Structural data creation support program, method, and device

Abstract

To support creation of structure data of a substance while reducing a work cost by a simple operation requiring no special technique.SOLUTION: A structural data creation support device receives a character string related to creation of structural data of a substance from a user, selects an operation type from the received character string by using a classifier predefined to select the operation type from the character string, and creates and outputs the structural data by using a module corresponding to the operation type selected by the classifier among a plurality of modules prepared in advance for each operation type and the received character string.SELECTED DRAWING: Figure 7

Description

[Technical Field]

[0001] The disclosed technology relates to a structured data creation support program, a structured data creation support method, and a structured data creation support device. [Background technology]

[0002] In fields such as materials engineering, materials science, life science, and drug discovery, computational techniques such as density functional theory calculations and molecular dynamics calculations are known as methods for calculating on a computer the properties of a substance whose structure is represented by a collection of atoms with a specific number, position, and atomic species. For example, in materials engineering or materials science, structural data on metal materials, catalysts, polymer compounds, etc. is handled, while in life science or drug discovery, structural data on proteins, etc. is handled. The content of the structural data representing the structure of a substance varies depending on the properties to be calculated and the computational method. For example, density functional theory calculations use approximately several hundred atoms. Known software for performing such calculations includes VASP (Vienna Ab initio Simulation Package), which is commercial software compatible with density functional theory calculations, molecular dynamics calculations, etc., and CP2K, which is open-source software compatible with density functional theory calculations, molecular dynamics calculations, etc. In order to use such software for calculating the properties of a substance, it is necessary to create or obtain structural data for the substance.

[0003] As a technology related to a method for obtaining structural data, for example, a search system including a physical property search unit and a user interface has been proposed. In this system, the physical property search unit performs a physical property search based on search conditions and outputs the search results. The user interface includes a mechanism estimation unit and a related factor identification unit. The mechanism estimation unit estimates candidate mechanisms that manifest the function represented by the input specifications, and the related factor identification unit identifies physical property parameters that contribute to at least one of the estimated candidate mechanisms. The estimated mechanism or the identified physical property parameters are reflected in the search conditions. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2020-64337 Summary of the Invention [Problem to be solved by the invention]

[0005] To create or obtain structural data for a substance, for example, one must search for and download structural data suitable for the purpose from a public database. Alternatively, one must generate new structural data using software that generates structural data. Furthermore, if necessary, one must use software that edits structural data to edit the searched or generated structural data into a form suitable for calculations. Because each of these operations requires dedicated software, the time and effort required to obtain the desired structural data is significant. Furthermore, specialized skills are required to operate these software programs.

[0006] In one aspect, the disclosed technology aims to support the creation of substance structure data with simple operations that do not require specialized skills and reduce work costs. [Means for solving the problem]

[0007] In one aspect, the disclosed technology receives a character string related to the creation of substance structure data from a user, selects an operation type from the received character string using a classifier predefined to select an operation type from the character string, and creates and outputs structure data using a module corresponding to the operation type selected by the classifier from among multiple modules predefined for each operation type, and the received character string. [Effects of the Invention]

[0008] As one aspect, the present invention has an effect of providing support for creating structure data of substances with simple operations that do not require specialized skills and reduce work costs. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a block diagram showing a schematic configuration of a structure data creation support system; [Figure 2] FIG. 10 is a diagram illustrating an example of structure data. [Figure 3] FIG. 2 is a functional block diagram of the structure data creation support device. [Figure 4] FIG. 10 is a diagram showing an example of a reception display screen. [Figure 5] FIG. 10 is a diagram for explaining modules according to operation types. [Figure 6] FIG. 10 is a diagram illustrating an example of program code generated by a generative model. [Figure 7] FIG. 10 is a diagram showing an example of a reception display screen on which structures 1 and 2 are displayed. [Figure 8] FIG. 10 is a diagram showing an example of a reception display screen on which structures 1 to 3 are displayed. [Figure 9] FIG. 2 is a block diagram showing a schematic configuration of a computer that functions as a structure data creation support device. [Figure 10] 10 is a flowchart illustrating an example of a structure data creation support process. [Figure 11] 10 is a flowchart showing an example of an editing process (1). [Figure 12] 10 is a flowchart showing an example of an editing process (2). [Figure 13] 10 is a flowchart showing an example of a generation process (1). [Figure 14] 10 is a flowchart showing an example of a generation process (2). [Figure 15] 10 is a flowchart showing an example of search processing (1). [Figure 16] 10 is a flowchart showing an example of search processing (2). DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, an example of an embodiment of the disclosed technology will be described with reference to the drawings.

