Virtual tabular data generation method and server performing the same

Abstract

The present invention discloses a method for generating virtual table data and a server for performing the same. The method for generating virtual table data comprises the steps of: generating a first prompt for generating a table schema; adjusting the table schema by comparing the table schema generated based on the first prompt with a predefined reference table schema; generating a second prompt by referencing the adjusted table schema; generating table condition data for the first table data generated based on the second prompt; generating a third prompt by referencing the table condition data and the adjusted table schema; and deriving final table data through a verification operation on the second table data generated based on the third prompt.

Description

Technology Field

[0001] The present invention relates to a method for generating virtual table data and a server for performing the same. Specifically, the present invention relates to a method for generating highly specialized virtual data related to a topic desired by a user through prompting on a deep learning model. Background Technology

[0003] The content described in this section merely provides background information regarding the present embodiment and does not constitute prior art.

[0004] Recently, as financial institutions or electronic financial companies provide financial products and services through computing devices, financial transactions conducted online without users directly interacting with employees of such institutions or electronic financial service providers are increasing. With the rise in these non-face-to-face transactions, the importance of technology capable of accurately and quickly processing users' sensitive information (e.g., personal information, financial information, etc.) is growing day by day.

[0005] Furthermore, various services incorporating AI and machine learning technologies are being rapidly developed in the financial industry recently. In this context, a large amount of training data is required to train deep learning modules (e.g., LLM (Large Language Model) or PLM (Pretrained Language Model)) in order to output the data desired by users or developers.

[0006] However, since financial institutions handle documents related to sensitive user information, data containing sensitive user data cannot be used as training data. To resolve this, financial institutions may use an indirect method of replacing sensitive user data with already publicly available data; however, there was a problem in that the publicly available data used in this case had low relevance to actual user data and did not contribute to improving the performance of the deep learning module. The problem to be solved

[0008] The objective of the present invention is to provide a method for generating virtual table data that can generate high-accuracy data usable to improve the performance of a business model for a new service by replacing data containing sensitive customer information.

[0009] In addition, the objective of the present invention is to provide a method for generating virtual table data that can verify and correct generated data so that the generated data possesses accuracy and diversity higher than a predetermined threshold and can reflect the characteristics of sensitive information data.

[0010] In addition, the objective of the present invention is to provide a method for generating virtual table data that can generate high-accuracy data in accordance with the data schema required by the user by automatically updating the prompt for generating virtual table data multiple times.

[0011] The objects of the present invention are not limited to those mentioned above, and other unmentioned objects and advantages of the present invention may be understood from the following description and will be more clearly understood by the embodiments of the present invention. Furthermore, it will be readily apparent that the objects and advantages of the present invention can be realized by the means and combinations thereof set forth in the claims. means of solving the problem

[0013] A method for generating virtual table data according to some embodiments of the present invention comprises: generating a first prompt for generating a table schema; adjusting the table schema by comparing the table schema generated based on the first prompt with a predefined reference table schema; generating a second prompt by referencing the adjusted table schema; generating table condition data for the first table data generated based on the second prompt; generating a third prompt by referencing the table condition data and the adjusted table schema; and deriving final table data through a verification operation on the second table data generated based on the third prompt.

[0014] Additionally, the step of adjusting the table schema may include the step of applying the first prompt to the deep learning module and receiving the table schema as the output of the deep learning module, the step of comparing the table schema with the predefined reference table schema, and the step of adjusting the table schema so that, if the table schema does not include an item included in the predefined reference table schema, the item is included.

[0015] Additionally, the step of generating the first prompt may include receiving a data generation request containing information about table data desired by the user from a user terminal linked to the server, and generating the first prompt based on the information included in the data generation request.

[0016] Additionally, the step of generating the second prompt includes generating the second prompt by referring to the adjusted table schema and the data generation request together, and the step of generating the third prompt may include generating the third prompt by referring to the table condition data, the adjusted table schema, and the data generation request all together.

[0017] Additionally, the step of generating the table condition data may include the step of applying the second prompt to the deep learning module and receiving the first table data as the output of the deep learning module, and the step of generating the table condition data by comparing each column included in the first table data with predefined condition data.

[0018] In addition, the table condition data includes a unary constraint or a binary constraint for each column included in the table data, wherein the unary constraint refers to a condition having any one of predetermined values, and the binary constraint may refer to a condition including an operation expression with another column.

[0019] Additionally, the step of deriving the final table data may include the step of applying the third prompt to the deep learning module and receiving the second table data as the output of the deep learning module, the step of applying the second table data to the data verification module and obtaining a data evaluation result as the output to the data verification module, and the step of determining the final table data based on the data evaluation result.

[0020] Additionally, the second table data includes a larger number of example data than the first table data, and the data verification module can perform diversity verification on the example data included in the second table data and perform verification of whether a condition is satisfied for columns to which a unary condition or a binomial condition is assigned within the example data.

[0021] Additionally, the step of obtaining the data evaluation result may include: a step of deriving a first evaluation value regarding the diversity of each row data or each column data included in the second table data; a step of deriving a second evaluation value regarding whether each column data included in the second table data satisfies a unary condition; a step of deriving a third evaluation value regarding whether each column data included in the second table data satisfies a binomial condition; and a step of determining whether a reference value for the first to third evaluation values ​​is satisfied and generating the data evaluation result including the result.

[0022] Additionally, the step of determining the final table data may include a step of correcting data included in the second table data that does not meet the standard value to a value that meets the standard value, and a step of determining the second table data with the corrections reflected as the final table data.

[0023] Additionally, the method may further include the steps of: correcting the third prompt based on the data evaluation result; applying the corrected third prompt to the deep learning module to regenerate the second table data; applying the regenerated second table data to the data verification module and re-obtaining the data evaluation result as an output to the data verification module; and determining the final table data based on the re-obtained data evaluation result.

[0024] Meanwhile, a server according to some embodiments of the present invention comprises a processor, a memory for loading a computer program executed by the processor, and a database for storing data generated during the execution of the computer program, wherein the computer program comprises the steps of: generating a first prompt for generating a table schema; adjusting the table schema by comparing the table schema generated based on the first prompt with a predefined reference table schema; generating a second prompt by referencing the adjusted table schema; generating table condition data for the first table data generated based on the second prompt; generating a third prompt by referencing the table condition data and the adjusted table schema; deriving final table data through a verification operation on the second table data generated based on the third prompt; and storing the final table data in the database.

