Ancient book calculation table identification and evaluation method and system and electronic equipment

Abstract

The invention discloses an ancient book calculation table recognition and evaluation method and system and electronic equipment, and relates to the technical field of ancient book calculation table recognized.The method comprises the steps that S1, a worker uploads an ancient book calculation table and a feature set corresponding to the ancient book calculation table, and a self-recognition model of the ancient book calculation table is determined; s2, randomly selecting a plurality of training samples from the ancient book calculation table, inputting the training samples into the self-recognition model, outputting objective recognition information of the ancient book calculation table, distributing the training samples of the ancient book calculation table to recognition personnel, and outputting subjective recognition information of the ancient book calculation table by the recognition personnel; and S3, evaluating the identification index of the self-identification model of the ancient book calculation table, determining the optimization direction of the self-identification model of the ancient book calculation table, and displaying the optimization direction of the self-identification model of the ancient book calculation table, so that the resource waste of the self-identification model is avoided, the identification efficiency of the ancient book calculation table is also improved, and the user experience is improved. According to the invention, the precision of the ancient book calculation table in character recognition, numerical recognition and the like is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of ancient book calculation table recognition, and in particular to a method, system and electronic device for recognizing and evaluating ancient book calculation tables. Background Art

[0002] As the crystallization of mathematical wisdom, ancient book calculation tables not only record the specific calculation methods of ancient mathematics, but also reflect the level, characteristics and trends of the development of mathematics at that time. Through the research on calculation tables, we can deeply understand the origin, evolution and achievements of ancient mathematics, providing physical evidence and theoretical support for the research of modern mathematical history. Through recognition and evaluation, these precious cultural heritages can be systematically sorted out and protected, avoiding information loss or damage caused by factors such as time and environment, and ensuring the continuous inheritance of mathematical knowledge. Therefore, it is extremely necessary to recognize and evaluate ancient book calculation tables.

[0003] However, for the current recognition of ancient book calculation tables, there are still some deficiencies in the existing technologies, which can be specifically reflected in the following aspects: In the existing technologies, it is rare to determine the self-recognition model of ancient book calculation tables based on ancient book calculation tables and their corresponding feature sets, so it is difficult to ensure the initial adaptability between the ancient book calculation tables and the self-recognition model, which is likely to cause waste of resources of the self-recognition model and reduce the recognition efficiency of ancient book calculation tables. At the same time, it is rare to perform post-evaluation on the training samples through the self-recognition model of ancient book calculation tables to confirm the optimization direction of the self-recognition model of ancient book calculation tables, making it difficult to ensure the recognition accuracy of the self-recognition model of ancient book calculation tables and difficult to capture the complex features of ancient book calculation tables. Summary of the Invention

[0004] The purpose of the present invention is to provide a method, system and electronic device for recognizing and evaluating ancient book calculation tables, which solves the problems existing in the background art.

[0005] To solve the above technical problems, the present invention adopts the following technical solutions: In the first aspect of the present invention, a method for recognizing and evaluating ancient book calculation tables is provided, including: S1. Staff upload ancient book calculation tables and their corresponding feature sets, and determine the self-recognition model of ancient book calculation tables.

[0006] The self-recognition model includes the combination of an image processing method and a recognition method.

[0007] S2. Randomly select several training samples from the ancient book calculation tables and input them into the self-recognition model to output the objective recognition information of the ancient book calculation tables, and distribute several training samples of the ancient book calculation tables to the recognition personnel, and the recognition personnel output the subjective recognition information of the ancient book calculation tables.

[0008] S3. Evaluate the recognition indexes of the self-recognition model of ancient book calculation tables, determine the optimization direction of the self-recognition model of ancient book calculation tables, and display the optimization direction of the self-recognition model of ancient book calculation tables.

