Structured data creation assistance device, structured data creation assistance method, and structured data creation assistance system

Abstract

A structured data creation assistance device assists with creation of structured data by using a large language model which performs processing according to a prompt. The device includes: a prompt generation unit that generates a prompt including a structured data rule, an instruction sentence, and table data; an answer reception unit that causes the large language model to perform processing according to the prompt and that receives an answer generated by the large language model; and an output unit that, according to the answer, outputs a proposal for improving the table data.

Description

Structured Data Creation Support Device, Structured Data Creation Support Method, and Structured Data Creation Support System

[0001] The present disclosure relates to a structured data creation support device, a structured data creation support method, and a structured data creation support system.

[0002] In data analysis such as statistical analysis or machine learning, data recorded according to a predetermined rule is required. The data used in data analysis such as statistical analysis or machine learning is, for example, structured data. Structured data has a predetermined rule in the description content of data and table configuration. Also, in data analysis such as statistical analysis or machine learning, a template for recording data according to a predetermined rule may be created, and the data recorded using the template may be analyzed.

[0003] For example, Patent Document 1 discloses a technique for detecting an error in an account item in spreadsheet data and recommending an amendment for the error in the account item.

[0004] Japanese Unexamined Patent Application Publication No. 2020 - 95682

[0005] However, recording data according to a predetermined rule has sometimes been difficult for workers who are not familiar with programming. When data is not recorded according to a predetermined rule, the worker who analyzes the data manually converts it into structured data, which takes time and effort and may also cause correction errors. Also, creating or modifying a template has sometimes been difficult for workers who are not familiar with programming.

[0006] An object of the present disclosure is to provide a structured data creation support device, a structured data creation support method, and a structured data creation support system that support the creation of structured data according to a predetermined rule.

[0007] The present disclosure has the following configuration.

[0008] [1] A structured data creation support device that assists in the creation of structured data using a large-scale language model that processes according to prompts, comprising: a prompt generation unit that generates prompts including rules, instructions, and table data for the structured data; a response receiving unit that causes the large-scale language model to process according to the prompts and receives the response generated by the large-scale language model; and an output unit that outputs suggestions for improving the table data according to the response.

[0009] [2] The structured data creation support device according to [1], wherein the structured data rules include items for checking the data and table structure, the response generated by the large-scale language model includes information about the items for which the table data violates and the table data improved in accordance with the rules, and the output unit outputs the information about the items for which the table data violates and the table data improved in accordance with the rules as a suggestion for improving the table data.

[0010] [3] The structured data creation support device according to [2], wherein the information relating to the check items in which the table data violates includes information explaining improvements to the check items in which the table data violates.

[0011] [4] The structured data creation support device according to [1], wherein the structured data rules include items for checking data description rules in accordance with the operation of a data processing program, the response generated by the large-scale language model includes information about the items for which the table data violates and the table data improved in accordance with the rules, and the output unit outputs the information about the items for which the table data violates and the table data improved in accordance with the rules as a suggestion for improving the table data.

[0012] [5] The structured data creation support device according to any one of [1] to [4], wherein the prompt generation unit reads a set of structured data rules and instruction statements corresponding to the type of structured data, and generates the prompt including the structured data rules, instruction statements, and table data.

[0013] [6] A structured data creation support device according to any one of [1] to [5], further comprising a template creation unit that creates a template of the structured data in accordance with the above answer.

[0014] [7] A structured data creation support device that assists in the creation of structured data using a large-scale language model that processes according to prompts, the structured data creation support method comprising: a prompt generation step of generating prompts that include rules, instructions, and table data for the structured data; a response receiving step of causing the large-scale language model to process according to the prompts and receiving the response generated by the large-scale language model; and an output step of outputting suggestions for improving the table data according to the response.

[0015] [8] A structured data creation support system that supports the creation of structured data using a large-scale language model that processes according to prompts, comprising: a prompt generation unit that generates prompts including rules, instructions, and table data for the structured data; a large-scale language model that processes according to the prompts and generates answers to the prompts; an answer receiving unit that receives the answers generated by the large-scale language model; and an output unit that outputs suggestions for improving the table data according to the answers.

