Data structure information generation device, data extraction device, data extraction system, data structure information generation method, and data extraction method

Abstract

Provided is a data structure information generation device for generating, by using a large-scale language model for processing according to a prompt, data structure information in which pieces of position information about keys and values included in a table portion of data described in a table format is designated in a text format. The device comprises: an input reception unit for receiving an input of the table portion of the data described in the table format; a proposal unit for proposing, to an operator, candidates of the keys and the values analyzed from the table portion; an instruction reception unit for receiving, from the operator, an instruction for correcting or confirming the candidates of the keys and the values; a range designation reception unit for receiving, from the operator, designation of a range of the table portion in the data described in the table format; a data structure information generation unit for generating the data structure information on the basis of the confirmed keys and values and the designation of the range of the table portion; and a presentation unit for presenting the generated data structure information to the operator.

Description

Data Structure Information Generation Device, Data Extraction Device, Data Extraction System, Data Structure Information Generation Method, and Data Extraction Method

[0001] The present disclosure relates to a data structure information generation device, a data extraction device, a data extraction system, a data structure information generation method, and a data extraction method.

[0002] For example, an experimenter inputs experimental data into format data described in a tabular format such as an Excel (registered trademark) format prepared for inputting experimental data. The format data described in the tabular format is prepared for each department of the experimenter, for example.

[0003] An analyst who acquires and analyzes the experimental data input by the experimenter into the format data extracts analysis data suitable for analysis using a data aggregation formatting program developed in accordance with the format data from the experimental data input by the experimenter into the format data.

[0004] Patent Document 1 discloses an experimental data management system and an electronic laboratory notebook system for integrating and utilizing data over a wider area, such as sharing data of multiple companies and academic institutions.

[0005] Japanese Patent Application Laid-Open No. 2023-147330

[0006] The data aggregation formatting program developed in accordance with the format data needs to be modified every time the format data is changed. Also, when the format data is prepared for each department of the experimenter, the development man-hours of the data aggregation formatting program increase due to an increase in the number of departments.

[0007] An object of the present disclosure is to provide a data structure information generation device, a data extraction device, a data extraction system, a data structure information generation method, and a data extraction method that reduce the labor of extracting data described as a combination of keys and values from the table part of data described in a tabular format.

[0008] The present disclosure has the following configuration.

[0009] [1] A data structure information generating device that generates data structure information in which the positional information of keys and values ​​included in the table portion of data described in tabular format is specified in text format, using a large-scale language model that processes according to prompts, the device comprising: an input receiving unit that receives input of the table portion of the data described in tabular format from the large-scale language model; a suggestion unit that proposes candidate keys and values ​​analyzed from the table portion to the operator, the large-scale language model; an instruction receiving unit that receives instructions from the operator to modify or confirm the candidate keys and values, the large-scale language model; a range specification receiving unit that receives instructions from the operator to specify the range of the table portion in the data described in tabular format from the operator; a data structure information generating unit that generates the data structure information based on the confirmed keys and values ​​and the specified range of the table portion, the large-scale language model; and a presentation unit that presents the generated data structure information to the operator.

[0010] [2] The data structure information generating apparatus according to [1], further comprising: [2] an illustrative section that illustrates key-value combinations included in the table portion of the data described in tabular format, based on the data structure information generated by the large-scale language model.

[0011] [3] The data structure information generating apparatus according to [1] or [2], wherein the prompt includes information defining the format of the data structure information, a processing procedure for generating the data structure information, and information relating to interaction with the operator.

[0012] [4] The data structure information generating device according to any one of [1] to [3], wherein the data described in the tabular format is tabular format data prepared for inputting predetermined data.

[0013] [5] The data structure information generation device according to [4], wherein the data structure information is generated after the format data has been created or modified.

[0014] [6] A data extraction device for extracting data described as key-value pairs from a table portion of data described in tabular format, comprising: an acquisition unit that acquires data structure information in which the positional information of keys and values ​​included in the table portion of the data described in tabular format is specified in text format; an extraction unit that extracts data described as key-value pairs from the table portion of the data described in tabular format according to the data structure information; and an output unit that outputs the extracted data.

