Apparatus and method for generating learning data by object diversification of chemical structural formulas

Abstract

According to one embodiment of the present invention, a training data generation device includes a processor and at least one memory electrically connected to the processor. The at least one memory stores at least one feature variable that determines a feature of an object representing a chemical structural formula, and a first setting value set that is predetermined for each of the at least one feature variable. The processor loads first chemical formula data including at least one node information and at least one edge information and the at least one feature variable, sets a setting value of the at least one feature variable to the first setting value set, generates a first chemical structural formula based on the first chemical formula data using an object set to and applied with the first setting value set, obtains a second setting value set in which the setting value of the corresponding feature variable is changed based on an input that changes the setting value of the at least one feature variable, generates a second chemical structural formula using the object set to and applied with the second setting value set, and generates training data including the generated second chemical structural formula.

Description

[Technical Field]

[0001] The present invention relates to a training data generation device and method that generates training data by diversifying objects of chemical structural formulas, and more particularly to a training data generation device and method that can mass-produce chemical structural formulas by diversifying setting values ​​of objects that constitute chemical structural formulas based on chemical formula data including node and edge information. [Background technology]

[0002] When developing an AI model that recognizes chemical structural formulas, generating training data solely using existing chemical structural formula generation programs cannot reflect the diversity of actual chemical structural formula data, making it difficult to expect high performance from the AI ​​model.

[0003] Existing chemical structure generation programs can only generate chemical structure formulas in a fixed format, making it difficult to diversify the thickness, color, and spacing of the lines representing bonds in chemical structure formulas, or the font, size, and color of the characters representing atoms.

[0004] Furthermore, improving the performance of an AI model requires a large amount of training data, but existing generation programs can only generate one image at a time, so it takes a lot of time to process the data to generate additional data.

[0005] Furthermore, while the learning of AI models using vision technology requires coordinate information and class information for atoms and bond regions in chemical structure training data, existing generation programs were unable to obtain this information, which meant that people had to manually obtain the coordinate information and class information for atoms and bond regions, creating a challenge. Summary of the Invention [Problem to be solved by the invention]

[0006] An object of the present invention is to provide a training data generation device and method capable of generating a large amount of training data in various formats for training an artificial intelligence model that recognizes chemical structural formulas.

[0007] The problems to be solved by the present invention are not limited to those mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the following description. [Means for solving the problem]

[0008] According to one aspect of the present invention, there is provided a training data generation device that includes a processor and at least one memory electrically connected to the processor. The at least one memory stores at least one feature variable that determines a feature of an object representing a chemical structural formula, and a first setting value set that is predetermined for each of the at least one feature variable. The processor loads first chemical formula data including at least one node information item and at least one edge information item and the at least one feature variable, sets a setting value of the at least one feature variable to the first set, generates a first chemical structural formula based on the first chemical formula data using an object set and applied with the first setting value set, obtains a second setting value set in which the setting value of the corresponding feature variable is changed in response to an input that changes the setting value of the at least one feature variable, generates a second chemical structural formula using the object set and applied with the second setting value set, and generates training data including the generated second chemical structural formula.

[0009] As another embodiment of the present invention, there is provided a method performed by a training data generation device, wherein the training data generation device may include the steps of: storing at least one feature variable that determines a feature of an object representing a chemical structural formula, and a first setting value set predetermined for each of the at least one feature variable, and loading first chemical formula data including at least one node information and at least one edge information, and the at least one feature variable; setting a setting value of the at least one feature variable to the first setting value set; generating a first chemical structural formula based on the first chemical formula data using an object set and applied with the first setting value set; acquiring a second setting value set in which the setting value of the at least one feature variable is changed by an input that changes the setting value of the at least one feature variable; generating a second chemical structural formula using an object set and applied with the second setting value set; and generating training data including the generated second chemical structural formula.

[0010] In addition, a computer program stored on a computer-readable recording medium may be provided in combination with a hardware device to execute the training data generation method of claim 9. [Effects of the Invention]

[0011] According to the present invention, it is possible to generate chemical structure training data in a variety of formats, not just a fixed format, and to generate a large amount of chemical structure training data from one chemical formula data.

[0012] In addition, a set value can be assigned to each feature variable of the object constituting the chemical structural formula generated by the training data generation device of the present invention, which facilitates annotation using the set value assigned to each object, reduces the error rate in annotation work, and reduces the cost of verifying whether the results of annotation work are performed correctly.