[0011] As shown in Fig. 1, the structured data creation support system 100 according to this embodiment includes a structured data creation support device 10, a user terminal 70, and an external device 80. The structured data creation support device 10, the user terminal 70, and the external device 80 are connected via a network. Note that the number of user terminals 70 and external devices 80 is not limited to the example in Fig. 1, and the structured data creation support system 100 may include a plurality of user terminals 70 and a plurality of external devices 80.

[0012] The user terminal 70 is an information processing terminal used by a user who uses the services provided by the structure data creation support system 100. The user terminal 70 includes a display device, an input device, a communication function, etc. The display device and the input device may be integrated into a touch panel display. The user terminal 70 may be, for example, a personal computer, a tablet terminal, a smartphone, etc.

[0013] The external device 80 is an information processing device such as a server device or a personal computer. A structure DB (database) 82 is stored in the external device 80. The structure DB 82 is a public database in which structural data of multiple substances is stored in a predetermined format, such as the Materials Project. The predetermined format may be, for example, the XYZ format as shown in the upper diagram of FIG. 2. The XYZ format is a format for structural data that includes the number of atoms, comments, and the three-dimensional coordinates of each atomic type. Note that the comments are not information used to calculate the properties of a substance, but rather information such as the name and description of the substance represented by the structural data. The lower diagram of FIG. 2 shows an example of structural data of a hydrogen molecule written in the XYZ format.

[0014] 3, the structure data creation support device 10 functionally includes a receiving unit 12, a classifying unit 14, and an output unit 16. In addition, a predetermined storage area of ​​the structure data creation support device 10 stores a classifier 22, a module DB 24, and a structure DB 28.

[0015] The receiving unit 12 receives a character string related to the creation of structure data of a substance from a user. Specifically, the receiving unit 12 displays a user interface on the display device of the user terminal 70 and receives a character string input by the user via the user interface.

[0016] FIG. 4 shows an example of a reception display screen 30 serving as a user interface. In the example of FIG. 4, the reception display screen 30 includes a character string input area 32, an execute button 34, and a display area 36. The character string input area 32 is an area in which a character string can be input as text data in a natural language. When the execute button 34 is selected, the character string input in the character string input area 32 is transmitted from the user terminal 70 to the structured data creation support device 10. The reception unit 12 receives the character string transmitted from the user terminal 70. The reception unit 12 passes the received character string to the classification unit 14.

[0017] Furthermore, the reception unit 12 may provide an auxiliary information input area (not shown) on the reception display screen 30 and receive auxiliary information related to the creation of structured data together with the character string. The auxiliary information may be, for example, information specifying a database to be used when searching for structured data, which will be described later. When receiving auxiliary information, the reception unit 12 passes the received auxiliary information together with the character string to the classification unit 14.

[0018] As will be described in detail later, the display area 36 of the reception display screen 30 displays an image of the structure of the substance representing the structure data created by the output unit 16.

[0019] The classification unit 14 selects an operation type from the received character string using a classifier 22 that is predefined to select an operation type from a character string. The operation type is the type of operation performed on the structural data of the substance. In this embodiment, a case will be described in which there are three types of operation types: "generation" of structural data corresponding to the character string, "editing" of structural data to be edited, and "searching" of structural data corresponding to the character string. However, the operation types are not limited to these three types.

[0020] The classifier 22 selects the above operation type regardless of the specific properties or functions of the substance. For example, even if the character string contains a description of a desired function for the substance indicated by the structural data to be created, such as being light-transmitting or conductive, the classifier 22 will not select that function.

[0021] The classifier 22 may be, for example, a machine learning model such as a neural network that is pre-trained to output an operation type corresponding to an input character string when the character string is input. In this case, the classifier 22 calculates, for example, the probability that the operation type corresponding to the input character string is "create," the probability that the operation type is "edit," and the probability that the operation type is "search," and outputs the operation type with the highest probability. Note that if none of the probabilities of the operation types calculated by the classifier 22 exceed a predetermined value, the classifier 22 outputs a message indicating that there is no corresponding operation type.