[0015] A data extraction system comprising: [7] A data extraction device described in [6]; and a data structure information generation device that generates the data structure information using a large-scale language model that processes according to prompts.

[0016] [8] The data structure information generation device comprises: an input receiving unit that receives input of a table portion of the data described in tabular format from a large-scale language model that processes according to prompts; a suggestion unit that proposes key and value candidates analyzed from the table portion to the operator; an instruction receiving unit that receives instructions from the operator to modify or confirm the key and value candidates to the large-scale language model; a range specification receiving unit that receives a range specification from the operator to the large-scale language model for the range of the table portion in the data described in tabular format; a data structure information generation unit that generates the data structure information based on the confirmed key and value and the range specification of the table portion to the large-scale language model; and a presentation unit that presents the generated data structure information to the operator.

[0017] [9] A data structure information generation method in which an information processing device generates data structure information in which the positional information of keys and values ​​included in the table portion of data described in tabular format is specified in text format, using a large-scale language model that processes according to prompts, the method comprising: an input reception procedure in which the large-scale language model receives input of the table portion of the data described in tabular format; a suggestion procedure in which the large-scale language model proposes candidate keys and values ​​analyzed from the table portion to an operator; an instruction reception unit in which the large-scale language model receives instructions from an operator to modify or confirm the candidate keys and values; a range specification reception procedure in which the large-scale language model receives a range specification from an operator to specify the range of the table portion in the data described in tabular format; a data structure information generation procedure in which the large-scale language model generates the data structure information based on the confirmed keys and values ​​and the range specification of the table portion; and a presentation procedure in which the large-scale language model presents the generated data structure information to an operator.

[0018]

[10] A data extraction method in which an information processing device extracts data described as key-value combinations from the table portion of data described in tabular format, comprising: an acquisition step of obtaining data structure information in which the positional information of keys and values ​​included in the table portion of the data described in tabular format is specified in text format; an extraction step of extracting data described as key-value combinations from the table portion of the data described in tabular format according to the data structure information; and an output step of outputting the extracted data.

[0019] According to this disclosure, the effort required to extract data described as key-value pairs from the table portion of data described in tabular format can be reduced.

[0020] This is a configuration diagram of an example of the data extraction system 1 according to this embodiment. This is a hardware configuration diagram of an example of the computer 500 according to this embodiment. This is an explanatory diagram of an example of the processing of the data extraction device 18 according to this embodiment. This is an explanatory diagram of an example of the automated or semi-automated process of schema generation using AI. This is an explanatory diagram of an example of the automated or semi-automated process of schema generation using AI. This is a functional configuration diagram of an example of the schema generation device 14 according to this embodiment. This is a configuration diagram of an example of the prompt according to this embodiment. This is a configuration diagram of an example of the prompt according to this embodiment. This is a configuration diagram of an example of the prompt according to this embodiment. This is an image diagram of an example of the application editing screen 1100. This is an example flowchart showing the processing of the schema generation device 14 according to this embodiment. This is an image diagram showing an example of the processing of the schema generation device 14 according to this embodiment. This is an image diagram showing an example of the processing of the schema generation device 14 according to this embodiment. This is an image diagram showing an example of the processing of the schema generation device 14 according to this embodiment. This is an image diagram showing an example of the processing of the schema generation device 14 according to this embodiment. This is an image diagram showing an example of the processing of the schema generation device 14 according to this embodiment. This is an image diagram showing an example of the processing of the schema generation device 14 according to this embodiment. This is an image diagram showing an example of the processing of the schema generation device 14 according to this embodiment. This is an example of the configuration diagram of the data extraction device 18 according to this embodiment. This is an example flowchart illustrating the processing of the data extraction system 1 according to this embodiment.

[0021] Next, embodiments of the present invention will be described in detail. However, the present invention is not limited to the following embodiments.