[0016] According to this disclosure, it is possible to assist in the creation of structured data in accordance with prescribed rules.

[0017] This is a diagram illustrating an example of a structured data creation support system according to this embodiment. This is a diagram illustrating a hardware configuration of an example of a computer according to this embodiment. This is a diagram illustrating a functional configuration of an example of a structured data creation support device according to this embodiment. This is a flowchart of an example of a structured data creation support system according to this embodiment. This is an explanatory diagram of an example of a prompt generated in step S14. This is an image diagram of an example of an improvement suggestion output in step S18. This is an image diagram of an example of a screen that displays the created template. This is an image diagram of an example of an improvement suggestion output in step S18. This is an image diagram of an example of a screen that displays the created template. This is a diagram illustrating an example of the normal operation of a data processing program. This is a diagram illustrating an example of application to data description rules in accordance with the operation of a data processing program.

[0018] Next, embodiments of the present invention will be described in detail. However, the present invention is not limited to the following embodiments.

[0019] <System Configuration> Figure 1 is a configuration diagram of an example of a structured data creation support system 1 according to this embodiment. The structured data creation support system 1 in Figure 1 supports the creation of structured data using a large-scale language model 16 that processes according to prompts. Structured data is data recorded in accordance with predetermined rules. Structured data is, for example, data arranged in rows and columns, and is data that has information about the items in the rows and columns.

[0020] Furthermore, structured data is data recorded according to predetermined rules for the purpose of use in data analysis such as statistical analysis or machine learning, and is, for example, data that is represented in a machine-readable format using applications such as data analysis tools. Machine-readable data is data whose content can be easily extracted or processed using a computer without human intervention, without losing its meaning.

[0021] Data that is presented in a machine-readable format is sometimes referred to as machine-readable data, machine-readable data, or machine-readable data. In Japan, the Ministry of Internal Affairs and Communications established "Methods for Presenting Machine-Readable Data in Statistical Tables" in 2020 as a unified rule for creating statistical tables for government statistics.

[0022] Data that is not recorded according to prescribed rules is data that is not machine-readable by applications such as data analysis tools, and is sometimes referred to as unstructured data.

[0023] The structured data creation support system 1 shown in Figure 1 comprises a worker terminal 10, a structured data creation support device 12, and a response device 14, all of which are connected via a network N to enable data communication. The network N is, for example, a local area network (LAN) or the internet.

[0024] The worker terminal 10 is an information processing device such as a PC (Personal Computer), tablet terminal, or smartphone operated by the worker. The worker can operate the worker terminal 10 and utilize the various functions provided by the structured data creation support device 12. The structured data creation support device 12 is an information processing device such as a workstation or PC. The worker may also operate the structured data creation support device 12 and utilize the various functions provided by the structured data creation support device 12. The response device 14 is an information processing device such as a workstation or PC. The response device 14 has a large-scale language model (LLM) that processes according to prompts.

[0025] The structured data creation support device 12 creates prompts as described later, and causes the large-scale language model 16 of the response device 14 to generate answers to those prompts. The large-scale language model 16 is a natural language processing model that has been trained using a large amount of text data. The large-scale language model 16 can utilize, for example, BERT (Bidirectional Encoder Representations from Transformers) or GPT (Generative Pre-Trained Transformer).

[0026] The structured data creation support device 12 generates prompts containing structured data rules, instructions, and tabular data, and causes the large-scale language model 16 to generate a response according to the generated prompts. The prompts generated by the structured data creation support device 12 cause the large-scale language model 16 to generate a response that outputs suggestions for improving the tabular data contained in the prompts. The structured data creation support device 12 outputs suggestions for improving the tabular data according to the response generated by the large-scale language model 16.

[0027] Note that the configuration of the structured data creation support system 1 shown in Figure 1 is just one example, and can vary depending on the application and purpose. For example, the structured data creation support device 12 or the response device 14 may be implemented using multiple computers, or it may be implemented as a cloud service.