[0013] Therefore, the present invention provides the effect of improving the performance of an artificial intelligence model that recognizes chemical structural formulas by utilizing chemical structural formulas generated by the training data generation device of the present invention.

[0014] The effects of the present invention are not limited to those mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the following description. [Brief explanation of the drawings]

[0015] [Figure 1] 1 is a block diagram of a training data generation device according to an embodiment of the present invention. [Figure 2] 1 is a diagram illustrating a process of generating a chemical structural formula based on chemical formula data according to an embodiment of the present invention. [Figure 3] (a) shows a first chemical structural formula generated using predetermined settings, and (b) shows a second chemical structural formula in which the settings of some letters, numbers, and symbols have been changed according to one embodiment of the present invention. [Figure 4] (a) shows a first chemical structural formula generated using predetermined setting values, and (b) shows a second chemical structural formula in which the setting values ​​of some of the figures have been changed according to one embodiment of the present invention. [Figure 5] (a) shows a first chemical structural formula generated using predetermined setting values, and (b) shows a second chemical structural formula in which the setting values ​​of some of the figures have been changed according to one embodiment of the present invention. [Figure 6](a) shows a first chemical structural formula generated using predetermined setting values, and (b) shows a second chemical structural formula in which the setting values ​​of some of the figures have been changed according to one embodiment of the present invention. [Figure 7] (a) shows a first chemical structural formula generated using predetermined setting values, and (b) shows a second chemical structural formula in which the setting values ​​of some of the figures have been changed according to one embodiment of the present invention. [Figure 8] (a) shows a first chemical structural formula generated using predetermined setting values, and (b) shows a second chemical structural formula in which the setting values ​​of some symbols have been changed according to one embodiment of the present invention. [Figure 9] (a) shows a first chemical structural formula generated using predetermined setting values, and (b) and (c) show a second chemical structural formula in which the setting values ​​of some of the figures have been changed according to one embodiment of the present invention. [Figure 10] 1 shows a chemical structure annotated with objects that make up a node according to one embodiment of the present invention. [Figure 11] 1 shows a chemical structure annotated with objects that make up edges according to one embodiment of the present invention. [Figure 12] 1 illustrates the types of objects that make up an edge according to one embodiment of the present invention. [Figure 13] 1 illustrates a chemical structure annotated with bracketed objects according to one embodiment of the present invention. [Figure 14] 1 illustrates a chemical structure annotated with bracket and object intersection points according to one embodiment of the present invention. [Figure 15] FIG. 2 is a block diagram illustrating a method for generating training data according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0016] The advantages and features of the present invention, as well as methods for achieving them, will become more apparent from the following detailed description of the embodiments in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, and can be realized in various different forms. The present embodiments are provided solely for the purpose of complete disclosure of the present invention and to fully convey the scope of the present invention to those skilled in the art to which the present invention pertains. The present invention is defined only by the scope of the claims.

[0017] The terms used in this specification are for the purpose of describing embodiments and are not intended to limit the present invention. In this specification, the singular includes the plural unless otherwise specified. The terms "comprises" and / or "comprising" used in this specification do not exclude the presence or addition of one or more other elements in addition to the elements referenced. The same reference numerals refer to the same elements throughout this specification, and "and / or" includes each and every combination of one or more of the referenced elements. While terms such as "first," "second," etc. are used to describe various elements, it goes without saying that these elements are not limited by these terms. These terms are merely used to distinguish one element from another. Therefore, it goes without saying that a first element referred to below may also be a second element within the technical spirit of the present invention.

[0018] Unless otherwise defined, all terms (including technical and scientific terms) used herein may be used in a manner commonly understood by those skilled in the art to which the present invention pertains. Furthermore, commonly used and predefined terms are not to be interpreted ideally or excessively unless specifically defined otherwise.

[0019] The terms used in this specification will be explained below.

[0020] An object represents a group of elements that make up a chemical structural formula. For example, a chemical structural formula consists of letters, numbers, symbols, and shapes. Specifically, letters represent atoms or compounds that make up the chemical structural formula, numbers represent the number of atoms or charges, symbols represent positive charges (+), negative charges (-), or brackets that represent specific repeating compound groups, and shapes represent single lines, double lines, triple lines, dotted lines, etc. that represent chemical bonds.