[0022] Alternatively, the classifier 22 may use a word dictionary predetermined for each operation type for pattern matching with words included in the received character string. If the character string contains a word that is unique to a word defined in the word dictionary for any of the operation types, the classifier 22 outputs the operation type. Note that if multiple words included in the character string match words defined in word dictionaries for different operation types, the classifier 22 may select an operation type to output according to a predetermined rule. The selection rule may be, for example, the operation type with the most matching words, or the operation type with a matching word of a predetermined high priority. Furthermore, if a word included in the character string does not match a word defined in the word dictionary for any of the operation types, the classifier 22 outputs a message indicating that there is no corresponding operation type.

[0023] Specifically, the classification unit 14 inputs the received character string into the classifier 22 and obtains a selection result indicating that there is no corresponding operation type or an operation type corresponding to the character string output from the classifier 22. For example, the classification unit 14 obtains the selection result "generation" for the character string "generate hydrogen molecules." For example, the classification unit 14 obtains the selection result "edit" for the character string "change the interatomic distance of the generated hydrogen molecules to 1 Å." For example, the classification unit 14 obtains the selection result "search" for the character string "search for materials containing carbon with a band gap of 1 eV or more." The classification unit 14 passes the obtained selection result, the character string, and auxiliary information, if accepted, to the output unit 16.

[0024] The output unit 16 creates and outputs structured data using a module and character string corresponding to the operation type selected by the classifier 22 from a plurality of modules prepared in advance for each operation type.

[0025] The plurality of modules are stored in the module DB 24. In this embodiment, as shown in Fig. 5, the module DB 24 includes a generation module 24A, an editing module 24B, and a search module 24C.

[0026] The generation module 24A is an example of a "first module" of the disclosed technology, and is a module that generates structured data. Specifically, the generation module 24A uses a generation model 26A to generate structured data represented by an accepted character string. The generation model 26A is a machine learning model such as a large language model (LLM) that is trained to generate and output structured data corresponding to an input character string. For example, the generation model 26A outputs structured data such as that shown in the lower diagram of FIG. 2 in response to an input character string "generate hydrogen molecules."

[0027] Alternatively, the generation module 24A may use the generation model 26A to generate program code for generating structure data represented by the character string, and execute the program based on the generated program code to generate the structure data. The program for generating the structure data may be, for example, ASE (Atomic Simulation Environment), which is open-source software that generates and edits structure data using the Python language. For example, the generation model 26A outputs program code such as that shown in FIG. 6 in response to an input of the character string "generate hydrogen molecule." The example in FIG. 6 is an example of program code for generating structure data of a hydrogen molecule using ASE 3.22.1. By executing this program code, the generation module 24A generates and outputs structure data such as that shown in the lower diagram of FIG. 2.

[0028] The editing module 24B is an example of a "second module" of the disclosed technology, and is a module that edits structured data. Specifically, the editing module 24B identifies structured data to be edited. The editing module 24B may identify structured data generated or searched by the generation module 24A or a search module 24C (described later) as the structured data to be edited.

[0029] The editing module 24B uses the editing model 26B to perform editing on the structural data to be edited, as represented by the received character string. The editing model 26B is a machine learning model such as an LLM that is trained to output structural data in which the structural data to be edited has been edited corresponding to the character string when the structural data to be edited and the character string are input. For example, as shown in the upper diagram of FIG. 7, it is assumed that the structural data generated by the generation module 24A (A in FIG. 7) is input to the editing model 26B as the structural data to be edited. It is also assumed that the character string "change the interatomic distance of the generated hydrogen molecule to 1 Å" is input to the editing model 26B. In this case, the editing model 26B outputs structural data as shown in B in FIG. 7.

[0030] Alternatively, the editing module 24B uses the editing model 26B to generate program code for performing editing on the structured data to be edited, as represented by the accepted character string. The editing module 24B may then perform editing on the structured data to be edited by inputting the structured data to be edited and executing a program based on the generated program code. The program for performing editing may be the above-mentioned ASE or the like.

