KR102539283B1 - Method of generating code using artificial intelligence - Google Patents

Abstract

Provided in one embodiment of the present invention is a method for generating a code by using artificial intelligence, which is executed by a computing device, and comprises: (a) a step of reading information of a marker attached to each of a plurality of objects constituting a graphic user interface; (b) a step of generating a graphic user interface structure based on the information of the marker of each of the plurality of objects; (c) a step of generating first modeling information corresponding to the graphic user interface by executing one of an image analysis model and a text analysis model by using artificial intelligence with respect to each of the plurality of objects based on the information of the marker of each of the plurality of objects; and (d) a step of generating a first code by using the artificial intelligence based on the graphic user interface structure and the first modeling information. Therefore, a code may be more precisely generated.

Description

Code generation method using artificial intelligence {METHOD OF GENERATING CODE USING ARTIFICIAL INTELLIGENCE}

The present disclosure relates to a code generation method using artificial intelligence, and more specifically, to a code generation method using marker information attached to an object and artificial intelligence.

When a designer creates a graphical user interface (hereinafter simply referred to as a "GUI") of a web page or application (hereinafter simply referred to as an "app"), a developer writes actual operating code based on the GUI. In other words, the developer writes the code for the web page or app.

In order to perform these tasks more smoothly and quickly, a method of generating codes using, for example, artificial intelligence has been developed. For example, a method such as "Pix2Code" has been developed.

Pix2Code refers to a method of analyzing a GUI image (also referred to as a “screen shot”) using a learned neural network (artificial intelligence) and automatically generating an appropriate code. For example, in a paper titled "pix2code: Generating Code from a Graphical User Interface Screenshot" published by Tony Beltramelli (see https://arxiv.org/abs/1705.07962), an overview of Pix2Code is described.

Pix2Code is a learning-based method, that is, a method of learning how to write codes by oneself by looking at a code corresponding to a screen shot of a GUI using a deep neural network. Therefore, it is different from a method of automating by inputting patterns or rules of GUI coding into an automation program, that is, a rule-based method.

More specifically, if the existing automation method (e.g., rule-based method) is implemented through the internal structure of the language or the internal API (Application Programming Interface), Pix2Code works like the actual tasks performed by front-end developers, expressed in pixels. The difference is that the code is written using the screen shot of the GUI.

According to Pix2Code, training processing and sampling processing are performed.

According to the learning process, the screen shot of the GUI is encoded by a convolutional neural network (CNN)-based vision model, and the context is a stack of long short-term memory (LSTM) layers ( It is encoded by a language model consisting of stack). A context is a set of encoded tokens corresponding to a domain specific language (DSL). The results encoded by the CNN-based vision model (eg feature vectors) and the language model (eg feature vectors) are combined and provided to the second stack (acting as a decoder) of the LSTM layer. The softmax layer samples one token at a time, and the size of the output of the softmax layer corresponds to the size of the DSL vocabulary. Thus, given an image and a sequence of tokens, it can be optimized end-to-end through gradient descent to predict the next token in the sequence. In the sampling process, the input context is updated for each prediction to include the last prediction token. The resulting sequence of DSL tokens is compiled into the desired target language using existing compiler design techniques.

However, Pix2Code has a disadvantage in that it is difficult to process GUI screen shots with complex structures and cannot extract specific information.

In addition, even if Pix2Code is used, pre-processing is required to separate the image part and the context part based on the GUI screen shot for smooth processing. In the preprocessing, for example, whether the object included in the screen shot of the GUI is an image object by identifying the outline (boundary) using a method of dividing the screen shot of the GUI and determining a feature value for each divided part. Classify whether it is a context object. However, this operation takes a lot of processing time. In particular, if there is an error in preprocessing, for example, a specific object is processed using a CNN-based vision model, and if the specific object corresponds to a text object, the specific object must be processed again using the LSTM layer stack. Therefore, unnecessary processing may be performed.

1. Korean Patent Publication No. 10-2022-0125282

1. https://arxiv.org/abs/1705.07962

The purpose of the technology described herein is to minimize the processing load required to analyze an object by using marker information attached to the object, and automatically more accurately code the code using artificial intelligence even for a GUI having a complex structure. It is to provide a code generation method using artificial intelligence that can generate.