[0021] In this specification, a chemical structural formula can be understood as a graph. A graph is a data structure composed of points and lines connecting these points. Here, the points are called nodes, and the lines are called edges. In other words, a chemical structural formula can be understood as a graph in which at least one atom or compound corresponding to a node is connected by at least one bond corresponding to an edge. A chemical structural formula may also include brackets. Brackets indicate groups of atoms and bonds that are repeated in a chemical structural formula.

[0022] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0023] FIG. 1 is a block diagram of a training data generation device according to an embodiment of the present invention.

[0024] The training data generation device 100 may include a processor 110, a memory 120, and a display 130. The processor 110, the memory 120, and the display 130 may be electrically connected to each other.

[0025] The processor 110 can load the first chemical formula data 200, including at least one node information 210 and at least one edge information 220, from the memory 120 or an external device. The processor 110 can also load at least one feature variable 300 from the memory 120 or an external device.

[0026] The processor 110 can set the setting value of each of the loaded at least one feature variable 300 to a predetermined first setting value set 400 .

[0027] The processor 110 can generate a chemical structural formula in which atoms or compounds are bonded by chemical bonds based on at least one node information and at least one edge information included in the chemical formula data. Therefore, the processor 110 can generate a first chemical structural formula based on the first chemical formula data 200 using objects set and applied in a predetermined first setting value set 400.

[0028] The processor 110 acquires the second set of setting values ​​500 or 600 in which the setting value of at least one feature variable 300 has been changed, in response to an input that changes the setting value of the feature variable.

[0029] An input for changing the set value of the feature variable 300 may be input by a user. Alternatively, the set value of the feature variable 300 may be randomly changed by the processor 110. At this time, the set value of each feature variable 300 may be randomly changed within a predetermined range or condition. The processor 110 may link an arbitrary chemical structural formula to the first chemical structural formula.

[0030] This provides the effect of generating a large amount of chemical structure formula learning data having a variety of forms.

[0031] In addition, the set values ​​of the feature variables 300 can be changed by combining user input and processing by the processor 110. For example, when the user selects a specific variable from the feature variables 300, the processor 110 can randomly change the set value of the feature variable within a predetermined range. In addition, when the user inputs the number of chemical structural formulas to be generated, the processor 110 can randomly change any of the feature variables so that the corresponding number of chemical structural formulas are generated. This can provide the effect of enabling the user to generate a large amount of chemical structural formula learning data having a desired form.

[0032] The processor 110 can obtain the second set of setting values ​​500 or 600 in which the setting value of at least one of the feature variables 300 has been changed.

[0033] The processor 110 may generate a second chemical structural formula using an object set and applied to the second setting value set 500 or the second setting value set 600. The processor 110 may generate training data including the generated second chemical structural formula.

[0034] The memory 120 can store at least one feature variable 300 that determines a feature of an object representing a chemical structure formula, and a first set of predetermined values ​​400 for each of the at least one feature variable 300.

[0035] The display 130 can output chemical structural formulas. Specifically, the display 130 can output a first chemical structural formula generated using an object set and applied to the first setting value set 400, and can output a second chemical structural formula generated using an object set and applied to the modified second setting value set 500 or the modified second setting value set 600.

[0036] In addition, the display 130 can output the first chemical formula data 200, the feature variables 300, the predetermined first set of setting values ​​400, and the changed second set of setting values ​​500, 600. The user can change the feature variables 300 of the desired object after checking the first chemical structural formula through the display 130, and can check the changed second chemical structural formula.

[0037] Furthermore, the display 130 may output the second chemical structure formula separately for each time point when the changed second set of setting values ​​is acquired.

[0038] FIG. 2 is a diagram illustrating a process of generating a chemical structural formula based on chemical formula data according to an embodiment of the present invention.

[0039] 2 shows a process of generating various types of chemical structural formulas based on the loaded first chemical formula data 200. Specifically, the process shows a process of setting a value of a feature variable 300 of an object for expressing the first chemical formula data 200 as a chemical structural formula to a predetermined first set of setting values ​​400, acquiring a second set of setting values ​​500 or 600 by inputting a change in the setting value of the feature variable 300, and generating a chemical structural formula using the object set in the second set of setting values ​​500 or 600.