[0031] The search module 24C is an example of a "third module" of the disclosed technology, and is a module that searches for structure data from a database. Specifically, the search module 24C generates structure data represented by an accepted character string using a search model 26C. In this case, the search model 26C is a machine learning model such as an LLM in which a structure DB such as the structure DB 28 or the structure DB 82 is incorporated as a parameter. The search model 26C is trained to output a search result obtained by searching the structure DB for structure data corresponding to an input character string.

[0032] Alternatively, the search module 24C uses the search model 26C to generate a query for searching the structure DB 28 or the structure DB 82 for structure data represented by the received character string, and searches for the structure data from the structure DB 28 or the structure DB 82 based on the generated query. In this case, the search model 26C does not incorporate a structure DB, and may be a machine learning model such as an LLM that is trained to output a query for searching for structure data corresponding to an input character string.

[0033] The query output by the search model 26C may be, for example, a SELECT statement, which is a command statement in the well-known database language SQL, for retrieving data from a database. More specifically, the structure DB 28 and the structure DB 82 are assumed to be SQL databases, with the composition formula represented in a "composition" column and the band gap represented in a "bandgap" column. Also, assume that the string is "Search for materials with a band gap of 1 eV or more and containing carbon." In this case, the search model 26C outputs a query such as "SELECT * FROM database WHERE bandgap >= 1 AND composition LIKE '%C%;'." The search module 24C uses this query to search the structure DB 28 or the structure DB 82 for structural data corresponding to materials with a band gap of 1 eV or more and containing carbon, and outputs the retrieved structural data.

[0034] If the auxiliary information received together with the character string specifies a database to be searched, the search module 24C searches the specified database for structured data.

[0035] The output unit 16 stores the structure data output from the module corresponding to the operation type selected by the classifier 22 in the structure DB 28. The structure DB 28 may store only the structure data created by the output unit 16, or may store structure data similar to that stored in the structure DB 82 stored in the external device 80. Note that the structure DB 28 is not limited to being stored in a predetermined storage area of ​​the structure data creation support device 10, and may also be stored in an external device connected via a network to the structure data creation support device 10.

[0036] Furthermore, the output unit 16 visualizes the structure data output from the module corresponding to the operation type selected by the classifier 22, for example, using a molecular structure model, and displays the visualized structure in the display area 36 of the reception display screen 30. At this time, the output unit 16 assigns a number to the displayed structure. When a new character string is accepted and structure data is created while a structure is displayed in the display area 36, ​​the output unit 16 assigns a new number to the newly created structure and adds it to the display area 36 for display.

[0037] The upper diagram in Figure 7 shows an example in which structure data for Structure 1 is created for the character string "Generate hydrogen molecules," and the created structure is displayed as Structure 1. In this state, if the character string "Change the interatomic distance of the generated hydrogen molecules to 1 Å" is further received, structure data edited so that the interatomic distance of the hydrogen molecules in Structure 1 is 1 Å is created and displayed as Structure 2, as shown in the lower diagram in Figure 7.

[0038] Furthermore, if the string "Insert oxygen molecules between the generated hydrogen molecules to turn them into water" is accepted from the state shown in the lower part of Figure 7, edited structure data will be created to insert oxygen molecules between the hydrogen molecules of Structure 2 to turn them into water, as shown in Figure 8, and displayed as Structure 3.

[0039] In this way, structured data can be created simply by inputting the character strings related to the creation of structured data, without requiring specialized skills such as the use of specialized software. Furthermore, by repeatedly inputting the character strings while checking the created structure, the desired structured data can be obtained.

[0040] The structure data creation support device 10 may be realized by, for example, a computer 40 shown in Fig. 9. The computer 40 includes a CPU (Central Processing Unit) 41, a GPU (Graphics Processing Unit) 42, a memory 43 as a temporary storage area, and a non-volatile storage device 44. The computer 40 also includes an input / output device 45 such as an input device and a display device, and an R / W (Read / Write) device 46 that controls reading and writing of data from and to a storage medium 49. The computer 40 also includes a communication I / F (Interface) 47 that is connected to a network such as the Internet. The CPU 41, GPU 42, memory 43, storage device 44, input / output device 45, R / W device 46, and communication I / F 47 are connected to one another via a bus 48.