[0025] Additionally, the step of adjusting the table schema includes applying the first prompt to the deep learning module and receiving the table schema as the output of the deep learning module, comparing the table schema with the predefined reference table schema, and adjusting the table schema so that the item is included if the table schema does not include an item included in the predefined reference table schema; the step of generating the table condition data includes applying the second prompt to the deep learning module and receiving the first table data as the output of the deep learning module, and generating the table condition data by comparing each column included in the first table data with the predefined condition data; and the step of deriving the final table data includes applying the third prompt to the deep learning module and receiving the second table data as the output of the deep learning module, applying the second table data to a data verification module loaded in memory and obtaining a data evaluation result as the output to the data verification module, and determining the final table data based on the data evaluation result.

[0026] Additionally, the step of obtaining the data evaluation result may include: a step of deriving a first evaluation value regarding the diversity of each row data or each column data included in the second table data; a step of deriving a second evaluation value regarding whether each column data included in the second table data satisfies a unary condition; a step of deriving a third evaluation value regarding whether each column data included in the second table data satisfies a binomial condition; and a step of determining whether a reference value for the first to third evaluation values ​​is satisfied and generating the data evaluation result including the result. Effects of the invention

[0028] The virtual table data generation method of the present invention can generate highly accurate virtual data capable of replacing data containing sensitive customer information (hereinafter referred to as sensitive information data), and can generate virtual data in accordance with a table schema of the form desired by the user. Through this, developers or operators can improve the performance of business models by utilizing highly accurate virtual data when testing or training business models applied to new services.

[0029] Furthermore, the virtual table data generated through the present invention can possess higher accuracy and diversity than predetermined thresholds, and can generate virtual data that reflects the characteristics of sensitive information data through data verification and correction processes. Through this, the present invention resolves security issues regarding sensitive information data while simultaneously generating data in a form desired by the user and utilizing it for operating business models.

[0030] In addition, the present invention can generate high-accuracy table data required by the user by automatically updating the prompt for generating virtual table data multiple times. Through this, the present invention can increase the work efficiency of developers or workers and optimize resources required for developing new services or analyzing existing services by omitting or automating the data collection process required for developing new services.

[0031] In addition to the above, the specific effects of the present invention are described together with the specific details for implementing the invention below. Brief explanation of the drawing

[0033] FIG. 1 is a conceptual diagram illustrating a virtual table data generation system according to some embodiments of the present invention. FIG. 2 is a block diagram illustrating components for a server according to some embodiments of the present invention. FIG. 3 is a flowchart illustrating a method for generating virtual table data according to some embodiments of the present invention. Figure 4 is a diagram illustrating an example of a table data generation process performed at each step of Figure 3. Figure 5 is a flowchart illustrating step S100 of Figure 3. Figure 6 is a block diagram for explaining the flowchart of Figure 5. Figure 7 is a diagram illustrating a neural network model applied to the deep learning module of Figure 6. Figure 8 is a flowchart for explaining step S200 of Figure 3. Figure 9 is a block diagram for explaining the flowchart of Figure 8. Figure 10 is a flowchart illustrating step S300 of Figure 3. Figure 11 is a block diagram for explaining the flowchart of Figure 10. FIG. 12 is a flowchart for illustrating an example of steps S323 and S325 of FIG. 10. Figure 13 is a block diagram for explaining the flowchart of Figure 12. FIG. 14 is a flowchart illustrating another embodiment of step S325 of FIG. 10. Figure 15 is a block diagram for explaining the flowchart of Figure 14. FIG. 16 is a diagram illustrating a hardware implementation of a device or system for performing a virtual table data generation method according to some embodiments of the present invention. Specific details for implementing the invention

[0034] Terms and words used in this specification and claims shall not be interpreted as being limited to their general or dictionary meanings. In accordance with the principle that an inventor may define the concept of a term or word to best describe their invention, they shall be interpreted in a meaning and concept consistent with the technical spirit of the invention. Furthermore, since the embodiments described in this specification and the configurations illustrated in the drawings are merely one embodiment of the invention and do not represent the entire technical spirit of the invention, it should be understood that various equivalents, modifications, and applicable examples capable of replacing them may exist at the time of filing this application.

[0035] The terms first, second, A, B, etc., as used in this specification and claims may be used to describe various components, but said components should not be limited by said terms. These terms are used solely for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be named the second component, and similarly, the second component may be named the first component. The term "and / or" includes a combination of a plurality of related described items or any of a plurality of related described items.

[0036] The terms used in this specification and claims are used merely to describe specific embodiments and are not intended to limit the invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this application, terms such as "comprising" or "having" should be understood as not precluding the existence or addition of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification.

[0037] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which this invention pertains.

[0038] Terms such as those defined in commonly used dictionaries should be interpreted as having meanings consistent with their meanings in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.

[0039] In addition, each component, process, procedure, or method included in each embodiment of the present invention may be shared within a scope that is not technically contradictory to one another.

[0040] Meanwhile, machine learning is a field of AI that develops algorithms and technologies enabling computers to learn from data. As a core technology in various fields such as data processing, image recognition, speech recognition, and internet search, it demonstrates excellent performance in extracting relevant information and generating new data.

[0041] Hereinafter, with reference to FIGS. 1 to 16, we will examine a method for generating virtual table data according to some embodiments of the present invention and a server for performing the same.

[0043] FIG. 1 is a conceptual diagram illustrating a virtual table data generation system according to some embodiments of the present invention.

[0044] Referring to FIG. 1, a virtual table data generation system according to some embodiments of the present invention includes a server (100) and a user terminal (200). Additionally, in some embodiments, the virtual table data generation system may further include a deep learning server (300). Here, the server (100) operates in conjunction with the user terminal (200) or the deep learning server (300) through a communication network (400) and can transmit and receive data with the user terminal (200) or the deep learning server (300).

[0045] First, the server (100) can generate data requested by the user terminal (200) using a deep-learning model. At this time, the user terminal (200) can transmit a data generation request to the server (100) that includes information about the type (e.g., financial data) and format (e.g., table data) of the data desired by the user.

[0046] For example, the data requested by the user terminal (200) may include personal information, financial activity information, etc. In this case, the financial activity information may include information on financial products such as deposits, savings, loans, etc., and customer consumption pattern information such as card payment history, etc. However, this is merely one example and the present invention is not limited thereto.

[0047] For convenience of explanation, the following description will be given as an example in which a user terminal (200) requests 'financial data' in the form of table data from a server (100).

[0048] The server (100) can generate a prompt reflecting a data generation request received from a user terminal (200) and apply it to a pre-trained deep learning model to generate virtual table data (i.e., virtual table data). At this time, the server (100) performs the operation of updating the prompt multiple times to obtain more accurate table data.

[0049] Here, 'virtual table data' includes virtual data generated by a deep learning model, unlike actual data regarding a specific user or anonymous data related to an anonymous user. In some embodiments of the present invention, the data generated by the deep learning model is not limited to table data.