[0009] In the second aspect of the present invention, a system for implementing the ancient book calculation table recognition and evaluation method described in the present invention is provided, including: an ancient book calculation table uploading module, which is used for staff to upload the ancient book calculation table and its corresponding feature set, and determine the self-recognition model of the ancient book calculation table.

[0010] The self-recognition model includes a combination of an image processing method and a recognition method.

[0011] An ancient book calculation table recognition module, which is used to randomly select several training samples from the ancient book calculation table and input them into the self-recognition model, output the objective recognition information of the ancient book calculation table, and assign several training samples of the ancient book calculation table to the recognition personnel, and the recognition personnel output the subjective recognition information of the ancient book calculation table.

[0012] An ancient book recognition evaluation module, which is used to evaluate the recognition index of the self-recognition model of the ancient book calculation table, determine the optimization direction of the self-recognition model of the ancient book calculation table, and display the optimization direction of the self-recognition model of the ancient book calculation table.

[0013] In the third aspect of the present invention, an electronic device is provided, including: a processor, a memory and a communication bus. A computer-readable program executable by the processor is stored on the memory. The communication bus realizes the connection and communication between the processor and the memory. When the processor executes the computer-readable program, the ancient book calculation table recognition and evaluation method described in the present invention is realized.

[0014] The beneficial effects of the present invention are as follows: (1) Based on the ancient book calculation table and its corresponding feature set, the present invention determines the self-recognition model of the ancient book calculation table, thereby ensuring the initial adaptability between the ancient book calculation table and the self-recognition model, avoiding waste of resources of the self-recognition model, and improving the recognition efficiency of the ancient book calculation table, laying a foundation for the subsequent confirmation of the optimization direction of the self-recognition model of the ancient book calculation table.

[0015] (2) After the present invention recognizes and evaluates the training samples through the self-recognition model of the ancient book calculation table, and then confirms the optimization direction of the self-recognition model of the ancient book calculation table, to a certain extent, the recognition accuracy of the self-recognition model of the ancient book calculation table is improved, the capture rate of complex features of the ancient book calculation table is increased, and the accuracy of the ancient book calculation table in character recognition, numerical recognition, etc. is ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1This is a schematic flowchart of the implementation steps of the method of the present invention.

[0018] Figure 2 This is a schematic connection diagram of the system structure of the present invention. Detailed implementation manners

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] Refer to Figure 1 As shown, a method for identifying and evaluating ancient Chinese arithmetic tables according to a first aspect of the present invention includes: S1. A staff member uploads an ancient Chinese arithmetic table and its corresponding feature set, and determines a self-identification model for the ancient Chinese arithmetic table.

[0021] It should be noted that the feature set includes a content complexity feature value, a table structure complexity feature value, a font complexity feature value, etc., and the content complexity feature value, the table structure complexity feature value, and the font complexity feature value are specifically numerical values from 0 to 1.

[0022] The self-identification model includes a combination of an image processing method and an identification method.

[0023] In a specific embodiment of the present invention, the process of determining the self-identification model for the ancient Chinese arithmetic table is as follows: S100. Identify the defect parameters of the ancient Chinese arithmetic table through image recognition technology.

[0024] The defect parameter is a general defect characterization parameter.

[0025] It should be noted that the general defect characterization parameter is specifically the total defect area after normalization. For example, if the total defect area is 10 square meters and the total area of the ancient Chinese arithmetic table is 80 square meters, then 0, 10, and 80 are normalized, and the total defect area after normalization is 0.125.

[0026] S101. Supplement the defect parameters of the ancient Chinese arithmetic table into the feature set of the ancient Chinese arithmetic table to obtain an updated feature set of the ancient Chinese arithmetic table.

[0027] S102. Compare the updated feature set of the ancient Chinese arithmetic table with the difficulty level classification table of the ancient Chinese arithmetic tables stored in the web data warehouse, and obtain the difficulty level of the ancient Chinese arithmetic table through matching.