In order to achieve the above object, according to one form of technology described herein, performed by a computing device, (a) reading marker information attached to each of a plurality of objects constituting a graphical user interface; (b) generating a graphic user interface structure based on the marker information of each of the plurality of objects; (c) By performing any one of an image analysis model and a text analysis model for each of the plurality of objects based on the marker information of each of the plurality of objects using artificial intelligence, corresponding to the graphic user interface generating first modeling information; and (d) generating a first code using the artificial intelligence based on the graphic user interface structure and the first modeling information.

According to the technology described herein, the processing load required to analyze an object is minimized by using marker information attached to the object, and the code is automatically more accurately generated using artificial intelligence even for a GUI having a complex structure. can create

1 is an exemplary flow diagram of a code generation method using artificial intelligence according to a first embodiment of the technology described herein;
2 is a diagram showing a schematic configuration of a computing device executing a code generation method using artificial intelligence according to a first embodiment of the technology described herein;
3 is a diagram exemplarily illustrating a plurality of objects constituting a graphic user interface and marker information attached to the objects in a code generation method using artificial intelligence according to a first embodiment of the technology described herein.
4 is a diagram showing an example of a graphical user interface structure in a code generation method using artificial intelligence according to a first embodiment of the technology described herein.
5 is an exemplary flow diagram of a code generation method using artificial intelligence according to a second embodiment of the technology described herein;
6 illustrates an example of replacing a first object with a second object generated by changing processing information of the first object in a code generation method using artificial intelligence according to a second embodiment of a technique described herein. floor plan.
7 is an exemplary flow diagram of a code generation method using artificial intelligence according to a third embodiment of the technology described herein;

Hereinafter, an embodiment of a code generation method using artificial intelligence according to the technology described herein will be described in more detail with reference to the accompanying drawings. Meanwhile, in the drawings for explaining the embodiments of the technology described herein, for convenience of description, only some of the actual configurations may be shown, some may be omitted, or may be shown in a modified or different scale.

<First Embodiment>

1 is an exemplary flowchart of a code generation method using artificial intelligence according to a first embodiment of the technology described herein, and FIG. 2 is a code generation method using artificial intelligence according to a first embodiment of the technology described herein. It is a diagram showing a schematic configuration of a computing device that executes.

First, referring to FIG. 2 , a computing device 100 executing a code generation method using artificial intelligence according to a first embodiment of the technology described herein will be described.

Referring to FIG. 2 , a computing device 100 executing a code generation method using artificial intelligence according to a first embodiment of the technology described herein includes a communication interface 110, an arithmetic processing unit 130, and storage may include section 150 .

The communication interface 110 is a communication interface based on a wired/wireless communication method. The communication interface 110 may be implemented by a semiconductor device such as a communication chip.

The arithmetic processing unit 130 may be implemented by a semiconductor device such as a central processing unit (CPU), a graphics processing unit (GPU), and an application specific integrated circuit (ASIC).

The arithmetic processing unit 130 may be implemented using, for example, a plurality of semiconductor devices. For example, the arithmetic processing unit 130 may include a first semiconductor device that performs a communication function, a second semiconductor device that performs an image analysis model described later using artificial intelligence, and a text analysis model described later using artificial intelligence. It may be implemented using a third semiconductor device.

The calculation processing unit 130 executes the code generation method using artificial intelligence according to the first embodiment, and the communication interface 110 and the storage unit 150 to execute the code generation method using artificial intelligence according to the first embodiment. ) can be controlled.

The storage unit 150 stores data. The storage unit 150 may be implemented by a semiconductor device such as, for example, a semiconductor memory. The storage unit 150 may store, for example, modeling information such as an object described later, marker information of an object, a graphic user interface structure, first modeling information, and a code such as a first code.

The storage unit 150 may be implemented using, for example, a plurality of semiconductor devices.

Hereinafter, a code generation method using artificial intelligence according to a first embodiment of the technology described herein will be described in more detail with reference to FIG. 1 .

First, the computing device 100, more specifically, the arithmetic processing unit 130 reads marker information attached to each of a plurality of objects constituting the graphic user interface (step S110).

3 is a diagram exemplarily illustrating a plurality of objects constituting a graphic user interface and marker information attached to the objects in the code generation method using artificial intelligence according to the first embodiment of the technology described herein.

Hereinafter, it will be described in more detail based on a graphic interface in the case of creating a web page, for example.

Referring to FIG. 3 , a number of objects, such as objects 210 to 280 , are shown.

Object 210, object 220, object 230, object 240, and object 250 are "PRODUCT", "PRICE", and "LATEST", respectively. It is a text object (can be inserted in the form of an image) representing "FREE TRIAL" and "BABY AND DOG".