[0050] However, for the sake of convenience of explanation, the deep learning model of the present invention generating virtual table data will be described below as an example. Additionally, in this specification, virtual table data and table data will be used interchangeably to refer to each other with the same meaning.

[0051] The virtual table data generated by the server (100) may have a very high degree of similarity to actual data (e.g., financial data containing sensitive personal information) and may include realistic, professional, and up-to-date financial activity information. The virtual table data generated by the server (100) may be widely used for conducting mock tests on new financial products or for research to improve the performance of new business models. A detailed explanation of the method for generating virtual table data performed by the server (100) will be provided below.

[0052] Meanwhile, the user terminal (200) may refer to a terminal that is used by a user (e.g., an administrator or a worker) and is capable of running an application in a wired or wireless communication environment. The user terminal (200) may include various types of electronic devices, such as a personal computer (PC), a laptop, a tablet, a mobile phone, a smartphone, or a wearable device (e.g., a watch-type terminal). Additionally, the user terminal (200) may not refer to a single specific terminal, but may be used as a general term for various electronic devices used by the user.

[0053] The communication network (400) performs the role of connecting the server (100) and the user terminal (200). That is, the communication network (400) refers to a communication network that provides a connection path so that the user terminal (200) can transmit and receive data after connecting to the server (100). The communication network (400) may include wired networks such as LANs (Local Area Networks), WANs (Wide Area Networks), MANs (Metropolitan Area Networks), and ISDNs (Integrated Service Digital Networks), or wireless networks such as wireless LANs, CDMA, Bluetooth, and satellite communication, but the scope of the present invention is not limited thereto.

[0055] Meanwhile, in some embodiments of the present invention, the server (100) may operate together with the deep learning server (300) to generate the aforementioned virtual table data. In this case, the deep learning server (300) may be a device (or server) that professionally operates a deep learning model including a neural network trained to generate virtual table data. According to some embodiments, the deep learning server (300) may use a transformer model based on an encoder-decoder structure. For example, the deep learning server (300) may use an open-source artificial intelligence model such as ChatGPT.

[0056] Here, the server (100) and the deep learning server (300) may each be configured as separate devices or may be configured to be included within the same hardware system. At this time, the server (100) or the deep learning server (300) may each be implemented as at least one of a workstation, a data center, an internet data center (IDC), a direct attached storage (DAS) system, a storage area network (SAN) system, a network attached storage (NAS) system, a redundant array of inexpensive disks, or a redundant array of independent disks (RAID) system, and an Electronic Document Management (EDMS) system, but the present embodiment is not limited thereto.

[0057] That is, the virtual table data generation method according to some embodiments of the present invention may be performed only on the server (100), or may be performed together on the server (100) and the deep learning server (300).

[0058] Additionally, the managing entities of the server (100) and the deep learning server (300) may be the same or different. For example, a specific financial institution may manage and operate the server (100) and the deep learning server (300) together. As another example, the managing entities of the server (100) and the deep learning server (300) may be different, in which case the deep learning server (300) may refer to an external server such as a ChatGPT server.

[0059] However, for the sake of convenience of explanation, the following description will be given as an example in which the entity performing the virtual table data generation method according to some embodiments of the present invention is a server (100), and the deep learning model used in the present invention uses a module included in the server (100). Below, with reference to FIG. 2, each component and module of the server (100) will be described.

[0061] FIG. 2 is a block diagram illustrating components for a server according to some embodiments of the present invention.

[0062] Referring to FIG. 2, a server (100) according to some embodiments of the present invention includes an interface (110), a database (120), a processor (130), and a memory (140).

[0063] At this time, a Prompt Generating Module (hereinafter PGM), a Deep-learning Method Module (hereinafter DM), a Calibration Module (hereinafter CM), a Condition Generating Module (hereinafter CGM), a Data Verification Module (hereinafter VM), or a Data Correction Module (hereinafter DCM) may be loaded into the memory (140) and driven (or executed) by the processor (130). According to an embodiment, the Data Verification Module (VM) may include a Diversity Verification Model (hereinafter DVM) or a Condition Verification Model (hereinafter CVM).

[0064] Each module may be stored in the form of a computer program in a database (120) or storage (not shown) included in the server (100) and used. Additionally, in some embodiments of the present invention, some of the aforementioned modules may be omitted.

[0065] Specifically, the interface (110) can transmit data received by the server (100) from a user terminal (200) or a deep learning server (300) to other components within the server (100). The interface (110) may be provided in the server (100) and connected to an input / output device for receiving user input. Additionally, the interface (110) may include various communication modules and can perform data exchange between the user terminal (200) or the deep learning server (300) through a communication network (400).

[0066] The database (120) performs the function of storing and managing data received through the interface (110). Multiple modules that are loaded into memory (140) and available for use may be stored in the database (120) in the form of programs. In addition, the database (120) may store and manage not only training data for training a deep learning model, but also weights of a neural network composed of multiple layers (i.e., neural network weights of a pre-trained deep learning model). However, this is merely an example, and the present invention is not limited thereto.

[0067] The processor (130) can execute software to control at least one other component (e.g., hardware or software component) of the server (100) and perform various data processing and operations. For example, the processor (130) can load information, commands, or data received from another component (e.g., database (120)) into memory (140), perform operations using the loaded information, commands, or data, and store the resulting data in the database (120) or storage (not shown).

[0068] The memory (140) can load and store various data used by at least one component of the server (100) (e.g., processor (130)). For example, the data may include input data or output data for software and related commands.

[0069] Accordingly, the processor (130) can load and use modules or instructions related to various operations of a virtual table data generation method according to some embodiment of the present invention on memory (140).

[0071] Specifically, the processor (130) can generate a prompt for generating virtual table data desired by the user by using a prompt generation module (PGM) loaded into memory (140). Specifically, the prompt generation module (PGM) can generate a first prompt for generating a table schema for virtual table data, a second prompt including an adjusted table schema, and a third prompt including an adjusted table schema and table condition data.

[0072] For example, the first prompt may include a query for generating a table schema (e.g., a list and description of attributes of card payment history) and detailed conditions therefor. The second prompt may include a query for generating table data based on the adjusted table schema (e.g., generate data for card payment history, but please adhere to the provided table schema). The third prompt may include a query for generating table data based on the adjusted table schema and conditions for specific columns of the table schema (e.g., generate data for card payment history, but please adhere to the provided table schema and observe unary and binary conditions). However, this is merely an example, and the present invention is not limited thereto.