[0028] Furthermore, the structured data creation support system 1 shown in Figure 1 may be implemented using a single computer. Also, the structured data creation support system 1 shown in Figure 1 may be implemented using a single computer for both the worker terminal 10 and the structured data creation support device 12. Furthermore, the structured data creation support system 1 shown in Figure 1 may be implemented using a single computer for both the structured data creation support device 12 and the response device 14.

[0029] <Hardware Configuration> The worker terminal 10, structured data creation support device 12, and response device 14 in Figure 1 can be realized by a computer 500 with the hardware configuration shown in Figure 2.

[0030] Figure 2 is a hardware configuration diagram of an example of a computer 500 according to this embodiment. The computer 500 includes, for example, an input device 501, an output device 502, an external interface 503, RAM (Random Access Memory) 504, ROM (Read Only Memory) 505, a CPU (Central Processing Unit) 506, a communication interface 507, and an auxiliary storage device 508, and each of these is interconnected via bus B. Note that the input device 501 and the output device 502 may be used in a connected configuration.

[0031] The input device 501 is a device that accepts user input, such as a touch panel, operation keys, buttons, keyboard, or mouse. The output device 502 has a device that displays a screen and a device that outputs sound. The device that displays a screen is, for example, a display such as an LCD. The device that outputs sound is, for example, a speaker. The communication I / F 507 is an interface for the computer 500 to perform data communication.

[0032] The auxiliary storage device 508 is an example of a non-volatile storage device that stores programs and data. The auxiliary storage device 508 is, for example, an HDD (Hard Disk Drive) or an SSD (Solid State Drive). The programs are, for example, the operating system (OS), which is basic software, and applications that provide various functions on the OS. The CPU 506 is an example of a processor and may have a device such as a GPU (Graphics Processing Unit).

[0033] The external interface 503 is an interface to an external device. The external device is a recording medium 503a, etc. The computer 500 can read programs and data from the recording medium 503a via the external interface 503. The recording medium 503a is, for example, a flexible disk, CD, DVD, SD memory card, or USB memory.

[0034] ROM 505 stores the BIOS, OS settings, and network settings that are executed when the computer 500 starts up. RAM 504 is working memory for temporarily storing programs and data. The CPU 506 can read programs and data from ROM 505 or auxiliary storage device 508 into RAM 504 and execute processing to realize the various functions described later.

[0035] <Functional Configuration> An example of the functional configuration of the structured data creation support device 12 according to this embodiment will be described. Figure 3 is a functional configuration diagram of an example of the structured data creation support device 12 according to this embodiment. In the configuration diagram of Figure 3, parts that are unnecessary for the explanation of this embodiment have been appropriately omitted.

[0036] The structured data creation support device 12 functions as a rule set acquisition unit 30, a table data acquisition unit 32, a prompt generation unit 34, a prompt transmission unit 36, a response reception unit 38, an output unit 40, a template creation unit 42, a rule set storage unit 46, and a table data storage unit 48, when a computer 500 executes a program such as an application.

[0037] The rule set acquisition unit 30 reads and acquires the rule set stored in the rule set storage unit 46. The rule set storage unit 46 stores rule sets that include structured data rules and instruction statements according to the type of structured data.

[0038] The types of structured data include, for example, rule sets that conform to the "Methods for Representing Machine-Readable Data in Statistical Tables" formulated by the Ministry of Internal Affairs and Communications, rule sets that conform to data entry rules in line with the operation of data processing programs, or rule sets for specific programs such as accounting data applications.

[0039] The table data acquisition unit 32 reads and acquires table data stored in the table data storage unit 48. The table data stored in the table data storage unit 48 is the table data to which improvement suggestions are to be output. The table data stored in the table data storage unit 48 is data arranged in rows and columns, and contains information about the items in the rows and columns. Note that the table data acquired by the table data acquisition unit 32 may be acquired based on input from the operator prior to the generation of the prompt described later. The input table data may be stored in the table data storage unit 48 by the table data acquisition unit 32.

[0040] The prompt generation unit 34 generates prompts using the structured data rules and instructions included in the rule set acquired by the rule set acquisition unit 30, and the table data acquired by the table data acquisition unit 32. Details of the prompts generated by the prompt generation unit 34 will be described later.