[0028] It should be noted that the difficulty level classification table of the ancient arithmetic table is the demand interval of each element in the feature set corresponding to each difficulty level. For example, the content complexity feature value is 0.1, the table structure complexity feature value is 0.1, the font complexity feature value is 0.1, and the overall defect characterization parameter is 0.2. If it is within the demand interval of each element in the feature set corresponding to the easy level of the ancient arithmetic table, the difficulty level of the ancient arithmetic table is recorded as easy.

[0029] S103. Compare the difficulty level of the ancient arithmetic table with the self-identification models corresponding to each difficulty level of the ancient arithmetic tables stored in the web data warehouse, and obtain the self-identification model of the ancient arithmetic table through matching.

[0030] It should be noted that the self-identification models corresponding to each difficulty level of the ancient arithmetic table. For example, if the difficulty level of the ancient arithmetic table is easy, the image processing is denoising + binarization + skew correction, and the recognition method is: traditional OCR + rule-based table structure recognition. If the difficulty level of the ancient arithmetic table is difficult, the image processing is super-resolution + contrast enhancement + table area detection, and the recognition method is deep learning OCR (such as CRNN or TrOCR) + deep learning table structure recognition (such as TableNet), which are specifically uploaded by the staff.

[0031] Based on the ancient arithmetic table and its corresponding feature set, the present invention determines the self-identification model of the ancient arithmetic table, thereby ensuring the initial adaptability between the ancient arithmetic table and the self-identification model, avoiding waste of resources of the self-identification model, and also improving the recognition efficiency of the ancient arithmetic table, laying a foundation for determining the optimization direction of the self-identification model of the ancient arithmetic table in the future.

[0032] S2. Randomly select several training samples from the ancient arithmetic table and input them into the self-identification model to output the objective recognition information of the ancient arithmetic table, and distribute several training samples of the ancient arithmetic table to the recognition personnel, and the recognition personnel output the subjective recognition information of the ancient arithmetic table.

[0033] In a specific embodiment of the present invention, the objective recognition information includes a text data set, a numerical data set, a symbol data set, and a table structure data set of several training samples.

[0034] The text data set includes a set of header keywords and a set of keywords for each cell. The symbol data set includes a set of operation symbols and a set of table delimiter symbols. The table structure data set includes the number of rows, the number of columns, each merged cell, and each split cell.

[0035] The subjective recognition information includes a text data set, a numerical data set, a symbol data set, and a table structure data set of several training samples.

[0036] S3. Evaluate the recognition metrics of the self-recognition model for ancient Chinese arithmetic tables, determine the optimization direction of the self-recognition model for ancient Chinese arithmetic tables, and display the optimization direction of the self-recognition model for ancient Chinese arithmetic tables.

[0037] In a specific embodiment of the present invention, the specific implementation steps for evaluating the recognition metrics at all levels of the self-recognition model for ancient Chinese arithmetic tables are as follows: S300. Based on the objective recognition information and subjective recognition information of the ancient Chinese arithmetic table, with a certain training sample as the specified sample, determine the first recognition metric α of the specified sample of the ancient Chinese arithmetic table _1 , the second recognition metric α _2 , the third recognition metric α _3 and the fourth recognition metric α _4 .

[0038] It should be noted that the specific determination method for the second recognition metric of the specified sample of the ancient Chinese arithmetic table is as follows: By using the set similarity algorithm, calculate the similarity between the numerical data set in the objective recognition information of the ancient Chinese arithmetic table and the numerical data set in the subjective recognition information, as the second recognition metric of the ancient Chinese arithmetic table.

[0039] It should also be noted that the set similarity algorithm is specifically where A and B respectively refer to two sets.

[0040] S301. By analogy, obtain the first recognition metric, second recognition metric, third recognition metric, and fourth recognition metric of each training sample of the ancient Chinese arithmetic table, and perform mean processing on them to obtain the mean value of the first recognition metric, mean value of the second recognition metric, mean value of the third recognition metric, and mean value of the fourth recognition metric of the ancient Chinese arithmetic table, as the first-level recognition metric β _1 , second-level recognition metric β _2 , third-level recognition metric β _3 , fourth-level recognition metric β _4 .