[0073] The first to third prompts generated by the prompt generation module (PGM) can be sequentially transmitted to the deep learning module (DM). However, the present invention is not limited thereto, and it is understood that the configuration and transmission order of each prompt can be implemented in various modified ways.

[0074] Next, the processor (130) can generate a table schema or table data based on a prompt generated by a prompt generation module (PGM) using a deep learning module (DM) loaded into memory (140). At this time, various types of neural network structures may be used for the deep learning module (DM). For example, the deep learning module (DM) may be implemented as a neural network model such as a Large Language Model (LM) or a Pretrained Language Model (PLM). However, this is merely one example, and the present invention is not limited thereto.

[0075] Additionally, weights for a neural network that has been pre-trained and derived may be applied to and utilized in the deep learning module (DM). As described above, the weights of the deep learning module (DM) or the neural network included in the deep learning module (DM) may be stored in the database (120) and utilized. In addition, the table schema or table data output from the deep learning module (DM) may also be stored in the database (120) and utilized and managed.

[0076] Meanwhile, the processor (130) can calibrate the table schema generated based on the first prompt in the deep learning module (DM) using the calibration module (CM) loaded in memory (140). At this time, the calibration module (CM) can calibrate the table schema output from the deep learning module (DM) using a reference table schema that is pre-stored in the database (120) or received from the user terminal (200). For convenience of explanation, the following description will use an example where the calibration module (CM) calibrates the table schema using a reference table schema pre-stored in the database (120). The specific calibration operation of the calibration module (CM) will be described later.

[0077] Additionally, the processor (130) can extract table condition data applicable to table data generated based on a second prompt in the deep learning module (DM) by using a condition generation module (CGM) loaded in memory (140). At this time, the condition generation module (CGM) can generate table condition data by comparing condition data received from the user terminal (200) or pre-stored in the database (120) with table data generated based on the second prompt. For convenience of explanation, the following description will be given as an example in which the condition generation module (CGM) generates table condition data using condition data pre-stored in the database (120).

[0078] The 'table condition data' output from the condition generation module (CGM) may include unary constraints or binary constraints for each column included in the table data.

[0079] Here, a 'unary condition' refers to a condition that takes one of a set of predetermined values. For example, a unary condition may be a condition that causes cells in a specific column to take binary values ​​such as 0 or 1, Yes or No, or Approve or Reject. As another example, a unary condition may be a condition that causes one of three predetermined options to take one of a value.

[0080] Furthermore, a 'binary condition' refers to a condition that includes an operation expression with another column. For example, the value of the first column may be determined by an operation between the value of the second column and the value of the second column (e.g., arithmetic operations or an operation using a predetermined function). However, this is merely a few examples of unary or binary conditions, and the present invention is not limited thereto.

[0081] Additionally, the processor (130) can perform data evaluation on table data generated based on a third prompt in the deep learning module (DM) by using a data verification module (VM) loaded in memory (140). At this time, the data verification module (VM) can perform verification of diversity for example data (V1 to Vn in FIG. 4) included in the table data, or verify whether constraint satisfaction is achieved for columns to which unary or binomial conditions are assigned within the example data. A detailed description of the operation of the data verification module (VM) will be provided below with reference to FIG. 10 to 15.

[0082] At this time, in one embodiment of the present invention, the processor (130) transmits the data evaluation result output from the data verification module (VM) to the prompt generation module (PGM), and the prompt generation module (PGM) can correct the third prompt based on the data evaluation result. Subsequently, the processor (130) can apply the corrected third prompt to the deep learning module (DM) and repeat the operation of regenerating table data.

[0083] Meanwhile, in another embodiment of the present invention, the processor (130) can perform correction on data included in table data generated based on a third prompt in the deep learning module (DM) by using a data correction module (DCM) loaded in memory (140). At this time, the data correction module (DCM) can perform an operation to correct data that does not meet a predetermined standard value among the table data to a value that meets the standard value, and such correction can be performed automatically according to a predetermined rule.

[0084] A detailed description of the process of verifying and correcting table data generated based on the third prompt in the processor (130) will be described in detail below with reference to FIGS. 12 to 15.

[0085] Below, we will examine in detail the operation of a virtual table data generation method according to several embodiments of the present invention.

[0087] FIG. 3 is a flowchart illustrating a method for generating virtual table data according to some embodiments of the present invention. FIG. 4 is a diagram illustrating an example of a table data generation process performed at each step of FIG. 3. For convenience of explanation, the method for generating virtual table data according to the present invention will be described below using the example of a server (100) or a processor (130) as the entity performing the method.

[0088] Referring to FIGS. 3 and 4, in a method for generating virtual table data according to some embodiments of the present invention, a server (100) generates a first prompt for generating a table schema (S110). Here, the first prompt may include a query for generating a table schema (e.g., a list and description of attributes of card payment history).

[0089] Next, the server (100) adjusts the table schema by comparing the table schema generated based on the first prompt with a predefined standard table schema (S120). For example, of FIG. 4 <a1>In this case, S represents a table schema generated with the first prompt as input, and of FIG. 4. <a2>In this, S' represents the adjusted table schema.

[0090] At this time, if the table schema generated based on the first prompt does not include an item of each column included in a predefined standard table schema, the server (100) may adjust the table schema to include such an item. A detailed explanation of this will be provided later with reference to FIGS. 5 and FIGS. 6.

[0091] Next, the server (100) generates a second prompt by referring to the adjusted table schema (S210). Here, the second prompt may include a query to generate table data based on the adjusted table schema (e.g., generate data for card payment history, but please comply with the provided table schema).

[0092] Next, the server (100) generates table condition data for the first table data generated based on the second prompt (S220). For example, of FIG. 4 <a2>represents first table data including an adjusted table schema (S') and example data (V1) for the same. FIG. 4 <a3>Columns to which table condition data (e.g., unary condition (Cu1) or binomial condition (Cb1~Cb3)) applied to the adjusted table schema (S') are disclosed.

[0093] Next, the server (100) generates a third prompt by referring to the table condition data and the adjusted table schema (S310). Here, the third prompt may include a query to generate table data based on the adjusted table schema and a condition on a specific column of the table schema (i.e., table condition data) (e.g., generate data for card payment history, but comply with the provided table schema and comply with unary and binary conditions).

[0094] Next, the server (100) derives the final table data through a verification operation on the second table data generated based on the third prompt (S320). For example, of FIG. 4 <a3>represents second table data including a modified table schema (S') and multiple example data (V1~Vn) to which table condition data (e.g., unary condition (Cu1) or binomial condition (Cb1~Cb3)) is applied.

[0095] In the present invention, the server (100) can generate high-accuracy table data required by the user by automatically updating the prompt for generating virtual table data multiple times while performing each of the aforementioned steps sequentially.