[0041] The prompt transmission unit 36 transmits the prompt generated by the prompt generation unit 34 to the large language model 16 of the answering device 14, and causes the large language model 16 to perform processing according to the prompt. The large language model 16 performs processing according to the prompt and generates an answer including a proposal for improving the table data.

[0042] The answer reception unit 38 receives an answer generated by the large language model 16, which includes a proposal for improving the table data.

[0043] The output unit 40 displays and outputs a proposal for improving the table data to the output device 502 of the structured data creation support device 12 or the output device 502 of the worker terminal 10 according to the answer generated by the large language model 16. Also, if there is no proposal for improving the table data, the output unit 40 may output, for example, "No problem".

[0044] The template creation unit 42 creates a template for the table data according to the answer generated by the large language model 16. The template created by the template creation unit 42 is displayed and output to the output device 502 of the structured data creation support device 12 or the output device 502 of the worker terminal 10.

[0045] <Processing> Figure 4 is a flowchart of an example of the structured data creation support system 1 according to the present embodiment.

[0046] In step S10, the worker operates the input device 501 of the structured data creation support device 12 or the input device 501 of the worker terminal 10 to specify the type of structured data for which a proposal for improving the table data is to be output. If there is only one type of structured data for which the rule set is stored in the rule set storage unit 46, the specification of the type of structured data may be omitted.

[0047] The rule set acquisition unit 30 of the structured data creation support device 12 acquires a rule set including the rules and instructions for the structured data corresponding to the specified type of structured data. The details of the rules and instructions for the structured data will be described later using FIG. 5.

[0048] In step S12, the operator operates the input device 501 of the structured data creation support device 12 or the input device 501 of the operator terminal 10 to specify the table data that is the target for outputting improvement proposals. The table data acquisition unit 32 of the structured data creation support device 12 acquires the specified table data from the table data storage unit 48. Note that the operator may input the table data that is the target for outputting improvement proposals each time.

[0049] In step S14, the prompt generation unit 34 of the structured data creation support device 12 uses the rules and instructions of the structured data included in the rule set acquired by the rule set acquisition unit 30 and the table data acquired by the table data acquisition unit 32 to generate, for example, the prompt 1000 shown in FIG. 5.

[0050] FIG. 5 is an explanatory diagram of an example of the prompt 1000 generated in step S14. The prompt 1000 shown in FIG. 5 illustrates an example where the rule set acquired in step S10 is a rule set that conforms to the "Method of Notation of Machine-Readable Data in Statistical Tables" formulated by the Ministry of Internal Affairs and Communications. The prompt 1000 in FIG. 5 includes an instruction 1002, a rule 1004 of structured data, and table data 1006 that is the target for outputting improvement proposals.

[0051] The instruction 1002 is the instruction included in the rule set acquired by the rule set acquisition unit 30 in step S10.

[0052] The rule 1004 of structured data is the rule of structured data included in the rule set acquired by the rule set acquisition unit 30 in step S10. The rule 1004 of structured data includes check items for data and check items for table configuration. The check items for data and table configuration are examples of the check items for data and table configuration.

[0053] The table data 1006 to which improvement suggestions are output is the table data acquired by the table data acquisition unit 32 in step S12. The table data 1006 represents the contents of a table in which data is arranged in rows and columns. The table data 1006 may also be in CSV (Comma-Separated Values) format. CSV format data can represent tabular data using comma-separated text data.

[0054] In step S16, the prompt transmission unit 36 ​​of the structured data creation support device 12 transmits the prompt 1000 shown in Figure 5, which was generated by the prompt generation unit 34 in step S14, to the large-scale language model 16 of the response device 14, causing the large-scale language model 16 to process according to the prompt 1000. The large-scale language model 16 processes according to the prompt 1000 shown in Figure 5 and generates a response that includes suggestions for improving the tabular data 1006 contained in the prompt 1000.

[0055] The response generated by the large-scale language model 16 includes information on the check items that the table data 1006 included in prompt 1000 violates, among the data check items and table structure check items included in prompt 1000, and the improved table data in accordance with the structured data rules included in prompt 1000.