[0022] <System Configuration> Figure 1 is a diagram showing an example of the configuration of the data extraction system 1 according to this embodiment. The data extraction system 1 in Figure 1 processes experimental data as an example of data described in tabular format. The experimental data is digital data in which experimental results, etc., by the experimenter are entered into a tabular format data such as the Excel® format prepared for inputting experimental data.

[0023] The data extraction system 1 comprises an experimenter terminal 10, an analysiser terminal 12, a schema generator 14, an experimental data storage device 16, and a data extraction device 18, and is 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 experimenter terminal 10 is an information processing device such as a PC (Personal Computer), tablet, or smartphone operated by the experimenter. The experimenter can operate the experimenter terminal 10 and utilize various functions provided by the schema generation device 14 or the experiment data storage device 16.

[0025] The analysis terminal 12 is an information processing device such as a PC, tablet, or smartphone operated by the analysis operator. The analysis operator can operate the analysis terminal 12 and utilize the various functions provided by the experimental data storage device 16 and the data extraction device 18.

[0026] The schema generator 14 uses a large-scale language model (hereinafter referred to as LLM) that processes according to prompts to generate a schema in which the key and value position information (such as cell positions) included in the table portion of the experimental data is specified in text format. A schema is an example of data structure information in which the key and value position information included in the table portion of data described in tabular format is specified in text format. The schema generator 14 stores the generated schema in the experimental data storage device 16. Note that the schema generator 14 is an example of a data structure information generator.

[0027] The experimental data storage device 16 stores experimental data entered by the experimenter in association with the schema of the format data in which the experimental data is entered. The schema stored by the experimental data storage device 16 may be generated by the schema generation device 14, generated manually by the experimenter, or generated by the experimenter using an auxiliary tool.

[0028] The experimental data storage device 16 can be any device or system that allows access to the schema and experimental data from the experimenter's terminal 10, the analysis terminal 12, the schema generation device 14, or the data extraction device 18, such as a NAS (Network Attached Storage) or an electronic laboratory notebook system.

[0029] The data extraction device 18 acquires experimental data from which data for analysis is to be extracted, and the schema of the format data into which the experimental data is entered. The data extraction device 18 extracts data for analysis from the table portion of the experimental data entered by the experimenter, according to the acquired schema. The data extraction device 18 outputs the extracted data for analysis in a way that makes it available to the analyst. For example, displaying the extracted data for analysis, or storing it in a storage area of ​​a device or system that the analyst can access, are examples of outputs that make the data available to the analyst. The data for analysis is an example of data described by a key-value combination.

[0030] The schema generation device 14, the experimental data storage device 16, and the data extraction device 18 may be integrated into a single configuration, either partially or entirely. LLM is a natural language processing model trained using a large amount of text data, and can utilize BERT (Bidirectional Encoder Representations from Transformers), GPT (Generative Pre-Trained Transformer), etc.

[0031] Furthermore, the schema generation device 14, experimental data storage device 16, and data extraction device 18 shown in Figure 1 can be implemented using a workstation or PC.

[0032] Note that the configuration of the data extraction system 1 shown in Figure 1 is just one example. The configuration of the data extraction system 1 can vary depending on the application and purpose. For example, the schema generation device 14, the experimental data storage device 16, and the data extraction device 18 may be implemented using multiple computers, or they may be implemented as a cloud service.

[0033] <Hardware Configuration> The experimenter terminal 10, the analysis terminal 12, the schema generation device 14, the experimental data storage device 16, and the data extraction device 18 shown in Figure 1 can be realized by a computer 500 with the hardware configuration shown in Figure 2.

[0034] 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, all of which are interconnected via bus B. The input device 501 and the output device 502 may also be used by connecting them to the computer 500 via the external interface 503.

[0035] 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.

[0036] 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).

[0037] 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.

[0038] 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.