[0040] The first chemical formula data 200 includes at least one piece of node information 210 and at least one piece of edge information 220. For example, the first chemical formula data 200 may include a list of atoms constituting the chemical structural formula, atoms connected by chemical bonds and the multiplicity of the chemical bonds (single, double, triple), coordinates in 2D or 3D space for each atom, the number of atoms, the number of chemical bonds, the total charge of the chemical structural formula, and attributes related to the charge and isomers of each atom. The first chemical formula data 200 may have the format of a MOL file.

[0041] The feature variables 300 may include character feature variables 310 , number feature variables 320 , symbol feature variables 330 and graphic feature variables 340 .

[0042] The character feature variables 310 may include character position, font type, character outline thickness, character size, word size, character spacing, character color, and character slant angle.

[0043] The number feature variables 320 may include the position of the number, the font type, the thickness of the outline of the number, the size of the number, the spacing between the number and the letter, the spacing between the numbers, the color of the number, and the angle of inclination of the number.

[0044] The symbol feature variables 330 may include symbol position, font type, symbol outline thickness, symbol size, spacing between symbol and letter, spacing between symbol and number, spacing between symbols, symbol color, and symbol tilt angle.

[0045] The graphic feature variables 340 may include the position of the graphic, the type, the thickness of the outline of the graphic, the size of the graphic, the spacing between the graphic and the letter, the spacing between the graphic and the number, the spacing between the graphic and the symbol, the spacing between the graphics, the color of the graphic, and the tilt angle of the graphic.

[0046] In this way, by providing a variety of feature variables 300, it is possible to generate a large number of chemical structural formulas from one chemical formula data. Meanwhile, it is understood that any variable that determines the features of the objects that make up the chemical structural formula, other than the variables listed above, can be included in the feature variables 300.

[0047] The first set of settings 400 is a group of predetermined settings. The first set of settings 400 may include settings 410 for character feature variables 310, settings 420 for number feature variables 320, settings 430 for symbol feature variables 330, and settings 440 for graphic feature variables 340. Although only eight settings a through h are shown in the drawing, it is understood that the present invention is not limited to these.

[0048] For example, the setting value a may be a value that determines the font of the characters and may have a setting value of "Arial," and the setting value b may be a value that determines the size of the characters and may have a setting value of "10."

[0049] The second setting value set 500 is a set including sets in which the setting values ​​a, c, e, f, and h of the first setting value set 400 have been changed to setting values ​​a', c', e', f', and h', respectively. On the other hand, the second setting value set 600 is a set including sets in which the setting values ​​a, b, and h of the first setting value set 400 have been changed to setting values ​​a', b', and h', respectively. This means that the remaining setting values ​​that have not been changed are the same as the setting values ​​of the first setting value set 400.

[0050] One chemical structure formula can be generated using the objects set and applied in the second setting value set 500, and another type of chemical structure formula can be generated using the objects set and applied in the second setting value set 600.

[0051] Although FIG. 2 shows that two sets of second setting values ​​500 and 600 are formed, it will be understood that an infinite number of sets of second setting values ​​can be formed depending on the combination of setting values ​​that are changed.

[0052] The present invention will be described below using specific embodiments of chemical structures.

[0053] 3(a) shows a first chemical structural formula generated using predetermined setting values, and FIG. 3(b) shows a second chemical structural formula in which the feature variable setting values ​​for some of the letters, numbers, and symbols have been changed according to an embodiment of the present invention. Specifically, the portions indicated by dotted lines in FIG. 3(b) indicate objects that have been changed compared to FIG. 3(a).

[0054] As described above, the processor 110 can load the first chemical formula data 200 and feature variables 300 including node information and edge information capable of generating the chemical structural formula of FIG. 3(a), and generate the first chemical structural formula as shown in FIG. 3(a) using the objects set and applied to the predetermined first setting value set 400.

[0055] The processor 110 can obtain a second set of setting values ​​500 in which the setting values ​​of at least one feature variable 300 have been changed by input that changes the setting values ​​of the feature variables, and can generate a second chemical structural formula as shown in (b) of Figure 3 using the object set and applied in the second set of setting values ​​500.