[0041] The storage device 44 is, for example, a hard disk drive (HDD), a solid state drive (SSD), a flash memory, etc. The storage device 44 serving as a storage medium stores a structured data creation support program 50 for causing the computer 40 to function as the structured data creation support device 10. The structured data creation support program 50 has a reception process control command 52, a classification process control command 54, and an output process control command 56. The storage device 44 also has an information storage area 60 in which information constituting each of the classifier 22, the module DB 24, and the structure DB 28 is stored.

[0042] The CPU 41 reads the structured data creation support program 50 from the storage device 44, loads it in the memory 43, and sequentially executes the control instructions of the structured data creation support program 50. The CPU 41 operates as the reception unit 12 shown in FIG. 3 by executing the reception process control instruction 52. The CPU 41 operates as the classification unit 14 shown in FIG. 3 by executing the classification process control instruction 54. The CPU 41 operates as the output unit 16 shown in FIG. 3 by executing the output process control instruction 56. The CPU 41 also reads information from the information storage area 60 and loads the classifier 22, the module DB 24, and the structure DB 28 into the memory 43. As a result, the computer 40 that has executed the structured data creation support program 50 functions as the structured data creation support device 10. The CPU 41 that executes the program is hardware. A portion of the program may be executed by the GPU 42.

[0043] The functions realized by the structure data creation support program 50 may be realized by, for example, a semiconductor integrated circuit, more specifically, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or the like.

[0044] Next, the operation of the structured data creation support system 100 according to this embodiment will be described. When an application for using the service provided by the structured data creation support system 100 is launched in the user terminal 70, the structured data creation support process shown in Fig. 10 is executed in the structured data creation support device 10. The structured data creation support process is an example of a structured data creation support method of the disclosed technology.

[0045] In step S10, the reception unit 12 displays the reception display screen 30, which serves as a user interface, on the display device of the user terminal 70, and receives a character string input into the character string input area 32 of the reception display screen 30. Next, in step S12, the classification unit 14 uses the classifier 22 to select an operation type from the character string received in step S10.

[0046] Next, in step S14, the output unit 16 determines whether the operation type selected in step S12 is "edit." If it is "edit," the process proceeds to step S20, and if it is not "edit," the process proceeds to step S16.

[0047] In step S16, the output unit 16 determines whether the operation type selected in step S12 is "generation." If it is "generation," the process proceeds to step S40, and if it is not "generation," the process proceeds to step S18.

[0048] In step S18, the output unit 16 determines whether the operation type selected in step S12 is "search." If it is "search," the process proceeds to step S50. If it is not "search," that is, if it does not correspond to any of the operation types, the process proceeds to step S64.

[0049] In step S20, the output unit 16 causes the editing module 24B to identify the structure data to be edited. Next, in step S30, the editing process is executed. Here, two examples of the editing process will be described. First, the editing process (1) shown in Fig. 11 will be described.

[0050] In step S32, the editing module 24B inputs the structured data and character string to be edited to the editing model 26B, causes the editing model 26B to output the structured data, and returns to the structured data creation support process (FIG. 10).

[0051] Next, the editing process (2) shown in FIG. 12 will be described.

[0052] In step S36, the editing module 24B uses the editing model 26B to generate and output program code for performing editing represented by the accepted character string on the structured data to be edited.

[0053] Next, in step S38, the editing module 24B inputs the structure data to be edited and executes the program based on the generated program code, thereby outputting the structure data edited for the structure data to be edited. Then, the process returns to the structure data creation support process (FIG. 10).

[0054] In step S40, the generation process is executed. Two examples of the generation process will now be described. First, the generation process (1) shown in FIG. 13 will be described.

[0055] In step S42, the generation module 24A inputs a character string to the generation model 26A, causes the generation model 26A to output structured data, and returns to the structured data creation support process (FIG. 10).

[0056] Next, the generation process (2) shown in FIG. 14 will be described.

[0057] In step S46, the generation module 24A uses the generation model 26A to generate and output program code for generating structure data represented by the character string.

[0058] Next, in step S48, the generation module 24A generates and outputs structured data by executing a program using software that generates structured data based on the generated program code, and the process returns to the structured data creation support process (FIG. 10).

[0059] In step S50, a search process is executed. Two examples of the search process will now be described. First, the search process (1) shown in FIG. 15 will be described.

[0060] In step S52, the search module 24C inputs the character string to the search model 26C in which the structure DB is incorporated as a parameter, causes the search model 26C to output structure data, and returns to the structure data creation support process (FIG. 10).