[0039] <Processing Overview> Figure 3 is an explanatory diagram of an example of the processing of the data extraction device 18 according to this embodiment. The data extraction device 18 extracts data for analysis from the table portion of experimental data written in tabular format, according to the schema of the experimental data written in tabular format. The schema is a file that describes the data structure of experimental data using a limited grammar.

[0040] For example, a schema specifies, in text format, the location information such as the absolute or relative coordinates of keys and values ​​included in the table portion of experimental data described in tabular form, and the range of the table portion in the experimental data.

[0041] Thus, the data extraction device 18 according to this embodiment extracts data for analysis from experimental data according to the acquired schema. For this reason, the data extraction device 18 does not need to modify its program in response to changes in the format data into which the experimental data is input; it only needs to respond by modifying the schema in a way that is easy for the operator to understand.

[0042] Furthermore, since the data extraction device 18 according to this embodiment does not require program modification in response to changes in the format data into which experimental data is input, schema management can be delegated to on-site personnel such as experimenters.

[0043] The data extraction device 18 according to this embodiment eliminates the need to modify the program that was required every time the format data into which experimental data is input is changed. Further, even when the number of departments increases when the format data is prepared for each department of the experimenter, the data extraction device 18 according to this embodiment does not increase the man-hours for program development.

[0044] By utilizing AI (Artificial Intelligence) such as generative AI including LLM, the schema can automate or semi-automate generation as shown below.

[0045] FIG. 4 is an explanatory diagram of an example of the process of automating or semi-automating the generation of a schema using AI. For example, converting the table part of the experimental data described in the table format of FIG. 4 into data as a table is possible by the structure analysis of AI, which is an existing technology.

[0046] The format of the data required as analysis data is data described by combinations of keys and values included in the table part of the experimental data. One example of the data described by combinations of keys and values extracted from the tabularized table of FIG. 4 is, for example, the key "anhydrous acid, anhydrous acid A, and planned charge amount" and the value "5.00".

[0047] As described above, since the keys in the tabularized table shown in FIG. 4 are complex, there are cases where AI cannot make a correct determination. Therefore, for the tabularized table shown in FIG. 4, it was necessary for the operator to specify the key range while saving labor with AI.

[0048] FIG. 5 is an explanatory diagram of an example of the process of automating or semi-automating the generation of a schema using AI. For example, in FIG. 5, the difference from FIG. 4 is that values are input to supplement the table part of the experimental data described in the table format.

[0049] One example of a key-value combination extracted from the digitized table in Figure 5 is the key "Acid anhydride, Acid anhydride A, manufactured by Company B, and planned batching amount" with the value "5.00". As shown in the digitized table in Figure 5, even for the same reagent, the amount of impurities and other properties may differ depending on the supplier, which can affect the properties. Therefore, it is desirable to include generally unnecessary supplementary information in the key.

[0050] Thus, as shown in Figure 5, the digitized table generally required supplementary or ancillary information on experimental data that is not typically considered necessary for analysis, or the purpose of the analysis might change. Therefore, while AI was used to streamline the process, the operator still needed to specify the key range.

[0051] Therefore, in this embodiment, the generation of the schema used by the data extraction device 18 is streamlined by utilizing AI, and the range of keys and the range of table portions in the experimental data are accepted from the operator. For example, the schema generation device 14 accepts the range of keys and the range of table portions in the experimental data from the operator using an LLM that processes according to prompts. Details of the prompts that cause the LLM to process the input to generate the schema after receiving the range of keys and the range of table portions in the experimental data from the operator will be described later.

[0052] In this embodiment, schema accuracy is improved by accepting specification of key ranges and ranges of table portions in experimental data from the operator. Because it is easy to generate highly accurate schemas, this embodiment facilitates the process of extracting data for analysis from the table portions of experimental data with high accuracy.

[0053] <Functional Configuration> An example of the functional configuration of the schema generation device 14 and data extraction device 18 according to this embodiment will be described below.

[0054] Figure 6 is a functional configuration diagram of an example of the schema generation device 14 according to this embodiment. Note that parts of the configuration diagram in Figure 6 that are not necessary for the explanation of this embodiment have been appropriately omitted. Figures 7 to 10 are configuration diagrams of an example of a prompt according to this embodiment.