[0096] Through this, the present invention can increase the work efficiency of developers or workers and optimize resources required for the development of new services or the analysis of existing services by omitting or automating the data collection process necessary for developing new services.

[0097] Below, we will examine in detail a method for adjusting a table schema according to some embodiments of the present invention (i.e., step S100 of FIG. 3).

[0099] FIG. 5 is a flowchart for explaining step S100 of FIG. 3. FIG. 6 is a block diagram for explaining the flowchart of FIG. 5. FIG. 7 is a diagram for explaining a neural network model applied to the deep learning module of FIG. 6.

[0100] Referring to FIGS. 5 and 6, the server (100) may receive a data creation request from a user terminal (200) that includes information about table data desired by the user (S111). Here, the user terminal (200) may receive a data creation request from the user that includes information about the type of data desired by the user (e.g., financial data), format (e.g., table data), or conditions (e.g., including amount and point columns), and may transmit the received data creation request to the server (100).

[0101] Next, the processor (130) of the server (100) inputs the received data generation request into the prompt generation module (PGM), and receives a first prompt generated based on the information included in the data generation request as the output of the prompt generation module (PGM) (S113).

[0102] Next, the processor (130) applies a first prompt generated to the deep learning module (DM) and receives a table schema as the output of the deep learning module (DM) (S121).

[0103] At this time, the deep learning module (DM) used in some embodiments of the present invention can output a table schema or table data corresponding to an input prompt using an artificial neural network that has been pre-trained based on big data. Here, table data refers to data including a table schema and example data generated in accordance with the format of the table schema.

[0104] Specifically, the deep learning module (DM) is implemented as a neural network structure and can train an artificial neural network using mapping values ​​for separate parameters derived based on input data. At this time, the deep learning module (DM) can perform machine learning on the parameters input as learning factors.

[0105] To explain in more detail, deep learning technology, a type of machine learning, learns by going down to deep levels in multiple stages based on data.

[0106] Deep learning refers to a set of machine learning algorithms that extract key data from multiple data by increasing the level.

[0107] The deep learning module (DM) can utilize various known artificial neural network structures. For example, the deep learning module (DM) can utilize structures such as CNN (Convolutional Neural Network), RNN (Recurrent Neural Network), DBN (Deep Belief Network), GNN (Graph Neural Network), and transformer.

[0108] Meanwhile, the artificial neural network training of the deep learning module (DM) can be achieved by adjusting the weights of the connections between nodes (and adjusting the bias values ​​if necessary) so that a desired output is produced for a given input. Additionally, the artificial neural network can continuously update its weight values ​​through learning. Furthermore, methods such as backpropagation can be used for the training of the artificial neural network.

[0109] At this time, an artificial neural network that has been pre-trained by machine learning may be installed in the memory (140) of the server (100). That is, data and result data used in machine learning may be stored in the memory (140).

[0110] In some embodiments of the present invention, the deep learning module (DM) may include a deep learning model based on a Transformer or a Large Language Model, which includes an encoder (EN) and a decoder (DE). However, this is merely one example of the present invention and the present invention is not limited thereto.

[0111] Referring to FIG. 7, the deep learning module (DM) includes an input layer with a text-based prompt as an input node, an output layer with virtual table data as an output node, and M hidden layers placed between the input layer and the output layer.

[0112] Here, weights can be set for the edges connecting the nodes of each layer. The presence or absence of such weights or edges can be added, removed, or updated during the learning process. Therefore, through the learning process, the weights of the nodes and edges placed between k input nodes and i output nodes can be updated.

[0113] Before the deep learning module (DM) performs training, all nodes and edges can be set to initial values. However, when information is input cumulatively, the weights of the nodes and edges are changed, and in this process, a matching can be made between the parameters input as learning factors (i.e., character recognition data) and the values ​​assigned to the output nodes (i.e., summary data).

[0114] Additionally, when using a cloud server (not shown) (e.g., deep learning server (300)), the deep learning module (DM) can receive and process a large number of parameters. In this case, in some embodiments of the present invention, the operation of the deep learning module (DM) may be performed in conjunction with the server (100) and / or a separate cloud server (not shown). Thus, the deep learning module (DM) can perform learning based on vast amounts of data.

[0115] The weights of nodes and edges between the input and output nodes constituting the deep learning module (DM) can be updated through the learning process of the deep learning module (DM). In addition, the parameters output from the deep learning module (DM) can be further extended with various data in addition to summary data.

[0116] Both semi-supervised learning and supervised learning can be used as machine learning methods in the deep learning module (DM). Additionally, the deep learning module (DM) can be controlled to automatically update the artificial neural network structure to output more accurate table data after learning, depending on the settings.

[0118] Referring again to FIGS. 5 and 6, the processor (130) compares the table schema output from the deep learning module (DM) with a reference table schema stored in advance in the database (120) using the adjustment module (CM) (S123). Here, the reference table schema may define items that must be included in the table data. Of course, the reference table schema may be defined in advance by an expert and stored in the database (120) for use, or it may be received separately from the user terminal (200) for use.

[0119] At this time, if the table schema received from the deep learning module (DM) does not include an item included in the reference table schema, the adjustment module (CM) can adjust the table schema so that the item is included.

[0120] For example, the table schema (S) generated in the deep learning module (DM) of FIG. 4 <a1>As disclosed in the example, if the standard table schema is composed only of Date, Amount, and Merchant items, and the standard table schema predefined in the database (120) is composed of Date, Amount, Merchant, Inquiry Type, Online Pay, Regular Pay, Ratio of Point, and Point, the adjustment module (CM) can adjust the adjusted table schema (S') to include all items of the standard table schema. However, this is merely an example and the present invention is not limited thereto.

[0121] Below, we will examine in detail a method for generating table condition data according to some embodiments of the present invention (i.e., step S200 of FIG. 3).

[0123] FIG. 8 is a flowchart for explaining step S200 of FIG. 3. FIG. 9 is a block diagram for explaining the flowchart of FIG. 8.

[0124] Referring to FIGS. 8 and 9, the processor (130) generates a second prompt using a prompt generation module (PGM) by referring to the table schema adjusted in step S100 (S211). At this time, although not clearly illustrated in the drawings, the prompt generation module (PGM) may generate the second prompt by referring together to the adjusted table schema and a data generation request received from the user terminal (200). However, this is merely an example and the present invention is not limited thereto.

[0125] Next, the processor (130) applies a second prompt to the deep learning module (DM) and receives first table data as the output of the deep learning module (DM) (S221). Here, the first table data may include first example data (V1) generated according to the format of the adjusted table schema (S') (Fig. 4 <a2>).

