Knowledge graph construction method and system based on digital archive resources

Abstract

The present invention relates to the technical field of digital archive resources, and discloses a method and system for constructing a knowledge graph based on digital archive resources, comprising the following steps: Step 1: converting paper archives into electronic documents, obtaining digital archive resources, and extracting content therein; Step 2: based on the extracted content, multiple recognitions of entities in the same electronic archive text are performed, and recognition result data is obtained, and the recognition result evaluation value is calculated; Step 3: based on the obtained recognition result evaluation value, the recognition result of the same electronic archive is evaluated, and the evaluation result includes generating a recognition accuracy signal; Step 4: based on the generated recognition accuracy signal. When the present invention identifies entities, multiple repeated recognitions are performed, and the results are comprehensively analyzed, which effectively improves the accuracy and reliability of entity recognition, and provides strong support for the construction of knowledge graphs based on digital archive resources.

Description

Technical Field

[0001] The present invention relates to the technical field of digital archive resources, and in particular to a method and system for constructing a knowledge graph based on digital archive resources. Background Art

[0002] Knowledge Graph is a structured semantic knowledge base, which is mainly used to describe entities and their relationships in the real world. It consists of nodes and edges, where nodes represent entities or abstract concepts, and edges represent attributes or relationships between entities.

[0003] However, in the process of constructing the knowledge graph of digital archives, due to the complexity and diversity of digital archive data, recognition errors may occur when performing entity recognition. The existing part lacks the process of monitoring and evaluating the recognition results, and cannot timely discover the abnormalities in the recognition results, which affects the subsequent construction of the knowledge graph. At the same time, the classification management of electronic archives is relatively simple, resulting in low retrieval efficiency.

[0004] In view of this, we proposed a knowledge graph construction method and system based on digital archive resources. Summary of the invention

[0005] The purpose of the present invention is to provide a method and system for constructing a knowledge graph based on digital archive resources to solve the problems in the above background.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] The method and system for constructing a knowledge graph based on digital archive resources include the following steps: Step 1: Convert paper archives into electronic documents, obtain digital archive resources, and extract the content therein; Step 2: Based on the extracted content, perform multiple recognitions of entities in the same electronic archive text, obtain recognition result data, process the data, and calculate the recognition result evaluation value; Step 3: Based on the obtained recognition result evaluation value, evaluate the recognition result of the same electronic archive, and the evaluation result includes generating a recognition accuracy signal; Step 4: Based on the generated recognition accuracy signal, construct a knowledge graph for the recognition result of the electronic archive.

[0008] As a further solution of the present invention: the process of obtaining the recognition result evaluation value is as follows:

[0009] Get the maximum value of the percentage of identical results, the percentage of groups, and the percentage of incomplete identifications;

[0010] Substitute into the formula The recognition result evaluation value JG is calculated, where XT represents the maximum value of the number of identical results, ZB represents the number of groups, BW represents the number of incomplete recognitions, and a1, a2, and a3 are preset proportional coefficients.

[0011] As a further solution of the present invention: the process of obtaining the maximum value of the number of identical results, the value of the number of groups and the value of the number of incomplete identifications is as follows:

[0012] Preset the number of recognition times, repeatedly perform entity recognition on the extracted key content, and obtain the recognition results each time;

[0013] Traverse all recognition results, mark the same recognition results as the same result group, obtain the number of the same result group, and perform ratio processing on the number of the same result group and the recognition times to obtain the group number ratio;

[0014] In each identical result group, obtain the number of recognition results, mark it as the identical result number value, perform ratio processing on the identical result number value and the recognition times to obtain the identical result number percentage value, extract the maximum value of the identical result number percentage value in all identical result groups, and mark it as the identical result number percentage maximum value;

[0015] The number of incomplete identifications in the identification results is extracted, and the number of incomplete identifications is processed with the number of identifications to obtain the ratio of incomplete identifications.

[0016] As a further solution of the present invention: the process of generating an accurate identification signal is:

[0017] Obtaining a recognition result evaluation value of each electronic file, and comparing the recognition result evaluation value with a recognition result evaluation threshold;

[0018] If the recognition result evaluation value is greater than or equal to the recognition result evaluation threshold, a recognition accuracy signal is generated.