Object 260 is, for example, an image object representing “baby”.

The object 270 is, for example, an image object representing "dog".

The object 280 is, for example, an image object representing an interface for content reproduction.

Referring to FIG. 3 , marker information 310 to marker information 380 corresponding to each of the objects 210 to 280 are shown.

The marker information 310 to marker information 380 may be designated according to a previously designated marker information designation rule. For example, the marker information 310 to marker information 380 may be designated using Cascading Style Sheets (CSS). Alternatively, the marker information 310 to marker information 380 may be designated using natural language.

That is, the marker information 310 to marker information 380 are specified in a form that the arithmetic processing unit 130 of the computing device 100 can understand. For example, when creating an object, a designer may designate marker information corresponding to an object according to a predetermined marker information designation rule or by using natural language.

When marker information is specified using CSS, since CSS tags can be used without restrictions, marker information can be specified more efficiently. Also, even when CSS is extended, marker information can be specified more efficiently by applying the extended CSS.

Referring to FIG. 3 , marker information 310 is attached to an object 210 . For example, the marker information 310 is "List combine_ProductDB_2022.12.16", meaning that a list combined in a database called "ProductDB_2022.12.16" is displayed.

Marker information 320 is attached to the object 220 . For example, the marker information 320 is "List combine_PriceDB_2022.12.16", which means that a list combined from a database called "PriceDB_2022.12.16" is displayed.

Marker information 330 is attached to the object 230 . For example, the marker information 330 is "Goto Home_Page5_LatestPage", and clicking the object 230 means moving to "Home_Page5_LatestPage".

Marker information 340 is attached to the object 240 . For example, the marker information 340 is "Goto Home_section4_TrialPage", and clicking the object 240 means moving to "Home_section4_TrialPage".

Marker information 350 is attached to the object 250 . For example, the marker information 350 is "Text_Image_scaleVector-Web", and the object 250 is a text image, meaning that it is in the form of a vector.

Marker information 360 is attached to the object 260 . For example, the marker information 360 is "Image of Baby", which means that the object 260 is an image object for "baby".

Marker information 370 is attached to the object 270 . For example, the marker information 370 is "Image of Dog", which means that the object 270 is an image object for "dog".

Marker information 380 is attached to the object 280 . For example, the marker information 380 is "Goto Home_section10_Play_Video_001", and clicking the object 280 means to play "Video_001" of "Home_section10".

Hereinafter, marker information will be described in more detail based on a first object that is one of a plurality of objects.

The marker information of the first object may include at least one of object classification information indicating whether the first object is an image object or text object and processing information related to the first object.

Object classification information may be set, for example, such as "Image of Baby", which is marker information 360, for the object 260, which is an image object.

The object classification information may be set as “Text_Image_scaleVector-Web” which is the marker information 360 for the object 250 which is a text object, for example.

Although not shown in FIG. 3 , the object classification information may be further included in the marker information 310 for the object 210 that is, for example, a text object.

The process information related to the first object is information indicating a process to be performed on the first object.

For example, when the first object is a text object, processing information of the first object may include at least one of text, used language, used font, color, size, display position, and display method.

For example, when the first object is an image object, processing information of the first object may include at least one of the image size, color, display position, and display method.

Also, the processing information of the first object may indicate processing performed when the first object is clicked, for example, as described above.

That is, as described above, processing information of the first object may be set so that when the first object is clicked, a specific page is moved or specific content is reproduced.

Meanwhile, some of the processing information may also be referred to as action information.

The action information of the first object is information indicating whether a specific process (action) is performed on the first object when a user input is received in relation to the first object or regardless of the user input.

For example, among the methods of displaying the first object, there are various display methods such as display in a pop-up form, display in a splash form, or rotated display.

The action information of the first object may indicate information on such a display method.

The processing information of the first object may be designated according to at least one of specifications of a device to be provided and specifications of a browser.

For example, the processing information of the first object is first processing information corresponding to when accessing the web page by a general computing device (eg, PC), and second processing information corresponding to when accessing the web page by a first mobile device (eg, smart phone). It may be designated in various forms, such as processing information and third processing information corresponding to accessing a web page by a second mobile device (eg, a tablet).

Similarly, the processing information of the first object may be designated in various forms, such as fourth processing information or fifth processing information, according to browser specifications (eg, manufacturer or browser version).