[0061] Next, the search process (2) shown in FIG. 16 will be described.

[0062] In step S56, the search module 24C uses the search model 26C to generate a query for searching the structure DB 28 or the structure DB 82 for structure data represented by the received character string.

[0063] Next, in step S58, the search module 24C searches for and outputs structure data from the structure DB 28 or the structure DB 82 based on the generated query, and the process returns to the structure data creation support process (FIG. 10).

[0064] Next, in step S60, the output unit 16 stores the structural data output in step S30, S40, or S50 in the structural DB 28. Next, in step S62, the output unit 16 visualizes the structural data output in step S30, S40, or S50 using, for example, a molecular structure model, and displays the visualized structure in the display area 36 of the reception display screen 30.

[0065] On the other hand, in step S64, the output unit 16 displays an error message such as prompting the user to input the character string again, since none of the operation types can be selected from the received character string.

[0066] Next, in step S66, the reception unit 12 determines whether to terminate the creation of the structured data by determining whether a command indicating the termination of the process of creating the structured data has been received, whether the reception display screen 30 has been closed, whether the application has been terminated, etc. If the creation of the structured data is not to be terminated, the process returns to step S10; if it is to be terminated, the process of supporting the creation of the structured data terminates.

[0067] For each of the generation process, editing process, and search process, example (1) or example (2) can be selected depending on feasibility. In the case of (1), one step is eliminated before the structured data is output, making it possible to output the structured data easily. Furthermore, problems such as the generation of programs not intended by this system by generative models 26A, 26B, and 26C do not occur. In the case of (2), more accurate structured data can be output by using dedicated software for generating or editing the structured data, searching from a database independent of the machine learning model, etc.

[0068] As described above, according to the structure data creation support system of this embodiment, the structure data creation support device receives from a user a character string related to the creation of structure data for a substance. The structure data creation support device then selects an operation type from the received character string using a classifier predefined to select an operation type from the character string. The structure data creation support device then creates and outputs structure data using a module corresponding to the operation type selected by the classifier from among multiple modules predefined for each operation type, and the received character string. This helps create structure data for substances with simple operations that do not require specialized skills and at reduced operating costs. This allows researchers in fields such as materials engineering and materials science to create structure data for substances suited to their purposes in a short amount of time.

[0069] In the above embodiment, the modules used to create structured data are described as a generation module, an editing module, and a search module, but this is not limiting. For example, a filter module that filters structured data may be included. The filter module is a module that filters structured data created in the past to output structured data corresponding to a received character string. The filter module is an example of a "fourth module" of the disclosed technology.

[0070] Furthermore, the number of structural data output from each module is not limited to one, and multiple structural data may be output. In this case, the multiple structural data may be output individually, or structural data obtained by combining multiple structural data under conditions specified by the user may be output. The specified conditions may indicate the positional relationship of bonds between structural data, such as specifying the atoms bonded between the structural data.

[0071] In the above embodiment, the structure data created by the structure data creation support device is displayed via a user interface of a user terminal, but this is not limiting. For example, the process of creating the structure data may be executed on the user terminal side. In this case, the user terminal side may be provided with each functional configuration, module, and structure DB of the structure data creation support device.

[0072] In the above embodiment, the structure data creation support program is stored (installed) in advance in a storage device, but this is not limiting. The program according to the disclosed technology may be provided in a form stored in a storage medium such as a CD-ROM, a DVD-ROM, or a USB memory.

[0073] The following additional notes are provided regarding the above-described embodiments.

[0074] (Appendix 1) Accepts character strings from the user regarding the creation of substance structure data, selecting an operation type from the received character string using a classifier predefined to select an operation type from the character string; creating and outputting structured data using a module corresponding to the operation type selected by the classifier from among a plurality of modules prepared in advance for each operation type and the received character string; A structured data creation support program for causing a computer to execute processing including the above.

[0075] (Appendix 2) The structured data creation support program according to Appendix 1, wherein the plurality of modules include a first module that generates structured data, a second module that edits the structured data, and a third module that searches for structured data from a database.

[0076] (Appendix 3) A structured data creation support program as described in Appendix 2, in which each of the plurality of modules creates the structured data using a machine learning model that outputs structured data when the accepted string is input.