[0056] In step S18, the response receiving unit 38 of the structured data creation support device 12 receives the response generated by the large-scale language model 16. The output unit 40 of the structured data creation support device 12 outputs improvement suggestions for the table data 1006 included in the prompt 1000 according to the received response. If there are no improvement suggestions for the table data 1006, the output unit 40 may output, for example, "No problem". The template creation unit 42 of the structured data creation support device 12 may also create a template for the table data 1006 according to the improvement suggestions for the table data 1006 included in the prompt 1000.

[0057] Figure 6 is an illustrative diagram of an example of an improvement suggestion output in step S18. Figure 6 shows an illustrative diagram of the screen 1100 on which the improvement suggestion is output. Input example 1102 is the table data 1006 included in prompt 1000. Output example 1104 is an example of table data improved in accordance with the structured data rules included in prompt 1000. Output example 1106 is an example of information regarding check items that the table data 1006 included in prompt 1000 violates. Output example 1106 may also be information explaining the improvements to the check items that the table data 1006 violates.

[0058] The prompt 1000 in Figure 5 includes data check items such as "each cell must contain one data item," "numerical data must not contain text," "cells must not be merged," and "the unit of data must be indicated in the column."

[0059] The input example 1102 shown in Figure 6 is an example of table data 1006 that violates the data check items "one data per cell", "numerical data does not contain text", "cells are not merged", and "the unit of data is indicated in the column".

[0060] The output example 1104 shown in Figure 6 is a table data that has been improved to comply with the data check items "one data item per cell," "numerical data does not contain text," "cells are not merged," and "the unit of data is indicated in the column."

[0061] The output example 1106 shown in Figure 6 provides information explaining the improvements (points for improvement) to the data check items that the table data 1006 violates: "one data item per cell," "numerical data does not contain text," "cells are not merged," and "the unit of data is indicated in the column."

[0062] By examining output examples 1104 and 1106 in Figure 6, the operator can easily understand the improvements needed to make the tabular data 1006 structured data. The improvement suggestions in Figure 6 are examples that mainly propose improvements to the way rows and columns are written in the tabular data 1006 (improvements to the data structure).

[0063] Furthermore, the template creation unit 42 of the structured data creation support device 12 can use the output example 1104 in Figure 6 to create a template 1204 from improved table data 1202 that does not violate the data and table structure check items, for example, as shown in the screen 1200 in Figure 7. The template 1204 may also be created by an operator by checking the output example 1104 in Figure 6. The template 1204 can then be used for subsequent data storage.

[0064] Figure 8 is an illustrative diagram of an example of an improvement suggestion output in step S18. Figure 8 shows an illustrative diagram of the screen 1100 on which the improvement suggestion is output. Input example 1102 is the table data 1006 included in prompt 1000. Output example 1104 is an example of table data improved in accordance with the structured data rules included in prompt 1000. Output example 1106 is an example of information regarding check items that the table data 1006 included in prompt 1000 violates. Output example 1106 may also be information explaining the improvements to the check items that the table data 1006 violates.

[0065] The prompt 1000 in Figure 5 includes data check items: "Numerical data must not contain text" and "The unit of data must be indicated in the column."

[0066] The input example 1102 shown in Figure 8 is an example of table data 1006 that violates the data check items "Numerical data should not contain text" and "Data units should be indicated in the column." The output example 1104 shown in Figure 8 is improved table data that does not violate the data check items "Numerical data should not contain text" and "Data units should be indicated in the column."

[0067] The output example 1106 shown in Figure 8 provides information explaining the improvements (points for improvement) to the data check items "Numerical data should not contain text" and "Data units should be indicated in the column" that the table data 1006 violates.

[0068] By examining output examples 1104 and 1106 in Figure 8, the operator can easily understand the improvements needed to make the tabular data 1006 structured data. The improvement suggestions in Figure 8 are examples that mainly propose improvements to the data (improvement of data content) written within a single cell of the tabular data 1006.

[0069] Furthermore, the template creation unit 42 of the structured data creation support device 12 can use the output example 1104 in Figure 8 to create a template 1204 from improved table data 1202 that does not violate the data and table structure check items, for example, as shown in the screen 1200 in Figure 9. The template 1204 may also be created by an operator by checking the output example 1104 in Figure 8. The template 1204 can then be used for subsequent data storage.