The size of the marker information of the first object is not particularly limited. Therefore, even in the case of having complex processing information (including action information), the marker information of an object can sufficiently include the aforementioned processing information. In particular, information necessary for generating a graphic user interface structure to be described later, information necessary for generating first modeling information (more specifically, information necessary for easily and quickly generating first modeling information using artificial intelligence) may be included in the marker information of the first object.

The object marker information (more specifically, object classification information, processing information, and action information) described with reference to FIG. 3 is merely an example and may be changed in various forms.

As in the example described with reference to FIG. 3 , marker information of an object is attached in advance to correspond to the object, and marker information of the object is read in step S110.

Next, the arithmetic processing unit 130 generates a graphic user interface structure based on marker information of each of a plurality of objects constituting the graphic user interface (step S120).

In step S120, a graphic user interface structure may be generated based on marker information of each of a plurality of objects and data attributes of each of a plurality of objects.

That is, when the marker information of the first object does not include the data attribute of the first object, the graphic user interface structure may be created by additionally using the data attribute of the first object.

Data attributes are properties such as text (i.e. content), used font, color, size, display position and display method in the case of a text object, for example.

The data properties may include, for example, at least one of properties such as size, color, display position, and display method of an image object.

4 is a diagram illustrating an example of a structure of a graphic user interface in a code generation method using artificial intelligence according to a first embodiment of a technique described herein.

The graphic user interface structure is a structure composed of, for example, data such as a plurality of screens (pages), data attributes of objects arranged in the plurality of screens, and correspondence between objects.

In the case of a web page, for example, the graphic user interface structure may have a form in which information on a plurality of objects is connected to a so-called “site map”.

For example, referring to FIG. 4 , for a category (“New Arrival”, etc.) and a sub-category (“Sweater”) displayed on a screen (“Women”, “Men”, etc.), a graphic in the form of a plurality of corresponding objects connected to each other. A user interface structure is shown.

If the graphic user interface structure is created through step S120 in this way, it is possible to improve the disadvantage that, for example, Pix2Code cannot extract detailed information because it is difficult to process a GUI screen shot having a complicated structure.

Next, the calculation processing unit 130 performs any one of an image analysis model and a text analysis model for each of a plurality of objects based on the marker information of each of the plurality of objects using artificial intelligence, thereby providing a graphical user interface. Corresponding first modeling information is generated (step S130).

The image analysis model may include at least one of a convolutional neural network (CNN)-based model, a region-based convolutional neural network (R-CNN)-based model, and a recurrent neural network (RNN)-based model.

The text analysis model may include a Long Short-Term Memory (LSTM) hierarchical model.

As described above, marker information of an object may include object classification information. Therefore, it can be clearly identified whether the object is an image object or a text object.

In the case of using Pix2Code, preprocessing is required to separate the image part and the context part based on the image (screen shot) of the GUI. That is, in order to determine whether an object is a text object or an image object, a process of identifying the outline (boundary) of each part of the screen shot must be performed in units of pixels.

However, in step S130, it is already possible to clearly confirm whether the object is an image object or a text object based on the marker information of the object, so that the process of analyzing the object can be easily performed.

Hereinafter, a description will be made based on a first object that is one of a plurality of objects.

The arithmetic processing unit 130 uses marker information of the first object (ie, object classification information of the first object) to perform an image analysis model on the first object when the first object represents an image object. .

The calculation processing unit 130 uses marker information of the first object (ie, object classification information of the first object) to perform a text analysis model on the first object when the first object represents a text object. .

Accordingly, it is possible to minimize the processing load required to analyze an object using artificial intelligence.

By analyzing a plurality of objects using artificial intelligence, first modeling information corresponding to a graphic user interface is generated.

The first modeling information includes learned feature information for each of a plurality of objects.

For example, the learned feature information on the object 250 described above may be information such as “a baby wearing a diaper”. The learned feature information of the object 250 may further include learned feature information such as baby's hair color information.

For example, the learned feature information for the object 260 described above may be information such as “puppy”. The learned feature information of the object 260 may further include learned feature information such as puppy's fur color information.

Next, the operation processing unit 130 generates a first code using artificial intelligence based on the graphical user interface structure generated through step S120 and the first modeling information generated through step S130 (step S140).

Since the graphical user interface structure can be more accurately generated through step S120 and the first modeling information (ie, learned feature information for a plurality of objects) generated through step S130 can be generated while minimizing the processing load, The first code can be easily and more accurately generated.

The first code includes at least one of HTML code and application code corresponding to the web page or application.