[0056] In (a) of Figure 3, SH is a combination of letters, and CH3 is a combination of letters and numbers. Therefore, processor 110 can match character feature variable 310 to SH, and character feature variable 310 and number feature variable 320 to CH3. (b) of Figure 3 shows that the setting values ​​indicating the font and degree of slant among character feature variable 310 have been changed for SH, and the setting values ​​indicating the font and degree of slant among character feature variable 310 and number feature variable 320 have been changed for CH3.

[0057] Also, in Fig. 3(a), the charge (-) of O is a symbol. Therefore, the symbol feature variable 330 can be matched to the charge (-) of O. Fig. 3(b) shows that the setting value indicating the position of the charge (-) of O has been changed.

[0058] Another embodiment will be described with reference to FIGS.

[0059] Figures 4(a), 5(a), 6(a), and 7(a) show a first chemical structural formula generated by a predetermined first set of setting values ​​400, and Figures 4(b), 5(b), 6(b), and 7(b) show a second chemical structural formula in which some of the setting values ​​of the shapes have been changed according to an embodiment of the present invention. Specifically, the portions indicated by dotted lines in Figures 4(b), 5(b), 6(b), and 7(b) indicate objects that have been changed compared to Figures 4(a), 5(a), 6(a), and 7(a), respectively.

[0060] The bonds in the first chemical structures shown in Figures 4(a), 5(a), 6(a), and 7(a) are graphical objects, and therefore the processor 110 can match the graphical feature variables 340 to each bond.

[0061] Figure 4(b) shows that the setting value indicating thickness among the graphic feature variables 340 for the single bond shown in the dotted line has been increased. Figure 5(b) shows that the setting values ​​indicating the inclination angle of the graphic and the spacing between the graphic features have been changed among the graphic feature variables 340 for the double bond shown in the dotted line. Figure 6(b) shows that the setting value indicating length among the graphic feature variables 340 for the single bond shown in the dotted line has been decreased. Figure 7(b) shows that the setting value indicating type among the graphic feature variables 340 for the single bond shown in the dotted line has been changed from a line to a triangle.

[0062] Another embodiment will be described with reference to FIG.

[0063] 8(a) shows a first chemical structural formula generated by a predetermined first set of setting values ​​400, and FIG. 8(b) shows a second chemical structural formula in which the setting values ​​of some symbols have been changed according to an embodiment of the present invention. Specifically, the portions indicated by dotted lines in FIG. 8(b) indicate objects that have been changed compared to FIG. 8(a).

[0064] In Figure 8(a), the charge (+, -) of each atom is a symbol object. Therefore, the processor 110 can match a symbol feature variable 330 to each symbol object. The value of the matched symbol feature variable 330 can be changed from a simple symbol (+, -) representing charge as in Figure 8(a) to a symbol as in Figure 8(b). In addition, the position of the symbol (+, -) can be changed as described above in the description of Figure 3.

[0065] Figures 9(a) to 9(c) show the case where two second chemical structural formulas are generated from one first chemical structural formula. Figure 9(b) shows that the set value indicating the spacing between the figures in the graphic feature variable 340 for the intersection between the bonds in the dotted line portion has been changed to increase. Figure 9(c) shows that the set value indicating the spacing between the lines in the graphic feature variable 340 for the double bond in the dotted line portion has been changed to decrease.

[0066] As described above, by changing the predetermined setting values ​​for the feature variables 300 of letters, numbers, symbols and figures, which are objects that make up chemical structural formulas, it is possible to generate a large number of chemical structural formulas in a variety of forms.

[0067] The second chemical structural formula generated by the processor 110 has a set value set for each of the feature variables 300 of the objects that make up the formula, and this can be utilized to annotate each object with the set value of the feature variable 300. Because each object is annotated, an artificial intelligence model that uses this as learning data can improve its performance.

[0068] Generally, data annotation is performed by creating bounding boxes for objects contained in an image and inputting attribute information for the bounding box-processed objects. This type of annotation is also called data labeling. The resulting dataset is then generated in the form of a JSON (Java Script Object Notation) file.

[0069] FIG. 10 shows a chemical structure annotated with respect to the objects that constitute the node of the second chemical structure.

[0070] The dotted lines in Figure 10 represent objects that make up a node. A feature variable 300 is matched to each object, and a specific setting value is set for each feature variable 300. For example, a setting value of a character feature variable 310 can be set for O, C, and H, a setting value of a number feature variable 320 can be set for the number 2, and a setting value of a graphic feature variable 340 can be set for a node without a letter or number. In this way, objects that make up a node of the second chemical structure formula can be annotated and formed using the setting values ​​set for the feature variables 300 of the corresponding objects.