[0055] The schema generation device 14 has an LLM 30 that processes according to prompts 46, for example, shown in Figures 7 to 10. Figures 7 to 10 show a series of consecutive prompts 46. A prompt is an instruction statement written in data such as a string that instructs the LLM 30 on conditions for generating a schema.

[0056] The prompt 46 shown in Figures 7 to 10 illustrates an example of generating a schema in a format that describes the structure, such as YAML. However, to fix the output, it is desirable to specify a format such as YAML or JSON.

[0057] The section 1000 of the prompt 46 shown in Figures 7 to 10 is a description that instructs the LLM 30 of the format of the schema to be output. Section 1000 is an example of information that defines the format of the schema. Section 1002 of the prompt 46 is a description that instructs the LLM 30 of the operation flow. Section 1002 is an example of a processing procedure for generating a schema.

[0058] The description section 1004 of prompt 46 is a description that teaches LLM 30 examples of interactions between the operator and the chatbot. Description section 1004 is an example of information regarding interactions with the operator. The chatbot can interact with the operator according to description section 1004. For example, description section 1004 includes a description that teaches LLM 30 to suggest key and value candidates analyzed by LLM 30 to the operator, to explain the reason for selecting the suggested keys, and to present alternative examples of how to select them. Description section 1004 also includes a description that teaches LLM 30 to confirm the range of the table portion in the experimental data with the operator.

[0059] The description section 1006 of prompt 46 provides an example of what key-value combinations will be extracted when data for analysis is extracted from experimental data according to the generated schema.

[0060] Returning to Figure 6, the LLM 30 of the schema generation device 14 functions as an input receiving unit 32, a suggestion unit 34, an instruction receiving unit 36, a range specification receiving unit 38, a schema generation unit 40, a presentation unit 42, and an example unit 44 by processing according to the prompt 46.

[0061] The input reception unit 32 receives input of the table portion of the experimental data, which is written in a tabular format, for example from the experiment staff terminal 10 operated by the experiment staff member or other worker. The suggestion unit 34 analyzes the table portion of the experimental data, which is written in a tabular format, and proposes candidate keys and values ​​to the worker by displaying them on the experiment staff terminal 10 operated by the experiment staff member or other worker.

[0062] The instruction receiving unit 36 ​​receives instructions to modify or confirm the proposed key and value candidates from, for example, the experimenter terminal 10 operated by an experimenter or other worker. The range specification receiving unit 38 receives the specification of the range of the table portion in the experimental data described in tabular format from, for example, the experimenter terminal 10 operated by an experimenter or other worker. The schema generation unit 40 generates a schema based on the key and value confirmed by the instructions received by the instruction receiving unit 36 ​​and the range specification of the table portion in the experimental data received by the range specification receiving unit 38. Note that the schema generation unit 40 is an example of a data structure information generation unit.

[0063] The presentation unit 42 displays the schema generated by the schema generation unit 40 on the experiment terminal 10 operated by the experiment manager or other worker, for example, to present it to the worker. The example unit 44, based on the schema generated by the schema generation unit 40, displays the key-value combinations included in the table portion of the experimental data, which is written in tabular format, on the experiment terminal 10 operated by the experiment manager or other worker, for example, to present them to the worker.

[0064] Figure 16 is a configuration diagram of an example of a data extraction device 18 according to this embodiment. Note that parts of the configuration diagram in Figure 16 that are not necessary for the explanation of this embodiment have been appropriately omitted. The data extraction device 18 functions as an acquisition unit 50, an extraction unit 52, and an output unit 54 by executing a program for the data extraction device 18.

[0065] The acquisition unit 50 acquires a schema in which the key and value position information included in the table portion of the experimental data described in tabular format is specified in text format, for example, from the experimental data storage device 16. The acquisition unit 50 acquires a schema specified from, for example, the analysis terminal 12 operated by an operator such as an analysis technician. Alternatively, the acquisition unit 50 may acquire a schema stored in the experimental data storage device 16 in association with the format data into which the experiment technician entered the experimental data.