Meanwhile, referring to FIG. 1 , the calculation processing unit 130 may remove marker information attached to each of a plurality of objects (step S150).

That is, once the graphical user interface structure is created (and the first modeling information is created), the marker information can be removed. Step S150 may be performed before step S140.

In this way, according to the first embodiment, the processing load required to analyze an object is minimized by using the marker information attached to the object, and the code is automatically more accurately coded using artificial intelligence even for a GUI having a complex structure. can create

<Second Embodiment>

5 is an exemplary flow diagram of a code generation method using artificial intelligence according to a second embodiment of the technology described herein.

The code generation method using artificial intelligence according to the second embodiment of the technology described herein may be performed after performing the code generation method using artificial intelligence according to the first embodiment described above.

A code generation method using artificial intelligence according to a second embodiment relates to generating a second code by replacing a first object with a second object.

Hereinafter, a second embodiment will be described assuming that the first object is a text object.

The calculation processing unit 130 converts the first modeling information into the second modeling information by replacing the first object with the second object generated by changing the processing information of the first object (step S210).

6 is a diagram illustrating an example in which a first object is replaced with a second object in a code generation method using artificial intelligence according to a second embodiment of a technique described herein.

For example, the object 250 illustrated with reference to FIG. 3 is a text object “BABY AND DOG”. That is, the processing information of the object 250 includes the text "BABY AND DOG" corresponding to English, which is the first language.

For example, to provide a web page or application in a second language other than English, such as German, a text object including processing information “BABY AND HUND” corresponding to the second language, German, should be used.

Accordingly, through step S210, the object 250 is replaced with the object 290 including processing information of “BABY AND HUND”.

Object 290 may be generated using artificial intelligence based on user input, for example.

Referring to FIG. 6 , an example in which an object 250 is replaced with an object 290 is shown, and in particular, marker information of the objects 210 to 240, the object 240, and the objects 260 to 290 is shown. A removed state is shown.

Meanwhile, FIG. 6 shows an example in which the language of the text of the object 250 is changed, but the present embodiment is not limited thereto. For example, replacing the object 250 with the object 290 created by changing at least one of text (ie, content), used font, color, size, display position, and display method for the text of the object 250. Even in this case, this embodiment can be preferably applied.

In addition, even when the first object is an image object rather than a text object and the first object is replaced with a second object generated by changing processing information of the first object, the present embodiment can be preferably applied.

As the object 250 is replaced with the object 290 , the first modeling information according to the first embodiment is converted into second modeling information replaced with the object 290 instead of the object 250 .

Next, the operation processing unit 130 generates a second code using artificial intelligence based on the graphic user interface structure and the second modeling information generated in step S210 (step S220).

The structure of the graphic user interface may be partially changed as the object 250 is replaced with the object 290 .

In this way, according to the second embodiment, in addition to the above effects according to the first embodiment, the second code can be automatically generated by replacing the first object with the second object.

<Third Embodiment>

7 is an exemplary flow diagram of a code generation method using artificial intelligence according to a third embodiment of the technology described herein.

The code generation method using artificial intelligence according to the third embodiment of the technology described herein is performed after performing the code generation method using artificial intelligence according to the first embodiment described above.

A code generation method using artificial intelligence according to the third embodiment is used to generate third codes by replacing a first object group with a second object group obtained by collectively changing action information of one or more objects.

The operation processing unit 130 converts the first modeling information into third modeling by replacing a first object group including one or more objects among a plurality of objects with a second object group in which action information of each of the one or more objects is collectively changed. It is converted into information (step S310).

For example, the first object group is a group in which “display in pop-up form” is set as action information. As the action information of the first object group, a second object group is created by collectively changing from “display in pop-up form” to “display in splash form”. Thereafter, the first modeling information is converted into third modeling information by replacing the first object group with the second object group.

Next, the calculation processing unit 130 generates a third code using artificial intelligence based on the graphic user interface structure and the third modeling information generated in step S310 (step S320).

The structure of the graphic user interface may be partially changed as the first object group is replaced with the second object group.

In this way, according to the third embodiment, in addition to the above effects according to the first embodiment, the third code can be automatically generated by replacing the first object group with the second object group.

<Another Example>

Although the embodiments of the technology described herein have been specifically described, this is merely an example of the technology described herein, and those of ordinary skill in the art to which the technology described herein belongs will Various modifications will be possible within a range that does not deviate from the essential characteristics of the technology.

For example, the technology described herein may be applied to a code generation device using artificial intelligence.