[0077] (Appendix 4) The structured data creation support program according to claim 3, wherein the first module is a module that uses a large-scale language model to generate structured data represented by the received character string, or a module that uses a large-scale language model to generate program code for generating structured data represented by the received character string, and generates structured data by executing a program based on the program code.

[0078] (Appendix 5) The structured data creation support program described in Appendix 3, wherein the second module is a module that identifies structured data to be edited and uses a large-scale language model to perform editing on the structured data to be edited in accordance with the received character string, or a module that uses a large-scale language model to generate program code for performing editing on the structured data to be edited in accordance with the received character string, and executes a program based on the program code to perform editing on the structured data to be edited.

[0079] (Appendix 6) 6. A structured data creation support program according to claim 5, which includes identifying the structured data generated or searched by the first module or the third module as the structured data to be edited.

[0080] (Appendix 7) The structured data creation support program according to Appendix 3, wherein the third module is a module that uses a large-scale language model incorporating the database to generate structured data represented by the received character string, or a module that uses a large-scale language model to generate a query for searching the database for structured data represented by the received character string, and searches the database for structured data based on the generated query.

[0081] (Appendix 8) the plurality of modules further includes a fourth module for filtering structure data; The structured data creation support program according to Appendix 2, wherein the fourth module is a module that filters and outputs structured data corresponding to the received character string from previously created structured data.

[0082] (Appendix 9) A structured data creation support program according to any one of appendices 1 to 8, which, when multiple structured data are created by a module corresponding to the operation type selected by the classifier, outputs each of the multiple structured data, or outputs structured data combining the multiple structured data under conditions specified by the user.

[0083] (Appendix 10) The structured data creation support program according to any one of Supplementary Note 1 to Supplementary Note 9, wherein the classifier is a machine learning model that is pre-trained to output the operation type corresponding to the input character string when the character string is input, or a word dictionary that is pre-defined for each operation type for pattern matching with words included in the received character string.

[0084] (Appendix 11) 11. The structured data creation assistance program according to claim 1, which receives auxiliary information related to creation of the structured data together with the character string.

[0085] (Appendix 12) Accepts character strings from the user regarding the creation of substance structure data, selecting an operation type from the received character string using a classifier predefined to select an operation type from the character string; creating and outputting structured data using a module corresponding to the operation type selected by the classifier from among a plurality of modules prepared in advance for each operation type and the received character string; A structured data creation support method in which a computer executes a process including the steps of:

[0086] (Appendix 13) The structured data creation support method according to claim 11, wherein the plurality of modules include a first module that generates structured data, a second module that edits the structured data, and a third module that searches for structured data from a database.

[0087] (Appendix 14) A structured data creation support method as described in Appendix 13, in which each of the multiple modules creates the structured data using a machine learning model that outputs structured data when the accepted string is input.

[0088] (Appendix 15) The structured data creation support method described in Appendix 14, wherein the first module is a module that uses a large-scale language model to generate structured data represented by the accepted character string, or a module that uses a large-scale language model to generate program code for generating structured data represented by the accepted character string, and generates structured data by executing a program based on the program code.

[0089] (Appendix 16) The structured data creation support method described in Appendix 14, wherein the second module is a module that identifies structured data to be edited and uses a large-scale language model to perform editing on the structured data to be edited in accordance with the received character string, or a module that uses a large-scale language model to generate program code for performing editing on the structured data to be edited in accordance with the received character string, and executes a program based on the program code to perform editing on the structured data to be edited.

[0090] (Appendix 17) A structured data creation support method according to appendix 16, including identifying the structured data generated or searched by the first module or the third module as the structured data to be edited.

[0091] (Appendix 18) The structured data creation support method described in Appendix 14, wherein the third module is a module that uses a large-scale language model incorporating the database to generate structured data represented by the received character string, or a module that uses a large-scale language model to generate a query for searching the database for structured data represented by the received character string, and searches the database for structured data based on the generated query.

[0092] (Appendix 19) the plurality of modules further includes a fourth module for filtering structure data; The structured data creation support method according to claim 13, wherein the fourth module is a module that filters and outputs structured data corresponding to the received character string from structured data created in the past.