[0041] In a specific embodiment of the present invention, the determination of the first recognition metric α _1 of the specified sample of the ancient Chinese arithmetic table, the specific determination method is as follows: Extract the set of header keywords and the set of keywords for each cell in the text data set from the objective recognition information of the specified sample of the ancient Chinese arithmetic table, and extract the set of header keywords and the set of keywords for each cell in the text data set from the subjective recognition information of the specified sample of the ancient Chinese arithmetic table.

[0042] By using the set similarity algorithm, calculate the similarity between the set of header keywords in the objective recognition information of the specified sample of the ancient Chinese arithmetic table and the set of header keywords in the subjective recognition information, as the header recognition accuracy of the specified sample of the ancient Chinese arithmetic table.

[0043] By using the set similarity algorithm, calculate the similarity between the keyword set of each cell in the objective recognition information of the specified sample of the ancient arithmetic table and the keyword set of the corresponding cell, which is used as the recognition accuracy of each cell in the specified sample of the ancient arithmetic table, and screen out the heterogeneous cells of the specified sample of the ancient arithmetic table.

[0044] It should be noted that the recognition accuracy is specifically a value between 0 and 1.

[0045] It should also be noted that the specific screening method for the heterogeneous cells of the specified sample of the ancient arithmetic table is as follows: compare the recognition accuracy of each cell in the specified sample of the ancient arithmetic table with the cell recognition accuracy threshold stored in the web data warehouse. If the recognition accuracy of a certain cell is less than the cell recognition accuracy threshold, then mark this cell as a heterogeneous cell, and screen out the heterogeneous cells of the specified sample of the ancient arithmetic table. The cell recognition accuracy threshold is specifically set by the staff of the ancient arithmetic table. For example, in order to improve the character recognition accuracy of the ancient arithmetic table, the cell recognition accuracy threshold is set to 0.95.

[0046] Summarize the number M of heterogeneous cells and the total number M' of cells in the specified sample of the ancient arithmetic table, and import them together with the header recognition accuracy ε' of the specified sample of the ancient arithmetic table into the first recognition index model to output the first recognition index of the specified sample of the ancient arithmetic table.

[0047] It should be noted that the first recognition index model uses the ln() function to control the value range within 0 - 1 for subsequent analysis.

[0048] In a specific embodiment of the present invention, the determination of the third recognition index α of the specified sample of the ancient arithmetic table _3 , and its specific determination method is as follows: extract the operation symbol set and the table separator symbol set in the symbol dataset from the objective recognition information of the specified sample of the ancient arithmetic table, and extract the operation symbol set and the table separator symbol set in the symbol dataset from the subjective recognition information of the specified sample of the ancient arithmetic table.

[0049] By using the set similarity algorithm, calculate the similarity between the operation symbol set in the objective recognition information and the operation symbol set in the subjective recognition information of the specified sample of the ancient arithmetic table, which is used as the operation symbol recognition accuracy of the specified sample of the ancient arithmetic table. Similarly, calculate the table separator symbol recognition accuracy of the specified sample of the ancient arithmetic table.

[0050] Import the operation symbol recognition accuracy η and the table separator symbol recognition accuracy μ of the ancient arithmetic table into the third recognition index model Output the third recognition index of the specified sample of the ancient Chinese arithmetic table. In the formula, η′ and μ′ respectively represent the recognition accuracy thresholds of operation symbols and table separation symbols stored in the web data warehouse, and ∧ and ∨ are logical symbols AND and OR respectively.

[0051] It should be noted that the recognition accuracy thresholds of the operation symbols and the table separation symbols are specifically set by the staff of the ancient Chinese arithmetic table. For example, in order to improve the recognition accuracy of the operation symbols and the table separation symbols of the ancient Chinese arithmetic table, both the recognition accuracy threshold of the operation symbols and the recognition accuracy threshold of the table separation symbols are set to 0.95.