[0070] According to the flowchart shown in Figure 4, a structured data creation support device 12, a structured data creation support method, and a structured data creation support system 1 can be provided to assist in the creation of structured data in accordance with predetermined rules.

[0071] <Application Example> In the above example, we described an example in which the rule set obtained in step S10 conforms to the "Method for Representing Machine-Readable Data in Statistical Tables" formulated by the Ministry of Internal Affairs and Communications. The structured data creation support system 1 according to this embodiment can also be applied to rule sets that conform to data description rules in line with the operation of the data processing program.

[0072] Figure 10 illustrates an example of the normal operation of a data processing program. The table data 1300 shown in Figure 10 includes multiple tables 1302A to 1302D. As an example of the normal operation of the data processing program, the positions of keywords "A" to "D" in the table data 1300 are identified, the range of rows and columns from the position of each keyword "A" to "D" to a blank row is identified, and the data within the identified range is extracted as tables 1302A to 1302D.

[0073] If the table data 1300 is recorded according to the data entry rules that align with the operation of the data processing program, the data processing program will be able to read it from the machine. Furthermore, by recording the data according to the data entry rules that align with the operation of the data processing program, the operator can create a data storage template in the desired format according to the purpose of use.

[0074] However, operators need to understand the data entry rules that align with the operation of the data processing program. Therefore, a system that provides suggestions for improving tabular data 1300 in accordance with the data entry rules that align with the operation of the data processing program would make it easier for operators to create structured data that follows the data entry rules that align with the operation of the data processing program.

[0075] Figure 11 illustrates an example of applying data entry rules to the operation of a data processing program. Figure 11 shows an example where the data entry rules, which are aligned with the operation of a data processing program, are "insert blank rows and columns between tables. Place keywords in the upper left corner of tables. Row length is arbitrary."

[0076] The processing of the application examples shown in Figures 10 and 11 will be explained using Figure 4.

[0077] In step S10, the operator operates the input device 501 of the structured data creation support device 12 or the input device 501 of the operator terminal 10 to specify a rule set that conforms to the data description rules in accordance with the operation of the data processing program in Figure 10 as the type of structured data for which improvement suggestions for table data should be output. If there is only one type of structured data for which a rule set is stored in the rule set storage unit 46, specifying the type of structured data may be omitted. The rule set acquisition unit 30 of the structured data creation support device 12 acquires a rule set that includes data description rules and instruction statements in accordance with the operation of the data processing program in Figure 10.

[0078] In step S12, the worker operates the input device 501 of the structured data creation support device 12 or the input device 501 of the worker terminal 10 to specify the table data 1300 in Figure 11, which is the target for outputting improvement suggestions. The table data acquisition unit 32 of the structured data creation support device 12 acquires the specified table data from the table data storage unit 48. In step S14, the prompt generation unit 34 of the structured data creation support device 12 generates a prompt using the data description rules and instructions in accordance with the operation of the data processing program in Figure 10 included in the rule set acquired by the rule set acquisition unit 30, and the table data 1300 in Figure 11 acquired by the table data acquisition unit 32.

[0079] In step S16, the prompt transmission unit 36 ​​of the structured data creation support device 12 transmits the prompt generated by the prompt generation unit 34 in step S14 to the large-scale language model 16 of the answer device 14, causing the large-scale language model 16 to process according to the prompt. The large-scale language model 16 processes according to the prompt and generates an answer that includes suggestions for improving the table data 1300 of Figure 11 that were included in the prompt. The answer generated by the large-scale language model 16 includes table data 1304 in which the table data 1300 of Figure 11 has been improved in accordance with the data description rules included in the prompt.

[0080] In step S18, the response receiving unit 38 of the structured data creation support device 12 receives the response generated by the large-scale language model 16. The output unit 40 of the structured data creation support device 12 outputs an improvement suggestion for the table data 1300 in Figure 11 included in the prompt, according to the received response. Alternatively, the template creation unit 42 of the structured data creation support device 12 may create a template for the table data 1300 according to the improvement suggestion for the table data 1300 in Figure 11 included in the prompt. Note that the table data 1304 in Figure 11 is an example of table data in which the table data 1300 has been improved in accordance with the data description rules included in the prompt.