[0093] (Appendix 20) a receiving unit that receives a character string related to the creation of substance structure data from a user; a classifier that selects an operation type from the received character string using a classifier that is predefined to select an operation type from the character string; an output unit that creates and outputs structured data using a module corresponding to the operation type selected by the classifier from among a plurality of modules prepared in advance for each operation type and the received character string; A structure data creation support device including: [Explanation of symbols]

[0094] 10. Structural data creation support device 12 Reception 14 Classification section 16 Output section 22 Classifier 24 Module DB 24A Generation Module 24B Editing Module 24C Search Module 26A Generative Model 26B Editing Model 26C Search Model 28, 82 Structure DB 30 Reception display screen 32 String input area 34 Execute button 36 Display area 40 Computer 41 CPU 42 GPU 43 Memory 44 Storage device 45 Input / Output Devices 46 R / W device 47 Communication I / F 48 Bus 49 Storage medium 50 Structural data creation support program 52 Reception process control instructions 54 Classification Process Control Instructions 56 Output Process Control Instructions 60 Information storage area 70 User terminals 80 External equipment 100 Structural Data Creation Support System

Claims

1. Accepts character strings from the user regarding the creation of substance structure data, selecting an operation type from the received character string using a classifier predefined to select an operation type from the character string; creating and outputting structured data using a module corresponding to the operation type selected by the classifier from among a plurality of modules prepared in advance for each operation type and the received character string; A structured data creation support program for causing a computer to execute processing including the above.

2. 2. The structured data creation support program according to claim 1, wherein the plurality of modules include a first module for generating structured data, a second module for editing the structured data, and a third module for searching for the structured data from a database.

3. The structured data creation support program according to claim 2 , wherein each of the plurality of modules creates the structured data using a machine learning model that outputs structured data when the accepted character string is input.

4. 4. The structured data creation support program according to claim 3, wherein the first module is a module that uses a large-scale language model to generate structured data represented by the received character string, or a module that uses a large-scale language model to generate program code for generating structured data represented by the received character string, and generates structured data by executing a program based on the program code.

5. The structured data creation support program according to claim 3, wherein the second module is a module that identifies the structured data to be edited and uses a large-scale language model to perform editing on the structured data to be edited in accordance with the received character string, or a module that uses a large-scale language model to generate program code for performing editing on the structured data to be edited in accordance with the received character string, and executes a program based on the program code to perform editing on the structured data to be edited.

6. 6. The structured data creation support program according to claim 5, further comprising specifying the structured data generated or searched by the first module or the third module as the structured data to be edited.

7. 4. The structured data creation support program according to claim 3, wherein the third module is a module that generates structured data represented by the received character string using a large-scale language model incorporating the database, or a module that generates a query for searching the database for structured data represented by the received character string using a large-scale language model, and searches the database for structured data based on the generated query.

8. the plurality of modules further includes a fourth module for filtering structure data; 3. The structured data creation support program according to claim 2, wherein the fourth module is a module that filters structured data created in the past to find structured data corresponding to the received character string and outputs the filtered structured data.

9. A structured data creation support program according to any one of claims 1 to 8, wherein when multiple structured data are created by a module corresponding to the operation type selected by the classifier, the program outputs each of the multiple structured data, or outputs structured data combining the multiple structured data under conditions specified by the user.

10. The structured data creation support program according to any one of claims 1 to 8, wherein the classifier is a machine learning model that is pre-trained to output the operation type corresponding to the input character string when the character string is input, or a word dictionary that is pre-defined for each operation type for pattern matching with words contained in the received character string.

11. 9. The structured data creation support program according to claim 1, further comprising: receiving auxiliary information relating to the creation of the structured data together with the character string.

12. Accepts character strings from the user regarding the creation of substance structure data, selecting an operation type from the received character string using a classifier predefined to select an operation type from the character string; creating and outputting structured data using a module corresponding to the operation type selected by the classifier from among a plurality of modules prepared in advance for each operation type and the received character string; A structured data creation support method in which a computer executes a process including the steps of:

13. a receiving unit that receives a character string related to the creation of substance structure data from a user; a classifier that selects an operation type from the received character string using a classifier that is predefined to select an operation type from the character string; an output unit that creates and outputs structured data using a module corresponding to the operation type selected by the classifier from among a plurality of modules prepared in advance for each operation type and the received character string; A structure data creation support device including:

Images