CN-121986363-A - Performance evaluation or learning system, method, and program for rendering artificial intelligence model using graph of constructed data set containing graph information

Abstract

A system, method, and program for evaluating the performance of or learning an artificial intelligence model for rendering a chart by constructing a dataset containing chart information is disclosed. The system includes a memory including a data set generation model that stores line information, which is information on at least one line of a graph, and meta information, which is information on meta data, as GT (Ground Truth, true value), and stores an image formed by the GT as a graph image, and outputs the GT and the graph image as a data set, and an AI model that accepts an input of the graph image stored in the data set, outputs a data format in which information of the graph is predicted, and a processor that operates the AI model or causes the AI model to learn and operate a performance evaluation model. The factors included in the line information may include an X-axis value of a line of a graph, a function, a coefficient (coefficient) of the function, a color or shape of a line or a point, and the like, and the factors included in the meta information may include a graph title, an X-axis name, a Y-axis name, a legend (legend), and the like. Also, the data format output from the AI model may be used to evaluate the performance of the AI model or to cause the AI model to learn.

Inventors

• LIU GUANGLU
• LI CHENGJUN
• LI CHUNYONG

Assignees

• (株)LG经营开发院

Dates

Publication Date

20260505

Application Date

20250124

Priority Date

20240125

Claims (19)

1. 1. A system for implementing a chart de-rendering model, comprising: At least one processor, and At least one memory storing instructions or information that cause the at least one processor to perform actions, The actions performed by the instructions include the steps of: Storing line information and meta information, which is information related to at least one line of a graph, as a true value by a dataset generation model, wherein the line information is information related to at least one line of the graph, the meta information is information related to meta data, and an image formed by using the GT is stored as a graph image, and the GT and the graph image are output as a dataset; Inputting the chart image stored in the dataset into an AI model, outputting a data format of information for predicting the chart, and Inputting the data format into a performance evaluation model, comparing the information of the data format with the GT stored in the data set, outputting a performance evaluation result aiming at the AI model, The values to which the factors included in the line information and the meta information are applied are selected from predetermined values, respectively.
2. 2. The system of claim 1, wherein, The method further comprises the step of inputting the data format output from the AI model into a performance evaluation model, comparing the information of the data format with the GT stored in the data set, and outputting a performance evaluation result aiming at the AI model.
3. 3. The system of claim 1, wherein, The method further comprises the steps of comparing information in the data format output from the AI model with the GT stored in the data set through the AI model, and learning the AI model by using a comparison result.
4. 4. The system of claim 1, wherein, The factors included in the line information are constituted to include the X-axis value of the graph line, the function, and the coefficient of the function, i.e., coefficient.
5. 5. The system of claim 4, wherein, The data set is constituted by a first data set and a second data set, The difference between the coefficients of the function contained in the second data set and the coefficients of the function contained in the first data set is less than a preset value.
6. 6. The system of claim 4, wherein, The maximum value of the function is larger than the preset maximum function value, and the minimum value of the function is smaller than the preset minimum function value.
7. 7. The system of claim 1, wherein, The factors included in the line information are constituted to include the color or shape of the line or the dot.
8. 8. The system of claim 1, wherein, The factors included in the meta information are constituted to include a chart title, an X-axis name, a Y-axis name, and a legend legend.
9. 9. The system of claim 2, 7 or 8, wherein, The value to which the factor is applied is selected for each predefined value with a predefined probability.
10. 10. A method for implementing a chart de-rendering model, comprising the steps of: storing line information and meta information as a true value, GT, which is information related to at least one line of a graph, by a data set generation model, storing an image formed by using GT as a graph image, and outputting GT and the graph image as a data set; Inputting the chart image stored in the dataset into an AI model, outputting a data format of information for predicting the chart, and Inputting the data format into a performance evaluation model, comparing the information of the data format with the GT stored in the data set, outputting a performance evaluation result aiming at the AI model, wherein, The values to which the factors included in the line information and the meta information are applied are selected from predetermined values, respectively.
11. 11. The method of claim 10, wherein, The method further comprises the steps of inputting the data format output from the AI model to a performance evaluation model, comparing the information of the data format with the GT stored in the data set, and outputting a performance evaluation result for the AI model.
12. 12. The method of claim 10, wherein, The method further comprises the steps of comparing information in the data format output from the AI model with the GT stored in the dataset by means of the AI model and using the comparison result to learn the AI model.
13. 13. The method of claim 10, wherein, The factors included in the line information are constituted to include the X-axis value of the graph line, the function, and the coefficient of the function, i.e., coefficient.
14. 14. The method of claim 13, wherein, The data set is constituted by a first data set and a second data set, The difference between the coefficients of the function contained in the second data set and the coefficients of the function contained in the first data set is less than a preset value.
15. 15. The method of claim 13, wherein, The maximum value of the function is larger than the preset maximum function value, and the minimum value of the function is smaller than the preset minimum function value.
16. 16. The method of claim 10, wherein, The factors included in the line information are constituted to include the color or shape of the line or the dot.
17. 17. The method of claim 10, wherein, The factors included in the meta information are constituted to include a chart title, an X-axis name, a Y-axis name, and a legend legend.
18. 18. The method of claim 13, 16 or 17, wherein, The value to which the factor is applied is selected for each predefined value with a predefined probability.
19. 19. A program stored in a computer-readable recording medium for executing the method of any one of claims 11 to 18 in conjunction with a computer.