[0066] The extraction unit 52 extracts analysis data, described as key-value pairs, from the tabular portion of the experimental data, which is written in a tabular format, according to the acquired schema. The output unit 54 outputs the analysis data extracted by the extraction unit 52 so that it can be used by the analyst. For example, the output unit 54 displays the analysis data on the analyst terminal 12 operated by the analyst or other worker. Alternatively, the output unit 54 may store the data in a storage area accessible from the analyst terminal 12 operated by the analyst or other worker.

[0067] <Processing> Figure 11 is an image diagram of an example of the application editing screen 1100. The schema generation device 14 can prepare a schema generation application incorporating a system prompt by, for example, entering a prompt 46 as a system prompt on the application editing screen 1100 in Figure 11 and pressing the execute button 1102. The schema generation device 14 executes the schema generation application incorporating the system prompt and processes the application according to the procedure shown in the flowchart in Figure 12.

[0068] Figure 12 is a flowchart illustrating an example of the processing of the schema generation device 14 according to this embodiment. Figures 13A to 13C, 14A to 14C, and 15 are illustrative diagrams illustrating an example of the processing of the schema generation device 14 according to this embodiment.

[0069] In step S10, the input receiving unit 32 of the schema generation device 14 receives input of the table portion of experimental data described in tabular format from, for example, the experimenter's terminal 10 operated by an experimenter or other worker. Figure 13A shows an example of experimental data described in tabular format. For example, the worker operating the experimenter's terminal 10 inputs text data of the table portion 1300, as shown in Figure 13B, into the schema generation application by copying and pasting the table portion 1300 of the experimental data in Figure 13A. For example, by enabling acceptance of the TSV (Tab Separated Values) file format, the schema generation application can accept input of text data of the table portion 1300 copied and pasted from experimental data in Excel format.

[0070] In step S12, the suggestion unit 34 of the schema generation device 14 analyzes candidate keys and values ​​from the text data of the table portion 1300 in Figure 13B, which was received as input in step S10, according to the prompt 46. The suggestion unit 34 proposes the candidate keys and values ​​analyzed from the text data of the table portion 1300 to the worker by displaying them on the experimenter terminal 10, which is operated by the worker such as the experimenter, as shown in Figure 13C. In Figure 13C, in addition to the candidate keys and values ​​analyzed from the text data of the table portion 1300, the reason for selecting the proposed keys (proposal reason) and alternative examples of how to select them (changed examples) are also proposed.

[0071] Furthermore, Figure 13C includes a message to receive confirmation from, for example, an experimenter or other worker, whether to confirm the proposed key and value candidates. The experimenter or other worker reviews the proposed content in Figure 13C and receives instructions to confirm or modify the proposed key and value candidates from the experimenter terminal 10 operated by the experimenter or other worker. The instructions to confirm or modify the proposed key and value candidates are, for example, given by the experimenter or other worker who specifies the key.

[0072] In step S14, the instruction receiving unit 36 ​​determines whether or not it has received an instruction to modify the proposed key and value candidates from the experiment operator terminal 10 operated by an experiment operator or other worker. If it has received an instruction to modify the proposed key and value candidates, the instruction receiving unit 36 ​​performs the process in step S16. In step S16, the instruction receiving unit 36 ​​modifies the key and value candidates according to the modification instruction, and then returns to the process in step S14.

[0073] If an instruction is received to confirm the proposed key and value candidates, the process proceeds from step S14 to step S18. In step S18, the range specification reception unit 38 accepts the specification of the range of the table portion 1300 in the experimental data, and for example, as shown in Figure 14A, it provides an example of how to specify the range of the table portion 1300.