[0052] In a specific embodiment of the present invention, the fourth recognition index α of the specified sample of the ancient Chinese arithmetic table is determined _4 , and its specific determination method is as follows: Extract the number of rows, the number of columns, each merged cell, and each split cell of the arithmetic table structure dataset from the objective recognition information of the specified sample of the ancient Chinese arithmetic table, and extract the number of rows, the number of columns, each merged cell, and each split cell of the arithmetic table structure dataset from the subjective recognition information of the specified sample of the ancient Chinese arithmetic table, and record them as the reference number of rows, the reference number of columns, each reference merged cell, and each reference split cell of the ancient Chinese arithmetic table respectively.

[0053] Through comparison and analysis, obtain the row deviation value, column deviation value, each deviation merged cell, and each deviation split cell of the specified sample of the ancient Chinese arithmetic table, and summarize to obtain the number of deviation merged cells and the number of deviation split cells of the specified sample of the ancient Chinese arithmetic table.

[0054] It should be noted that the specific comparison process for obtaining the row deviation value, column deviation value, logical feedback values of each row and each column, each deviation merged cell, and each deviation split cell of the specified sample of the ancient Chinese arithmetic table through comparison and analysis is as follows: Subtract the number of rows and the number of columns of the specified sample of the ancient Chinese arithmetic table from the reference number of rows and the reference number of columns respectively to obtain the row deviation value and column deviation value of the specified sample of the ancient Chinese arithmetic table. Compare each merged cell of the ancient Chinese arithmetic table with each reference merged cell. If a certain merged cell fails to match with each reference merged cell, then record this merged cell as a deviation merged cell to obtain each deviation merged cell. Similarly, obtain each deviation split cell.

[0055] Summarize to obtain the number of reference merged cells and the number of reference split cells of the specified sample of the ancient Chinese arithmetic table, and through numerical processing, obtain the fourth recognition index of the specified sample of the ancient Chinese arithmetic table.

[0056] It should be noted that the specific processing method for obtaining the fourth recognition index of the specified sample of the ancient arithmetic table is as follows: divide the number of deviation merged cells of the specified sample of the ancient arithmetic table by the number of reference merged cells, and divide the number of deviation split cells of the specified sample of the ancient arithmetic table by the number of reference split cells, respectively, to obtain the deviation merged cell ratio and the deviation split cell ratio of the specified sample of the ancient arithmetic table.

[0057] Compare the row deviation value, column deviation value, deviation merged cell ratio, and deviation split cell ratio of the specified sample of the ancient arithmetic table with the allowable row deviation value, allowable column deviation value, deviation merged cell ratio threshold, and deviation split cell ratio threshold stored in the web data warehouse respectively. If the row deviation value is less than the allowable row deviation value, the column deviation value is less than the allowable column deviation value, the deviation merged cell ratio is less than the deviation merged cell ratio threshold, and the deviation split cell ratio is less than the deviation split cell ratio threshold, then record the fourth recognition index of the specified sample of the ancient arithmetic table as 1; otherwise, record it as 0.

[0058] In a specific embodiment of the present invention, the specific determination method for determining the optimization direction of the self-recognition model of the ancient arithmetic table is as follows: based on the first recognition index α of the ancient arithmetic table _1 , the second recognition index α _2 , the third recognition index α _3 and the fourth recognition index α _4 .

[0059] If (β _1 < β _ ′ 1 ), then record the optimization direction of the self-recognition model of the ancient arithmetic table as text optimization.

[0060] If (β _2 < β _ ′ 2 ), then record the optimization direction of the self-recognition model of the ancient arithmetic table as numerical optimization.

[0061] If (β _3 = 0), then record the optimization direction of the self-recognition model of the ancient arithmetic table as symbol optimization.