Description

Performance evaluation or learning system, method, and program for rendering artificial intelligence model using graph of constructed data set containing graph information Technical Field The present invention relates to a performance evaluation or learning system, method, and program for a graph rendering model by constructing a data set containing graph information, and more particularly, to a system, method, and program for evaluating the performance of an artificial intelligence model for rendering a graph or learning an artificial intelligence model by constructing a data set containing graph information. Background Chart De-Rendering (De-Rendering) is a process that is the reverse of chart Rendering (Rendering) and refers to a job of analyzing and grouping visual patterns or information of charts to extract key information, and extracting data-related information (e.g., numbers, groups, etc.), chart layout-related information, etc. from the key information. For chart De-rendering (De-rendering), a dataset (Data Set) composed of a chart image including text, lines, etc., GT (Ground Truth, true value) including information related to a chart, etc., is input to the AI model. The data set must be cleaned so that the AI model can be identified, and must contain data of various styles so that the AI model can be evaluated or learned from multiple angles. The conventional method for constructing a data set employs a method of collecting data by crawling crawling a specific website or the like (PlotQA: reasoning over Scientific Plots, NITESH METHANI, 3 persons out of NITESH METHANI, 2020.). A problem with this approach is that the amount of data contained in the data set is limited, and often only data tailored to a particular style is collected, and thus the collected data lacks diversity. Further, there is a problem in that the chart rendering AI model recognizes meta information together with numerical information contained in the data, and the data collected by crawling often does not completely contain meta information (for example, names of X-axis and Y-axis, names of entity groups recorded in legend (legend), etc.). Further, in order to evaluate or learn the AI model more accurately, it is necessary to divide the chart images drawn in the similar style into comparison products of the experimental group and the control group, but it is difficult to collect the chart images in the similar style according to the conventional data set construction method, so that the comparison method cannot be adopted. Prior art literature Non-patent literature PlotQA: reasoning over Scientific Plots, NITESH METHANI, 3 others, 2020. Disclosure of Invention Problems to be solved by the invention The present invention provides a system, method and program for evaluating performance of an artificial intelligence model for rendering a chart or learning the artificial intelligence model by constructing a data set containing chart information in a large amount. The problems to be solved by the present invention are not limited to the above-mentioned problems, and other problems not mentioned will be clearly understood by those skilled in the art from the following description. Means for solving the problems A system for implementing a chart de-rendering model includes at least one or more processors and at least one or more memories for storing instructions or information for causing the at least one or more processors to execute operations including the steps of storing, by a data set generation model, line information, which is information related to at least one or more lines of a chart, and meta information, which is information related to meta data, as GT (Ground Truth), and storing an image formed by using GT as a chart image, and outputting the GT and the chart image as a data set, inputting the chart image stored in the data set into an AI model, outputting a data format in which information of the chart is predicted, and inputting the data format into a performance evaluation model, comparing the information of the data format with the GT stored in the data set, and outputting performance evaluation results for the AI model, wherein factors included in the line information and factors included in the meta information are each selected from predetermined values applied respectively. In the system, the method may further include the step of inputting the data format output from the AI model to a performance evaluation model, comparing information of the data format with the GT stored in the data set, and outputting a performance evaluation result for the AI model. In the system, it is possible that the action further includes the steps of comparing, by an AI model, information of the data format output from the AI model with the GT stored in the dataset, and using a result of the comparison to learn the AI model. In the system, the factor included in the line information may be configured to include an X-axis value of a