A code generating device using artificial intelligence according to the technology described herein may be implemented by, for example, the computing device 100 described above.

That is, according to another form of the technology described herein, the communication interface 110, the calculation processing unit 130, and the storage unit 150 are included, and the calculation processing unit 130 includes (a) a graphic user processing of reading marker information attached to each of a plurality of objects constituting the interface; (b) processing of generating a graphical user interface structure based on the marker information of each of the plurality of objects; (c) By performing any one of an image analysis model and a text analysis model for each of the plurality of objects based on the marker information of each of the plurality of objects using artificial intelligence, corresponding to the graphic user interface processing of generating first modeling information; and (d) generating a first code using the artificial intelligence based on the graphic user interface structure and the first modeling information.

Other features of the code generation method using artificial intelligence according to the technology described herein may also be applied to the code generation device using artificial intelligence according to the technology described herein.

Therefore, the embodiments described herein are intended to explain, not limit, the technology described herein, and the spirit and scope of the technology described herein are not limited by these embodiments. The scope of rights of the technology described herein should be construed by the claims below, and all technologies within the equivalent range should be construed as being included in the scope of the technology described herein.

According to the technology described herein, the processing load required to analyze an object is minimized by using marker information attached to the object, and the code is automatically more accurately generated using artificial intelligence even for a GUI having a complex structure. can create

100: computing device 110: communication interface
130: calculation processing unit 150: storage unit
210, 220, 230, 240, 250, 260, 270, 280, 290: Object
310, 320, 330, 340, 350, 360, 370, 380: marker information

Claims (16)

performed by a computing device;
(a) reading marker information attached to each of a plurality of objects constituting the graphic user interface;
(b) generating a graphic user interface structure based on the marker information of each of the plurality of objects;
(c) By performing any one of an image analysis model and a text analysis model for each of the plurality of objects based on the marker information of each of the plurality of objects using artificial intelligence, corresponding to the graphic user interface generating first modeling information; and
(d) generating a first code based on the graphic user interface structure and the first modeling information using the artificial intelligence;
including,
The marker information of a first object among the plurality of objects includes at least one of object classification information indicating whether the first object is an image object or a text object and processing information related to the first object,
In the step (c),
(c-1) if the object classification information of the first object indicates the image object, performing the image analysis model on the first object; and
(c-2) if the object classification information of the first object indicates the text object, performing the text analysis model on the first object;
Code generation method using artificial intelligence that includes a. According to claim 1,
The code generation method of claim 1 , wherein the marker information of each of the plurality of objects is designated according to a previously designated marker information designation rule. According to claim 2,
Code generation method using artificial intelligence, wherein the marker information designation rule includes Cascading Style Sheets (CSS). According to claim 1,
Code generation method using artificial intelligence, wherein the marker information of each of the plurality of objects is specified using a natural language. According to claim 1,
The image analysis model includes at least one of a Convolutional Neural Network (CNN)-based model, a Region-based Convolutional Neural Network (R-CNN)-based model, and a Recurrent Neural Network (RNN)-based model. How to generate the code used. According to claim 1,
The text analysis model includes a long short-term memory (LSTM) hierarchical model. Code generation method using artificial intelligence. According to claim 1,
(e) removing the marker information prior to step (d)
Code generation method using artificial intelligence further comprising a. According to claim 1,
Wherein the processing information of the first object is specified according to at least one of a specification of a device to be provided and a specification of a browser. According to claim 1,
(f) converting the first modeling information into second modeling information by replacing the first object with a second object generated by changing the processing information of the first object; and
(g) generating a second code based on the graphic user interface structure and the second modeling information using the artificial intelligence;
Code generation method using artificial intelligence further comprising a. According to claim 9,
The first object is the text object, and the processing information of the first object includes at least one of text, a language used, a font used, a color, a size, a display position, and a display method. How to generate the code used. According to claim 1,
The processing information of the first object includes action information related to the first object. According to claim 11,
(h) By replacing a first object group including one or more objects among the plurality of objects with a second object group in which action information of each of the one or more objects is collectively changed, the first modeling information is replaced with third modeling information. converting into information; and
(i) generating a second code based on the graphic user interface structure and the third modeling information using the artificial intelligence;
Code generation method using artificial intelligence further comprising a. According to claim 1,
Wherein the first modeling information includes learned feature information for each of the plurality of objects. According to claim 1,
Wherein the first code includes at least one of HTML code and application code. delete delete

Images