[0071] Figure 11, Figure 13(a) and Figure 14 show the same second chemical structure.

[0072] Figure 11 shows a chemical structural formula annotated with respect to objects that form edges of the second chemical structural formula, Figure 13 shows a chemical structural formula annotated with respect to objects that form brackets of the second chemical structural formula, and Figure 14 shows a chemical structural formula annotated with points where objects that form brackets of the second chemical structural formula intersect with other objects.

[0073] 11 indicates the objects that make up the edge. For example, a single line segment and a double line segment can be set to a set value of the graphic feature variable 340.

[0074] Figure 12 shows the types of objects that make up an edge according to one embodiment of the present invention. Figure 12(a) shows a single bond, Figure 12(b) shows a double bond, Figure 12(c) shows a triple bond, Figure 12(d) shows a crossed bond, Figure 12(e) and Figure 12(f) show bond forms for three-dimensionally expressing bonds, Figure 12(g) shows a bond represented by a dotted line, Figure 12(h) shows a wavy bond, and Figure 12(i) shows a bond with a resonance structure.

[0075] In this way, the objects constituting the edges of the second chemical structural formula can be annotated and formed with the setting values ​​set for the feature variables 300 of the corresponding objects.

[0076] The dotted lines shown in FIG. 13 represent the objects that make up the brackets.

[0077] The brackets in FIG. 13(a) indicate a chemical structure in which the entire chemical structural formula included in the brackets is repeated n times and linked.

[0078] Figure 13(b) shows a part of Figure 13(a). The brackets in Figure 13(b) indicate that the chemical structural formula contained in the brackets is repeated 12 times and linked together.

[0079] Figure 13(c) shows a part of Figure 13(a). The brackets in Figure 13(c) indicate that the chemical structural formula contained in the brackets is repeated eight times and linked together.

[0080] In this way, the object that constitutes the parentheses of the second chemical structural formula can be annotated and formed with the setting value set for the feature variable 300 of the corresponding object.

[0081] The dotted lines in Figure 14 indicate the points where the objects forming the bracket and the objects forming the edge intersect with each other. Unlike the intersection points between edges in Figure 12(d), the intersection points between the bracket and the edge can be annotated as shown in Figure 14.

[0082] In this manner, the objects constituting the brackets and the objects constituting the edges of the second chemical structural formula can be formed by annotating the points where the objects constituting the brackets and the objects constituting the edges intersect with each other.

[0083] Such annotation work can be performed automatically by the processor 110 of the training data generation device 100.

[0084] FIG. 15 is a block diagram illustrating a method for generating training data according to an embodiment of the present invention.

[0085] The training data generation device 100 loads first chemical formula data 200 including at least one node information and at least one edge information, and at least one feature variable 300 (S100).

[0086] The training data generation device 100 sets the setting value of at least one feature variable 300 in the first setting value set 400 (S200).

[0087] The training data generation device 100 generates a first chemical structural formula based on the first chemical formula data 200, using the objects set and applied in the first setting value set 400 (S300).

[0088] The training data generation device 100 acquires a second set of setting values ​​500 in which the setting values ​​of at least one feature variable 300 have been changed in response to an input that changes the setting values ​​of the feature variables (S400).

[0089] The training data generation device 100 generates a second chemical structural formula using the object set and applied in the second setting value set 500 (S500).

[0090] The training data generation device 100 generates training data including the generated second chemical structural formula (S600). As described above, the generated second chemical structural formula may be annotated with a setting value set for each object.

[0091] The memory may include at least one type of storage medium of a flash memory type, a hard disk type, a solid state disk type (SSD type), a silicon disk drive type (SDD type), a multimedia card micro type, a card-type memory (such as SD or XD memory), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, and an optical disk.

[0092] Meanwhile, the disclosed embodiments may be realized in the form of a recording medium storing computer-executable instructions. The instructions may be stored in the form of program code, which, when executed by a processor, may generate program modules to perform the operations of the disclosed embodiments. The recording medium may be realized as a computer-readable recording medium.