[0126] Next, the processor (130) generates table condition data by using a condition generation module (CGM) to compare each column included in the first table data with predefined condition data (S223). At this time, the table condition data includes a unary constraint and / or a binary constraint for each column included in the table data.

[0127] At this time, the condition generation module (CGM) may use condition data stored in advance in the database (120) or condition data received separately from the user terminal (200). Here, the condition data may be data stored in a table format containing multiple items (or columns) and whether or not a condition (i.e., a unary condition or a binary condition) is applied to each item. The condition generation module (CGM) may generate table condition data by comparing the condition data with each column included in the first table data and determining which condition to apply to which column of the first table data.

[0128] For example, the table condition data may include a unary condition that causes the Inquiry Type column to have either an online or offline value. Additionally, the table condition data may include a binomial condition that causes the Amount, Ratio of Point, and Point columns to be cross-referenced by an expression (Fig. 4). <a3>(Reference). However, this is merely one example and the present invention is not limited thereto.

[0129] Below, we will examine in detail a method for generating second table data and confirming final table data through a verification operation according to some embodiments of the present invention (i.e., step S300 of FIG. 3).

[0131] FIG. 10 is a flowchart for explaining step S300 of FIG. 3. FIG. 11 is a block diagram for explaining the flowchart of FIG. 10.

[0132] Referring to FIGS. 10 and 11, the processor (130) generates a third prompt using a prompt generation module (PGM) by referring to the table schema adjusted in step S100 and the table condition data derived in step S200 (S311). At this time, although not clearly illustrated in the drawings, the prompt generation module (PGM) may generate the third prompt by referring together to the adjusted table schema, the table condition data, and a data generation request received from the user terminal (200). However, this is merely an example and the present invention is not limited thereto.

[0133] Next, the processor (130) applies a third prompt to the deep learning module (DM) and receives second table data as the output of the deep learning module (DM) (S321). Here, the second table data may include a plurality of example data (V1 to Vn) generated according to the format of the adjusted table schema (S') (Fig. 4 <a3>). At this time, the second table data (of FIG. 4 <a3>) is the first table data (of FIG. 4 <a2>It can include a larger number of example data than ).

[0134] Next, the processor (130) applies the second table data to the data verification module (VM) and obtains a data evaluation result as an output to the data verification module (VM) (S323). At this time, the data verification module (VM) may perform a diversity verification on the example data included in the second table data and perform a verification of whether the condition is satisfied for a column to which a unary condition or a binomial condition is assigned within the example data. Accordingly, the data evaluation result may include a first evaluation value for diversity, a second evaluation value for whether the unary condition is satisfied, and / or a third evaluation value for whether the binomial condition is satisfied.

[0135] Next, the processor (130) determines the final table data based on the acquired data evaluation results (S325).

[0136] In the following, the process of determining final table data based on data evaluation results in some embodiments of the present invention will be described in detail.

[0138] FIG. 12 is a flowchart for illustrating an embodiment of steps S323 and S325 of FIG. 10. FIG. 13 is a block diagram for illustrating the flowchart of FIG. 12.

[0139] Referring to FIGS. 12 and 13, first, the processor (130) transmits the second table data obtained from the deep learning module (DM) to the data verification module (VM).

[0140] At this time, the data validation module (VM) derives a first evaluation value for the diversity of each row data or each column data included in the second table data using a diversity validation model (DVM) (S323a).

[0141] Specifically, in one embodiment, the data validation module (VM) can derive the uniformity of each row of data included in the second table data by concatenating all values ​​(e.g., using a concat function), embedding the concatenated values ​​into a specific domain, extracting the uniformity of the embedded values, and normalizing them using a normalization function (e.g., a Sigmoid function). Through this, the data validation module (VM) can derive the diversity between instances included in each row of data as the first evaluation value.

[0142] Meanwhile, in another embodiment, the data validation module (VM) can derive the diversity of each column data included in the second table data (i.e., diversity between attributes within an instance) as a first evaluation value. Here, the first evaluation value can be derived by the following <Equation 1>.

[0143]

[0144] Here, represents a first evaluation value indicating the degree of diversity of the data, and I represents a set of unique values ​​corresponding to a specific column, and i represents a value included in that set, represents the probability that each i appears.

[0146] In addition, the data validation module (VM) uses a condition validation model (CVM) to derive a second evaluation value regarding whether each column data included in the second table data satisfies a unary condition (S323b). Here, the second evaluation value can be derived by the following <Equation 2>.

[0147]

[0148] Here, represents a second evaluation value regarding whether a unary condition is satisfied, and represents the number of unary condition items, and represents the set of values ​​to which a unary condition applies, and represents a unary operator that checks for the existence of a predetermined unary item, and It refers to each value to which the unary condition applies.

[0150] In addition, the data validation module (VM) uses a condition validation model (CVM) to derive a third evaluation value regarding whether each column data included in the second table data satisfies a binomial condition (S323c). Here, the third evaluation value can be derived by the following <Equation 3>.

[0151]

[0152] Here, represents a third evaluation value regarding whether the binomial condition is satisfied, and represents the number of binomial condition items, and represents the set of values ​​to which the binary condition applies, and represents each value to which the binomial condition applies.

[0154] Next, the data verification module (VM) determines whether the criteria for the first to third evaluation values ​​are satisfied and generates a data evaluation result including the result (S323d). At this time, the data verification module (VM) determines whether the first to third evaluation values ​​and each evaluation value satisfy a predetermined criteria (or threshold value) and generates a data evaluation result including the result.

[0155] At this time, the data verification module (VM) may perform the aforementioned steps S323a through S323c sequentially or in parallel, and some steps may be omitted or modified. Furthermore, the data verification module (VM) may perform step S323d based on the evaluation values ​​derived from the aforementioned steps.

[0157] Next, the processor (130) checks whether the first to third evaluation values ​​satisfy a predetermined threshold by referring to the data evaluation results (S323e). At this time, the processor (130) can determine whether the threshold is satisfied based on whether each of the first to third evaluation values ​​satisfies all of the predetermined thresholds, or whether more than a predetermined number of evaluation values ​​among the first to third evaluation values ​​satisfy the thresholds.

[0158] If the first to third evaluation values ​​do not meet the criteria, the processor (130) may perform an operation to regenerate the second table data. At this time, the processor (130) may regenerate all or part of the second table data.

[0159] When regenerating only a portion of the second table data, the processor (130) can store only the data that passed the threshold in the database (120), discard the remaining data, and then regenerate the second table data.