[0081] By examining the improved tabular data 1304 in Figure 11, the operator can easily understand the improvements needed to create structured data that conforms to the data entry rules in line with the operation of the data processing program in Figure 10.

[0082] Furthermore, the template creation unit 42 of the structured data creation support device 12 can create a template from the improved table data 1304 in Figure 11, ensuring that it does not violate the check items of the data entry rules in accordance with the operation of the data processing program. The template may also be created by an operator by checking the improved table data 1304 in Figure 11. The created template can then be used for subsequent data storage.

[0083] As described above, this embodiment can be understood to be capable of various modifications to its form and details without departing from the spirit and scope of the claims.

[0084] For example, the tabular data improved by the structured data creation support system 1 according to this embodiment, in accordance with the rules of structured data, is machine-readable data. For example, by using a template created from the improved tabular data for data storage, an operator can easily extract or process the stored machine-readable data using a computer and perform data analysis such as statistical analysis or machine learning.

[0085] Furthermore, since the tabular data improved by the structured data creation support system 1 according to this embodiment in accordance with the rules of structured data is machine-readable data, it may be used, for example, to control a mixed material generation device that generates mixed materials by specifying the mixing ratio of mixed materials.

[0086] Although the present invention has been described above based on examples, the present invention is not limited to the above examples, and various modifications are possible within the scope of the claims. This application claims priority to U.S. Patent and Trademark Office Provisional Application No. 63 / 743,354, filed on 9 January 2025, which is incorporated herein by reference in its entirety.

[0087] 1 Structured data creation support system 10 Worker terminal 12 Structured data creation support device 14 Response device 16 Large-scale language model 34 Prompt generation unit 38 Response receiving unit 40 Output unit 42 Template creation unit 1000 Prompt

Claims

1. A structured data creation support device that uses a large language model to perform processing according to a prompt, the device comprising: A prompt generation unit that generates the prompt including rules, instructions, and table data of the structured data; An answer reception unit that causes the large language model to perform processing according to the prompt and receives an answer generated by the large language model; An output unit that outputs a proposal for improving the table data according to the answer. A structured data creation support device having the above components.

2. The rules of the structured data include items for checking data and table configurations. The answer generated by the large language model includes information on the items of the check that the table data violates and the table data improved according to the rules. The output unit outputs, as a proposal for improving the table data, information on the items of the check that the table data violates and the table data improved according to the rules. The structured data creation support device according to claim 1.

3. The information on the items of the check that the table data violates includes information explaining the improvement points of the items of the check that the table data violates. The structured data creation support device according to claim 2.

4. The rules of the structured data include items for checking data description rules in accordance with the operation of a data processing program. The answer generated by the large language model includes information on the items of the check that the table data violates and the table data improved according to the rules. The output unit outputs, as a proposal for improving the table data, information on the items of the check that the table data violates and the table data improved according to the rules. The structured data creation support device according to claim 1.

5. The prompt generation unit reads out a rule set including the rules and instructions of the structured data according to the type of the structured data and generates the prompt including the rules, instructions, and table data of the structured data. The structured data creation support device according to any one of claims 1 to 4.

6. The structured data creation support device according to any one of claims 1 to 5, further comprising a template creation unit that creates a template of the structured data according to the answer.

7. A structured data creation support method performed by a structured data creation support device that uses a large language model to perform processing according to a prompt, the method comprising: a prompt generation step of generating the prompt including rules, instructions, and table data of the structured data; an answer reception step of causing the large language model to perform processing according to the prompt and receiving an answer generated by the large language model; and an output step of outputting an improvement proposal for the table data according to the answer.

8. A structured data creation support system that uses a large language model to perform processing according to a prompt, the system comprising: a prompt generation unit that generates the prompt including rules, instructions, and table data of the structured data; a large language model that performs processing according to the prompt and generates an answer to the prompt; an answer reception unit that receives the answer generated by the large language model; and an output unit that outputs an improvement proposal for the table data according to the answer.

Images