[0062] If (β _4 = 0), then record the optimization direction of the self-recognition model of the ancient arithmetic table as arithmetic table structure optimization.

[0063] β _ ′ 1 and β _ ′ 2 are respectively the first-level recognition index convergence value and the second-level recognition index convergence value stored in the web data warehouse.

[0064] It should be noted that the convergence values of the first recognition index and the second recognition index are specifically set by the staff of the ancient book calculation table. For example, in order to ensure the text recognition effect and numerical recognition effect of the self-recognition model of the ancient book calculation table, the convergence values of the first recognition index and the second recognition index can both be set to 0.9.

[0065] Summarize to obtain the optimization direction of the self-recognition model of the ancient book calculation table.

[0066] In the present invention, after the training samples are recognized and evaluated by the self-recognition model of the ancient book calculation table, the optimization direction of the self-recognition model of the ancient book calculation table is confirmed, which improves the recognition accuracy of the self-recognition model of the ancient book calculation table to a certain extent, increases the capture rate of complex features of the ancient book calculation table, and ensures the accuracy of the ancient book calculation table in text recognition, numerical recognition, etc.

[0067] Refer to Figure 2 As shown, the second aspect of the present invention provides a system for executing the ancient book calculation table recognition and evaluation method of the present invention, including:

[0068] An ancient book calculation table uploading module, which is used for the staff to upload the ancient book calculation table and its corresponding feature set, and determine the self-recognition model of the ancient book calculation table.

[0069] The self-recognition model includes a combination of an image processing method and a recognition method.

[0070] An ancient book calculation table recognition module, which is used to randomly select several training samples from the ancient book calculation table and input them into the self-recognition model, output the objective recognition information of the ancient book calculation table, and distribute several training samples of the ancient book calculation table to the recognition personnel, and the recognition personnel output the subjective recognition information of the ancient book calculation table.

[0071] An ancient book recognition and evaluation module, which is used to evaluate the recognition index of the self-recognition model of the ancient book calculation table, determine the optimization direction of the self-recognition model of the ancient book calculation table, and display the optimization direction of the self-recognition model of the ancient book calculation table.

[0072] It should be noted that the present invention also includes a web data warehouse. The ancient book calculation table uploading module is connected to the ancient book calculation table recognition module, the ancient book calculation table recognition module is connected to the ancient book recognition and evaluation module, and the web data warehouse is respectively connected to the ancient book calculation table uploading module and the ancient book recognition and evaluation module.

[0073] The third aspect of the present invention provides an electronic device, including: a processor, a memory and a communication bus. A computer-readable program executable by the processor is stored on the memory. The communication bus realizes the connection and communication between the processor and the memory. When the processor executes the computer-readable program, the ancient book calculation table recognition and evaluation method as described in the present invention is realized.

[0074] The above content is only an example and illustration of the concept of the present invention. Those skilled in the art of the present technology can make various modifications, supplements, or use similar methods to replace the specific embodiments described, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A method for identifying and evaluating ancient book arithmetic tables, characterized in that: include: S1. The staff uploads the ancient book calculation table and its corresponding feature set, and determines the self-recognition model of the ancient book calculation table; The self-recognition model includes a combination of an image processing method and a recognition method; S2. randomly select a number of training samples from the ancient book table and input them into the self-recognition model, output the objective recognition information of the ancient book table, and distribute a number of training samples of the ancient book table to the recognition personnel, who then output the subjective recognition information of the ancient book table; S3. Evaluate the recognition index of the self-recognition model of the ancient book arithmetic table, determine the optimization direction of the self-recognition model of the ancient book arithmetic table, and display the optimization direction of the self-recognition model of the ancient book arithmetic table.