[0074] Operators such as experiment managers can specify the range of the table portion 1300 in the experimental data, for example as shown in Figure 14B, by confirming the method for specifying the range of the table portion 1300 as exemplified in Figure 14A. The range specification reception unit 38 receives the specification of the range of the table portion 1300 in the experimental data from the experiment manager terminal 10 operated by the operator such as the experiment manager.

[0075] In step S20, the schema generation unit 40 generates a schema, for example, as shown in Figure 14C, based on the key and value determined by the instruction received by the instruction receiving unit 36 ​​and the range specification of the table portion 1300 in the experimental data received by the range specification receiving unit 38.

[0076] In step S22, the presentation unit 42 presents the schema generated in step S20 to the experimenter or other operator by displaying it in the display area 2002 of the screen 2000, for example, as shown in Figure 15. In addition, the display area 2004 of the screen 2000 shows examples of key-value combinations extracted from the table portion 1300 based on the schema generated in step S20. The example unit 44 extracts key-value combinations extracted from the table portion 1300 based on the schema generated in step S20 and presents them to the experimenter or other operator by displaying them in the display area 2004 of the screen 2000.

[0077] According to the flowchart shown in Figure 12, when an experimenter or other worker creates or modifies format data containing experimental data, they can easily generate a schema for the newly created or modified format data.

[0078] Figure 17 is a flowchart illustrating an example of the processing of the data extraction system 1 according to this embodiment.

[0079] In step S30, for example, an experimenter or other worker defines format data, such as an Excel format, for inputting experimental data. The format data for inputting experimental data can be any data written in a table format, and may be defined in a file such as a spreadsheet or database software file.

[0080] In step S32, for example, an experimenter or other worker transmits the format data for inputting experimental data, as defined in step S30, from the experimenter terminal 10 operated by the experimenter or other worker to the schema generation device 14. The input receiving unit 32 of the schema generation device 14 receives the format data defined in step S30. The schema generation device 14 generates a schema for the format data for inputting experimental data using the procedure shown in Figure 12, and stores it in the experiment data storage device 16 in association with the format data for inputting experimental data.

[0081] In step S34, for example, an experimenter or other worker inputs experimental data into the format data defined in step S30 and transmits it from the experimenter terminal 10 operated by the experimenter or other worker to the experiment data storage device 16. The experiment data storage device 16 stores the experimental data input by the experimenter or other worker into the format data and associates it with the schema of the format data into which the experimental data is input. The processing in step S34 is accumulated as the experimenter or other worker performs their duties.

[0082] The data extraction device 18 starts processing in step S36 at the time when an operator, such as an analyst, performs the analysis of the experimental data accumulated in step S34.

[0083] In step S36, the acquisition unit 50 of the data extraction device 18 acquires the experimental data to be analyzed and the schema of the format data in which the experimental data is input from the experimental data storage device 16.

[0084] In step S38, the extraction unit 52 extracts analysis data described as key-value pairs from the table portion of the experimental data, according to the schema obtained from the experimental data storage device 16 in step S36.

[0085] In step S40, the output unit 54 outputs the analysis data extracted in step S38. For example, the output unit 54 displays the analysis data on the analysis terminal 12 operated by an operator such as an analyst. Alternatively, the output unit 54 may store the data in a storage area accessible from the analysis terminal 12 operated by an operator such as an analyst.

[0086] According to the flowchart shown in Figure 17, operators such as analysts can easily output analysis data suitable for analysis, described as key-value pairs, from the experimental data entered into the format data by operators such as experimenters.

[0087] Although this embodiment has been described above, it will be understood that various modifications to the form and details are possible without departing from the spirit and scope of the claims. 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 described in the claims. This application claims priority to Basic Application No. 2025-003466 filed with the Japan Patent Office on January 9, 2025, the entire contents of which are incorporated herein by reference.

