may further include:

the editing page display unit is used for displaying an application editing page, and the application editing page contains application information related to the current editing in the target application.

Optionally, the apparatus of the present application may further include:

a structured processing unit for processing the application development requirements into structured text before processing by the standard executable code generation unit;

the process of the standard executable code generating unit calling the preconfigured artificial intelligent model to analyze the application development requirement to generate the executable code of the standard protocol comprises the following steps:

Optionally, the process of the standard executable code generating unit calling the artificial intelligence model and converting the structured text into the executable code of the standard protocol includes:

Optionally, if the editing intention of the user is more than one, the editing intention identifying unit identifies the editing intention of the user based on the application editing requirement, including:

Optionally, when a hierarchical order exists between the editing intents in the intent list, the editing intent identifying unit invokes the artificial intelligence model to identify, with reference to the intent list, a process of identifying each editing intent of the user in the application editing requirement, including:

Optionally, the process of editing the target application by the application editing unit based on the editing intention of the user to obtain the edited target application includes:

Optionally, the built target application is released in the application market, and the device may further include:

the application use requirement acquisition unit is used for acquiring application use requirements set by a user in the application market;

a matching application determining unit, configured to determine an application matching the application use requirement in the application market;

the application form page filling unit is used for extracting and filling the business data on which the form page of the matched application depends in combination with the application use requirement and the matched application information;

and the application pulling unit is used for starting the matched application and filling the form page of the matched application by using the service data.

Optionally, the process of determining, by the matching application determining unit, an application matching the application use requirement in the application market includes:

The application building device provided by the embodiment of the application can be applied to application building equipment such as terminals, servers, cloud ends and the like. Alternatively, fig. 14 shows a block diagram of a hardware structure of the application building apparatus, and referring to fig. 14, the hardware structure of the application building apparatus may include: at least one processor 1, at least one communication interface 2, at least one memory 3 and at least one communication bus 4;

in the embodiment of the application, the number of the processor 1, the communication interface 2, the memory 3 and the communication bus 4 is at least one, and the processor 1, the communication interface 2 and the memory 3 complete the communication with each other through the communication bus 4;

processor 1 may be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement embodiments of the present application, etc.;

The memory 3 may comprise a high-speed RAM memory, and may further comprise a non-volatile memory (non-volatile memory) or the like, such as at least one magnetic disk memory;

wherein the memory stores a program, the processor is operable to invoke the program stored in the memory, the program operable to:

acquiring application development requirements input by a user;

Alternatively, the refinement function and the extension function of the program may be described with reference to the above.

The embodiment of the present application also provides a storage medium storing a program adapted to be executed by a processor, the program being configured to:

acquiring application development requirements input by a user;

Finally, it is further noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

In the present specification, each embodiment is described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and may be combined according to needs, and the same similar parts may be referred to each other.

The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An application building method, comprising:

acquiring application development requirements input by a user;

2. The method as recited in claim 1, further comprising:

3. The method as recited in claim 1, further comprising:

4. The method as recited in claim 1, further comprising:

5. The method as recited in claim 4, further comprising:

6. The method of claim 4, wherein the operation object comprises a function page and content or components in the function page, wherein the function page comprises any one of a form page, a statistics page, and a flow page.

7. The method of claim 1, further comprising, prior to said invoking the preconfigured artificial intelligence model to parse the application development requirements to generate executable code for a standard protocol:

processing the application development requirements into a structured text;

8. The method of claim 7, wherein invoking the artificial intelligence model to convert the structured text into executable code of a standard protocol comprises:

9. The method of claim 4, wherein the user's editing intent is more than one, and wherein the process of identifying the user's editing intent based on the application editing requirements comprises:

10. The method of claim 4, wherein editing the target application based on the user's editing intent, the process of obtaining an edited target application, comprises:

11. The method of claim 7, wherein the application development requirements are multimodal information comprising: application development requirements of any one of a text modality, an audio modality, and a visual modality.

12. The method according to any one of claims 1-11, wherein the built target application is published in an application marketplace, the method further comprising:

13. The method of claim 12, wherein determining in the application marketplace an application that matches the application usage requirement comprises:

14. An application building device, comprising:

15. An application building apparatus, comprising: a memory and a processor;

the memory is used for storing programs;

the processor is configured to execute the program to implement the respective steps of the application building method according to any one of claims 1 to 13.

16. A storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the application building method according to any of claims 1-13.