2. The method for identifying and evaluating ancient book arithmetic tables according to claim 1, characterized in that: The specific determination process of the self-recognition model for determining the ancient book calculation table is as follows: S100, identifying defective parameters of ancient book calculation tables by using image recognition technology; The defect parameter is the overall defect characterization parameter; S101, adding defect parameters of the ancient book calculation table to the feature set of the ancient book calculation table to obtain an updated feature set of the ancient book calculation table; S102, comparing the updated feature set of the ancient book calculation table with the difficulty classification table of the ancient book calculation table stored in the web data warehouse, and obtaining the difficulty level of the ancient book calculation table through matching; S103, comparing the difficulty level of the ancient book calculation table with the self-recognition models corresponding to the various difficulty levels of the ancient book calculation tables stored in the web data warehouse, and obtaining the self-recognition model of the ancient book calculation table through matching.

3. The method for identifying and evaluating ancient book arithmetic tables according to claim 1, characterized in that: The objective recognition information includes a text data set, a numerical data set, a symbol data set and a table structure data set of a number of training samples; The subjective recognition information includes a text data set, a numerical data set, a symbol data set and a calculation table structure data set of several training samples.

4. The method for identifying and evaluating ancient book arithmetic tables according to claim 3, characterized in that: The specific implementation steps of the self-recognition model for evaluating the ancient book calculation table are as follows: S300, based on the objective recognition information and subjective recognition information of the ancient book calculation table, taking a certain training sample as a designated sample, determining a first recognition index α of the designated sample of the ancient book calculation table _1 , the second identification index α _2 , the third identification index α _3 and the fourth identification index α _4 ; S301, and so on, obtain the first recognition index, the second recognition index, the third recognition index and the fourth recognition index of each training sample of the ancient book calculation table, and perform mean processing on them to obtain the first recognition index mean, the second recognition index mean, the third recognition index mean and the fourth recognition index mean of the ancient book calculation table as the first-level recognition index β of the self-recognition model of the ancient book calculation table _1 , Secondary identification index β _2 , three-level identification index β _3 , four-level identification index β _4 .

5. The method for identifying and evaluating ancient book arithmetic tables according to claim 4, characterized in that: The first identification index α of the designated sample of the ancient book calculation table is determined _1 , the specific determination method is: Extracting the table header keyword set and the keyword set of each cell in the text data set from the objective recognition information of the designated sample of the ancient book arithmetic table, and extracting the table header keyword set and the keyword set of each cell in the text data set from the subjective recognition information of the designated sample of the ancient book arithmetic table; By using a set similarity algorithm, the similarity between the table header keyword set in the objective identification information and the table header keyword set in the subjective identification information of the designated sample of the ancient book arithmetic table is calculated as the table header recognition accuracy of the designated sample of the ancient book arithmetic table; By using a set similarity algorithm, the similarity between the keyword set of each cell in the objective identification information of the designated sample of the ancient book calculation table and the keyword set of the corresponding cell is calculated as the recognition accuracy of each cell of the designated sample of the ancient book calculation table, and the heterogeneous cells of the designated sample of the ancient book calculation table are screened; Summarize the number of heterogeneous cells M and the total number of cells M′ of the specified sample of the ancient book calculation table, and import them and the table header recognition accuracy ε′ of the specified sample of the ancient book calculation table into the first recognition index model , output the first recognition index of the specified sample of the ancient book calculation table.

6. The method for identifying and evaluating ancient book arithmetic tables according to claim 4, characterized in that: The third recognition index α of the specified sample of the ancient book calculation table is determined _3 , the specific determination method is: Extracting the set of operation symbols and the set of table separator symbols in the symbol data set from the objective recognition information of the designated sample of the ancient book arithmetic table, and extracting the set of operation symbols and the set of table separator symbols in the symbol data set from the subjective recognition information of the designated sample of the ancient book arithmetic table; By using a set similarity algorithm, the similarity between the set of operation symbols in the objective identification information and the set of operation symbols in the subjective identification information of the designated sample of the ancient book arithmetic table is calculated as the recognition accuracy of the operation symbols of the designated sample of the ancient book arithmetic table. Similarly, the recognition accuracy of the table separator symbol of the designated sample of the ancient book arithmetic table is calculated. Import the recognition accuracy of the operation symbols of the ancient arithmetic tables η and the recognition accuracy of the table separators μ into the third recognition index model The third recognition index of the specified sample of the ancient book arithmetic table is output, where η′ and μ′ represent the recognition accuracy threshold of the operation symbol and the recognition accuracy threshold of the table separator symbol stored in the web data warehouse, respectively, and ∧ and ∨ represent the logical symbols and and or, respectively.