[0088] 1 Data Extraction System 10 Experimenter Terminal 12 Analysiser Terminal 14 Schema Generation Device 16 Experimental Data Storage Device 18 Data Extraction Device 30 LLM 32 Input Reception Unit 34 Proposal Unit 36 ​​Instruction Reception Unit 38 Range Specification Reception Unit 40 Schema Generation Unit 42 Presentation Unit 44 Example Unit 50 Acquisition Unit 52 Extraction Unit 54 Output Unit

Claims

1. A data structure information generating device that generates data structure information in which the positional information of keys and values ​​included in the table portion of data described in tabular format is specified in text format, using a large-scale language model that processes according to prompts, the device comprising: an input receiving unit that receives input of the table portion of the data described in tabular format from the large-scale language model; a suggestion unit that proposes candidate keys and values ​​analyzed from the table portion to the operator, the large-scale language model; an instruction receiving unit that receives instructions from the operator to modify or confirm the candidate keys and values, the large-scale language model; a range specification receiving unit that receives a range specification from the operator for the table portion of the data described in tabular format from the large-scale language model; a data structure information generating unit that generates the data structure information based on the confirmed keys and values ​​and the specified range of the table portion, the large-scale language model; and a presentation unit that presents the generated data structure information to the operator.

2. The data structure information generation device according to claim 1, further comprising: an illustrative section which illustrates key-value combinations included in the table portion of the data described in tabular format, based on the generated data structure information of the large-scale language model.

3. The data structure information generating apparatus according to claim 1 or 2, wherein the prompt includes information defining the format of the data structure information, a processing procedure for generating the data structure information, and information relating to interaction with the operator.

4. The data structure information generation device according to any one of claims 1 to 3, wherein the data described in the tabular format is tabular format data prepared for inputting predetermined data.

5. The data structure information generating apparatus according to claim 4, wherein the data structure information is generated after the format data has been created or modified.

6. A data extraction device for extracting data described as key-value pairs from a table portion of data described in tabular format, comprising: an acquisition unit that acquires data structure information in which the positional information of keys and values ​​included in the table portion of the data described in tabular format is specified in text format; an extraction unit that extracts data described as key-value pairs from the table portion of the data described in tabular format according to the data structure information; and an output unit that outputs the extracted data.

7. A data extraction system comprising: a data extraction device according to claim 6; and a data structure information generation device that generates the data structure information using a large-scale language model that processes according to prompts.

8. The data structure information generation device comprises: an input receiving unit that receives input of a table portion of the data described in tabular format from a large-scale language model that processes according to prompts; a suggestion unit that proposes candidate keys and values ​​analyzed from the table portion to the operator; an instruction receiving unit that receives instructions from the operator to modify or confirm the candidate keys and values ​​to the large-scale language model; a range specification receiving unit that receives a range specification from the operator to the large-scale language model for the range of the table portion in the data described in tabular format; a data structure information generation unit that generates the data structure information based on the confirmed keys and values ​​and the specified range of the table portion to the large-scale language model; and a presentation unit that presents the generated data structure information to the operator to the large-scale language model.

9. A data structure information generation method in which an information processing device generates data structure information in which key and value position information included in the table portion of data described in tabular format is specified in text format, using a large-scale language model that processes according to prompts, the method comprising: an input reception procedure in which the large-scale language model receives input of the table portion of the data described in tabular format; a suggestion procedure in which the large-scale language model proposes candidate keys and values ​​analyzed from the table portion to the operator; an instruction reception unit in which the large-scale language model receives instructions from the operator to modify or confirm the candidate keys and values; a range specification reception procedure in which the large-scale language model receives a range specification from the operator for the table portion of the data described in tabular format; a data structure information generation procedure in which the large-scale language model generates the data structure information based on the confirmed keys and values ​​and the range specification of the table portion; and a presentation procedure in which the large-scale language model presents the generated data structure information to the operator.

10. A data extraction method in which an information processing device extracts data described as key-value pairs from the table portion of data described in tabular format, comprising: an acquisition step of obtaining data structure information in which the positional information of keys and values ​​included in the table portion of the data described in tabular format is specified in text format; an extraction step of extracting data described as key-value pairs from the table portion of the data described in tabular format according to the data structure information; and an output step of outputting the extracted data.

Images