[0160] Specifically, the processor (130) transmits the data evaluation result received from the data verification module (VM) to the prompt generation module (PGM), and the prompt generation module (PGM) corrects the third prompt based on the data evaluation result (S324a). At this time, the prompt generation module (PGM) may correct the third prompt to include a conditional statement regarding the replacement or exclusion of data that does not meet the criteria. However, this is merely an example and the present invention is not limited thereto.

[0161] Next, the processor (130) applies the corrected third prompt again to the deep learning module (DM) to regenerate the second table data (S324b).

[0162] Next, the processor (130) re-applies the regenerated second table data to the data verification module (VM) and re-obtains the data evaluation result as an output to the data verification module (VM). At this time, the operation of the data verification module (VM) may be substantially the same as the steps S323a to S323d described above.

[0163] Next, the processor (130) checks whether the first to third evaluation values ​​satisfy a predetermined standard value by referring to the re-acquired data evaluation result.

[0165] On the other hand, if the first to third evaluation values ​​satisfy a predetermined standard, the processor (130) can finalize the second table data as final table data (S325a), and can store the finalized table data in the database (120) (S325b).

[0166] In addition, it goes without saying that the operation of generating the aforementioned second table data can be performed repeatedly until the final table data is finalized.

[0168] FIG. 14 is a flowchart illustrating another embodiment of step S325 of FIG. 10. FIG. 15 is a block diagram illustrating the flowchart of FIG. 14. In the following, details that overlap with the previously described content will be omitted, and the differences will be explained in detail.

[0169] Referring to FIGS. 14 and FIGS. 15, in another embodiment of the present invention, the processor (130) performs steps S323a to S323d using a data verification module (VM).

[0170] Next, the processor (130) transmits the data evaluation result received from the data verification module (VM) to the data correction module (DCM).

[0171] At this time, the data correction module (DCM) performs correction on data that does not meet the threshold included in the second table data (S424). Specifically, the data correction module (DCM) corrects values ​​that do not meet the threshold to values ​​that meet the threshold. At this time, the data correction module (DCM) can correct values ​​that do not meet the threshold to values ​​that meet the threshold according to a predetermined rule.

[0172] Next, the processor (130) can confirm the second table data, which reflects the corrections performed by the data correction module (DCM), as the final table data (S425a). The confirmed final table data can be stored in the database (120) (S425b).

[0173] Therefore, virtual table data generated through the aforementioned method can possess accuracy and diversity higher than predetermined thresholds, and through data verification and correction processes, virtual data reflecting the characteristics of the customer's sensitive information data can be generated. Through this, the present invention solves security issues regarding sensitive information data while simultaneously generating data in a form desired by the user and utilizing it for operating business models.

[0174] In summary, some embodiments of the present invention can automatically generate highly accurate virtual data capable of replacing data containing sensitive customer information, and can generate virtual data to fit a table schema of the form desired by the user. Through this, developers or operators can use the present invention to test or train business models applied to new services and can immediately generate and utilize highly accurate virtual data.

[0176] FIG. 16 is a diagram illustrating a hardware implementation of a device or system for performing a virtual table data generation method according to some embodiments of the present invention.

[0177] Referring to FIG. 16, a server (100) performing a virtual table data generation method according to some embodiments of the present invention may be implemented as an electronic device (1000). The electronic device (1000) may include a processor (1010), an input / output device (1020, I / O), a memory (1030, memory), an interface (1040, interface), a storage (1050, storage), and a bus (1060, bus). The processor (1010), the input / output device (1020), the memory (1030), the interface (1040), and / or the storage (1050) may be connected to each other through the bus (1060). The bus (1060) corresponds to a path through which data is moved.

[0178] Specifically, the processor (1010) may include at least one of a CPU (Central Processing Unit), MPU (Micro Processor Unit), MCU (Micro Controller Unit), GPU (Graphic Processing Unit), microprocessor, digital signal processor, microcontroller, application processor (AP), and logic elements capable of performing similar functions.

[0179] The input / output device (1020) may include at least one of a keypad, a keyboard, a touchscreen, and a display device.

[0180] The memory (1030) can load data and / or programs, etc. At this time, the memory (1030) is an operational memory for enhancing the operation of the processor (1010) and may include high-speed DRAM and / or SRAM, etc. The memory (1030) may include one or more volatile memory devices such as DDR SDRAM (Double Data Rate Static DRAM) and SDR SDRAM (Single Data Rate SDRAM) and / or one or more non-volatile memory devices such as EEPROM (Electrical Erasable Programmable ROM) and flash memory.

[0181] The interface (1040) can perform the function of transmitting data to a communication network or receiving data from a communication network. The interface (1040) may be in a wired or wireless form. For example, the interface (1040) may include an antenna or a wired / wireless transceiver.

[0182] Storage (1050) can store and retain data and / or programs, etc. Storage (1050) may include one or more non-volatile memory devices such as a solid-state drive (SSD), a hard drive, and a flash memory. In the present invention, storage (1050) can store a computer program consisting of instructions for performing a virtual table data generation method.

[0183] Alternatively, the server (100) and the deep learning server (300) according to embodiments of the present invention may each be a system formed by connecting a plurality of electronic devices (1000) to each other through a network. In such cases, each module or combination of modules may be implemented as an electronic device (1000). However, the present embodiment is not limited thereto.

[0184] Additionally, the server (100) may be implemented as at least one of a workstation, a data center, an internet data center (IDC), a direct attached storage (DAS) system, a storage area network (SAN) system, a network attached storage (NAS) system, and a redundant array of inexpensive disks, or a redundant array of independent disks (RAID) system, but the present embodiment is not limited thereto.

[0185] Additionally, the server (100) can transmit data through a network. The network may include a network based on wired internet technology, wireless internet technology, and local area communication technology. Wired internet technology may include, for example, at least one of a local area network (LAN) and a wide area network (WAN).

[0186] Wireless internet technology may include, for example, at least one of Wireless LAN (WLAN), DMNA (Digital Living Network Alliance), WiBro (Wireless Broadband), WiMAX (World Interoperability for Microwave Access), HSDPA (High Speed ​​Downlink Packet Access), HSUPA (High Speed ​​Uplink Packet Access), IEEE 802.16, Long Term Evolution (LTE), LTE-A (Long Term Evolution-Advanced), Wireless Mobile Broadband Service (WMBS), and 5G NR (New Radio) technology. However, the present embodiment is not limited thereto.

[0187] Short-range communication technology may include, for example, at least one of Bluetooth, RFID (Radio Frequency Identification), Infrared Data Association (IrDA), Ultra-Wideband (UWB), ZigBee, Near Field Communication (NFC), Ultra Sound Communication (USC), Visible Light Communication (VLC), Wi-Fi, Wi-Fi Direct, and 5G NR (New Radio). However, the present embodiment is not limited thereto.