7. The method for identifying and evaluating ancient book arithmetic tables according to claim 4, characterized in that: The fourth recognition index α of the specified sample of the ancient book calculation table is determined _4 , the specific determination method is: Extracting the number of rows, columns, merged cells, and split cells of the table structure data set from the objective identification information of the designated sample of the ancient book table, and extracting the number of rows, columns, merged cells, and split cells of the table structure data set from the subjective identification information of the designated sample of the ancient book table, and recording them as the reference number of rows, reference number of columns, reference merged cells, and reference split cells of the ancient book table, respectively; Comparison and analysis are performed to obtain the row number deviation value, column number deviation value, each deviation merged cell, and each deviation split cell of the specified sample of the ancient book calculation table, and the number of the deviation merged cells and the number of the deviation split cells of the specified sample of the ancient book calculation table are summarized; The number of reference merged cells and the number of reference split cells of the designated sample of the ancient book calculation table are obtained by summarizing, and the fourth recognition index of the designated sample of the ancient book calculation table is obtained by numerical processing, and the fourth recognition index includes the values ​​of 0 and 1.

8. The method for identifying and evaluating ancient book arithmetic tables according to claim 3, characterized in that: The specific method for determining the optimization direction of the self-recognition model of the ancient book calculation table is as follows: The first recognition index α based on ancient book calculation table _1 , the second identification index α _2 , the third identification index α _3 and the fourth identification index α _4 ; If (β _1 <β _ ′1), then the optimization direction of the self-recognition model of the ancient book table is recorded as text optimization; If (β _2 <β _ ′2), then the optimization direction of the self-recognition model of the ancient book table is recorded as numerical optimization; If (β _3 =0), the optimization direction of the self-recognition model of the ancient book arithmetic table is recorded as symbol optimization; If (β _4 =0), then the optimization direction of the self-recognition model of the ancient book arithmetic table is recorded as the arithmetic table structure optimization; β _ ′1, β _ ′2 are the convergence values ​​of the first-level identification index and the second-level identification index stored in the web data warehouse respectively; The optimization direction of the self-recognition model of ancient book arithmetic tables is summarized.

9. A system for executing the ancient book arithmetic table recognition and evaluation method according to any one of claims 1 to 8, characterized in that: include: The ancient book calculation table uploading module is used for the staff to upload the ancient book calculation table and its corresponding feature set, and determine the self-recognition model of the ancient book calculation table; The self-recognition model includes a combination of an image processing method and a recognition method; The ancient book arithmetic table recognition module is used to randomly select a number of training samples from the ancient book arithmetic table and input them into the self-recognition model to output the objective recognition information of the ancient book arithmetic table, and distribute the training samples of the ancient book arithmetic table to the recognition personnel, who then output the subjective recognition information of the ancient book arithmetic table; The ancient book recognition evaluation module is used to evaluate the recognition index of the self-recognition model of the ancient book calculation table, determine the optimization direction of the self-recognition model of the ancient book calculation table, and display the optimization direction of the self-recognition model of the ancient book calculation table.

10. An electronic device, characterized in that: include: Processor, memory and communication bus; The memory stores a computer-readable program that can be executed by the processor; the communication bus realizes the connection and communication between the processor and the memory; when the processor executes the computer-readable program, it implements the ancient book arithmetic table recognition and evaluation method as described in any one of claims 1-8.

Images