[0093] Computer-readable recording media include all types of recording media that store computer-readable instructions, such as ROM (Read Only Memory), RAM (Random Access Memory), magnetic tape, magnetic disk, flash memory, and optical data storage devices.

[0094] The disclosed embodiments have been described above with reference to the accompanying drawings. Those skilled in the art will understand that the present invention can be implemented in forms different from the disclosed embodiments without changing the technical idea or essential features of the present invention. The disclosed embodiments are illustrative and should not be interpreted as limiting. [Industrial Applicability]

[0095] INDUSTRIAL APPLICABILITY The present invention relates to a training data generation device that performs machine learning on chemical structure training data and an artificial intelligence model trained by the training data generation device, and therefore has industrial applicability.

Claims

1. a processor; at least one memory electrically connected to the processor; The at least one memory storing at least one characteristic variable that determines a characteristic of the object representing the chemical structure formula, and a first set of predetermined values ​​for each of the at least one characteristic variable; The processor: Loading first chemical formula data including at least one node information and at least one edge information and the at least one feature variable; setting the setting value of the at least one characteristic variable to the first set of settings; generating a first chemical structural formula based on the first chemical formula data using the object set and applied to the first setting value set; acquiring a second set of setting values ​​in which the setting values ​​of the at least one characteristic variable have been changed by an input for changing the setting values ​​of the at least one characteristic variable; generating a second chemical structure using the object configured and applied to the second set of settings; generating training data including the generated second chemical structural formula; Training data generation device.

2. The object constituting the node of the second chemical structural formula is Annotated with the setting values ​​set for the feature variables of the object, The training data generating device according to claim 1 .

3. The object constituting the edge of the second chemical structure formula is Annotated with the setting values ​​set for the feature variables of the object, The training data generating device according to claim 1 .

4. The object constituting the brackets of the second chemical structural formula is Annotated with the setting values ​​set for the feature variables of the object, The training data generating device according to claim 1 .

5. The objects constituting the brackets of the second chemical structural formula and the objects constituting the edges are The intersection point between the object constituting the bracket and the object constituting the edge is annotated to form the annotation. The training data generating device according to claim 1 .

6. the learning data generation device, a display that outputs a chemical structure; The display includes: The first chemical structural formula and the second chemical structural formula are output, and the second chemical structural formula is output separately for each time point when the changed second set of setting values ​​is acquired. The training data generating device according to claim 1 .

7. The at least one feature variable is The information includes at least one of the type, thickness, length, size, color, spacing between objects, and tilt angle of the objects. The training data generating device according to claim 1 .

8. The learning data is The method is used to train an artificial intelligence model for predicting chemical formula data including at least one node information and at least one edge information; The training data generating device according to claim 1 .

9. A method performed by a training data generation device, the training data generation device stores at least one feature variable that determines a feature of an object that represents a chemical structural formula, and a first set of setting values ​​that are predetermined for each of the at least one feature variable; Loading first chemical formula data including at least one node information and at least one edge information, and the at least one feature variable; setting a setting value of the at least one characteristic variable to the first set of settings; generating a first chemical structural formula based on the first chemical formula data using objects configured and applied to the first set of settings; acquiring a second set of setting values ​​in which the setting values ​​of the at least one characteristic variable have been changed by an input that changes the setting value of the at least one characteristic variable; generating a second chemical structure using the object set and applied to the second set of settings; generating training data including the generated second chemical structural formula; How to generate training data.

10. The object constituting the node of the second chemical structural formula is Annotated with the setting values ​​set for the feature variables of the object, The training data generating device according to claim 9 .

11. The object constituting the edge of the second chemical structure formula is Annotated with the setting values ​​set for the feature variables of the object, The training data generating device according to claim 9 .

12. The object constituting the brackets of the second chemical structural formula is Annotated with the setting values ​​set for the feature variables of the object, The training data generating device according to claim 9 .

13. The objects constituting the brackets of the second chemical structural formula and the objects constituting the edges are The intersection point between the object constituting the bracket and the object constituting the edge is annotated to form the annotation. The training data generating device according to claim 9 .

14. The at least one feature variable is The information includes at least one of the type, thickness, length, size, color, spacing between objects, and tilt angle of the objects. The training data generating device according to claim 9 .

15. A computer program stored on a computer-readable recording medium for executing the training data generation method of claim 9 when combined with a hardware device.

Images