[0188] A server (100) communicating through a network may comply with technical standards and standard communication methods for mobile communication. For example, a standard communication method may include at least one of GSM (Global System for Mobile communication), CDMA (Code Division Multi Access), CDMA2000 (Code Division Multi Access 2000), EV-DO (Enhanced Voice-Data Optimized or Enhanced Voice-Data Only), WCDMA (Wideband CDMA), HSDPA (High Speed ​​Downlink Packet Access), HSUPA (High Speed ​​Uplink Packet Access), LTE (Long Term Evolution), LTEA (Long Term Evolution-Advanced), and 5G NR (New Radio). However, the present embodiment is not limited thereto.

[0189] The above description is merely an illustrative explanation of the technical concept of the present embodiment, and a person skilled in the art to which the present embodiment belongs would be able to make various modifications and variations within the scope of the essential characteristics of the present embodiment. Accordingly, the present embodiments are intended to explain, not limit, the technical concept of the present embodiment, and the scope of the technical concept of the present embodiment is not limited by these embodiments. The scope of protection of the present embodiment shall be interpreted by the claims below, and all technical concepts within an equivalent scope shall be interpreted as being included within the scope of rights of the present embodiment.

Claims

Claim 1 A method for generating virtual table data performed on a server using a deep learning module, comprising: a step of generating a first prompt for generating a table schema; a step of comparing a table schema generated based on the first prompt with a predefined reference table schema, and if the table schema does not include an item included in the predefined reference table schema, a step of adjusting the table schema so that the item is included; a step of generating a second prompt by referencing the adjusted table schema; a step of generating table condition data for the first table data generated based on the second prompt; a step of generating a third prompt by referencing the table condition data and the adjusted table schema; and a step of deriving final table data through a verification operation on the second table data generated based on the third prompt. Claim 2 A method for generating virtual table data according to claim 1, wherein the step of adjusting the table schema comprises the step of applying the first prompt to the deep learning module and receiving the table schema as the output of the deep learning module. Claim 3 A method for generating virtual table data according to claim 1, wherein the step of generating the first prompt comprises: receiving a data generation request containing information about table data desired by a user from a user terminal linked to the server; and generating the first prompt based on the information included in the data generation request. Claim 4 A virtual table data generation method according to claim 3, wherein the step of generating the second prompt includes generating the second prompt by referring together with the adjusted table schema and the data generation request, and the step of generating the third prompt includes generating the third prompt by referring together with the table condition data, the adjusted table schema and the data generation request. Claim 5 A method for generating virtual table data according to claim 1, wherein the step of generating table condition data comprises: applying the second prompt to the deep learning module and receiving the first table data as the output of the deep learning module; and generating table condition data by comparing each column included in the first table data with predefined condition data. Claim 6 A method for generating virtual table data according to claim 5, wherein the table condition data includes a unary constraint or a binary constraint for each column included in the table data, wherein the unary constraint means a condition having any one of predetermined values, and the binary constraint means a condition including an operation expression with another column. Claim 7 A method for generating virtual table data according to claim 1, wherein the step of deriving the final table data comprises: applying the third prompt to the deep learning module and receiving the second table data as the output of the deep learning module; applying the second table data to the data verification module and obtaining a data evaluation result as the output to the data verification module; and determining the final table data based on the data evaluation result. Claim 8 A method for generating virtual table data according to claim 7, wherein the second table data includes a greater number of example data than the first table data, and the data verification module performs diversity verification on the example data included in the second table data and performs verification of whether a condition is satisfied for a column to which a unary condition or a binomial condition is assigned within the example data. Claim 9 A method for generating virtual table data according to claim 7, wherein the step of obtaining the data evaluation result comprises: a step of deriving a first evaluation value for the diversity of each row data or each column data included in the second table data; a step of deriving a second evaluation value for whether each column data included in the second table data satisfies a unary condition; a step of deriving a third evaluation value for whether each column data included in the second table data satisfies a binomial condition; and a step of determining whether the first to third evaluation values ​​satisfy a reference value and generating the data evaluation result including the result. Claim 10 A method for generating virtual table data according to claim 9, wherein the step of determining the final table data comprises: a step of correcting data included in the second table data that does not meet the standard value to a value that meets the standard value; and a step of determining the second table data with the corrections reflected as the final table data. Claim 11 A method for generating virtual table data according to claim 7, further comprising: a step of correcting the third prompt based on the data evaluation result; a step of applying the corrected third prompt to the deep learning module to regenerate the second table data; a step of applying the regenerated second table data to the data verification module and re-obtaining the data evaluation result as an output to the data verification module; and a step of determining the final table data based on the re-obtained data evaluation result. Claim 12 A server comprising: a processor; a memory for loading a computer program and a deep learning module executed by the processor; and a database for storing data generated during the execution of the computer program, wherein the computer program comprises: a step of generating a first prompt for generating a table schema; a step of comparing a table schema generated based on the first prompt with a predefined reference table schema, and if the table schema does not include an item included in the predefined reference table schema, a step of adjusting the table schema so that the item is included; a step of generating a second prompt by referencing the adjusted table schema; a step of generating table condition data for the first table data generated based on the second prompt; a step of generating a third prompt by referencing the table condition data and the adjusted table schema; a step of deriving final table data through a verification operation on the second table data generated based on the third prompt; and a step of storing the final table data in the database. Claim 13 In claim 12, the step of adjusting the table schema comprises: applying the first prompt to the deep learning module and receiving the table schema as the output of the deep learning module; comparing the table schema with the predefined reference table schema; and, if the table schema does not include an item included in the predefined reference table schema, adjusting the table schema so that the item is included; the step of generating the table condition data comprises: applying the second prompt to the deep learning module and receiving the first table data as the output of the deep learning module; and generating the table condition data by comparing each column included in the first table data with the predefined condition data; and the step of deriving the final table data comprises: applying the third prompt to the deep learning module and receiving the second table data as the output of the deep learning module; applying the second table data to the data verification module loaded in memory and obtaining a data evaluation result as the output to the data verification module. A server comprising the step of determining the final table data based on the above data evaluation results. Claim 14 In claim 13, the step of obtaining the data evaluation result comprises: a step of deriving a first evaluation value for the diversity of each row data or each column data included in the second table data; a step of deriving a second evaluation value for whether each column data included in the second table data satisfies a unary condition; a step of deriving a third evaluation value for whether each column data included in the second table data satisfies a binomial condition; and a step of determining whether the first to third evaluation values ​​satisfy a reference value and generating the data evaluation result including the result. Claim 15 A computer-readable recording medium having a program capable of executing a method described in any one of claims 1 to 11.

Images