[0019] As a further solution of the present invention: the process of constructing a knowledge graph includes:

[0020] Based on the generated accurate recognition signal, a target recognition result is obtained;

[0021] Based on the obtained target recognition results, the electronic files are classified into the first level;

[0022] Based on the first-level classification results, the electronic archives are classified into the second level;

[0023] Based on the results of primary and secondary classification, the relationships between entities are defined, and graph visualization tools are used to construct electronic archives into the form of knowledge graphs.

[0024] As a further solution of the present invention: the process of obtaining the target recognition result is:

[0025] Obtain the generated recognition accurate signal and the corresponding electronic file, obtain the same result group corresponding to the maximum value of the number of same results in the electronic file, extract the recognition result corresponding to the same result group, and mark it as the target recognition result.

[0026] As a further solution of the present invention: the process of primary classification of electronic archives is:

[0027] Obtain target recognition results of all electronic archives, and classify the electronic archives into primary categories, wherein the primary category is the subject to which the electronic archives belong;

[0028] Extract the subject information of the electronic files in the recognition results and classify the electronic files according to the categories of the first-level classification.

[0029] As a further solution of the present invention: the process of secondary classification of electronic archives is as follows:

[0030] After classifying the subject matter, the electronic archives are classified into a secondary classification, wherein the secondary classification is the time period in which the electronic archives were created;

[0031] Get the number of electronic archives in each category with the same subject, and mark them as the number of electronic archives with the same subject;

[0032] In the same subject category, obtain the creation timestamp of each electronic file and classify each electronic file according to the set time period;

[0033] In each identical topic, all comparison ratio values ​​are obtained, and the comparison ratio values ​​are compared with a first comparison ratio threshold and a second comparison ratio threshold respectively;

[0034] If the comparison ratio value is greater than or equal to the second comparison ratio threshold, a dense number signal is generated;

[0035] Based on the densely populated signals generated above, the corresponding secondary classification time period is obtained, and the current time period is shortened;

[0036] If the comparison ratio value is less than or equal to the first comparison ratio threshold, a sparse number signal is generated;

[0037] Based on the sparse signal generated above, a corresponding secondary classification time period is obtained, marked as an analysis time period, and an adjacent time period of the analysis time period is obtained, wherein the adjacent time period includes a previous time period and a subsequent time period of the analysis time period;

[0038] Obtain the comparison ratio value of the time period before the analysis time period and the comparison ratio value of the time period after the analysis time period, and compare them;

[0039] If the two values ​​are different, the time period corresponding to the smaller comparison ratio value is extracted and marked as the time period to be merged;

[0040] If the two values ​​are the same, the time period corresponding to one of the comparison ratio values ​​is randomly extracted and marked as the time period to be merged;

[0041] Merge the analysis time period with the time period to be merged to obtain the integrated time period;

[0042] If the first comparison ratio threshold is less than the comparison ratio value and the second comparison ratio threshold is less than the comparison ratio threshold, a moderate number of signals is generated.

[0043] As a further solution of the present invention: the process of obtaining the comparison ratio value is:

[0044] In the same time period category, obtain the number of electronic archives in each category, mark it as the number of the same time period, compare the number of the same time period with the number of the same subject, and obtain the proportion of the same time period, which is marked as the comparison proportion.

[0045] As a further solution of the present invention: specifically including: an electronic archive acquisition module, a recognition result acquisition processing module, a recognition result evaluation module, a knowledge graph construction module and a cloud management and control platform;

[0046] Electronic archive acquisition module: convert paper archives into electronic documents, obtain digital archive resources, and extract the content;

[0047] Recognition result acquisition and processing module: Based on the extracted content, multiple entities in the same electronic archive text are recognized, the recognition result data is obtained, processed, and the recognition result evaluation value is calculated;

[0048] Recognition result evaluation module: based on the obtained recognition result evaluation value, the recognition result of the same electronic file is evaluated, and the evaluation result includes generating a recognition accuracy signal;

[0049] Knowledge graph construction module: Based on the generated recognition accuracy signals, the knowledge graph of the recognition results of electronic archives is constructed.

[0050] Beneficial effects of the present invention:

[0051] (1) The present invention obtains multiple results by performing multiple repetitions of recognition on the same archive, and analyzes the recognition result data to calculate the maximum value of the number of identical results, the number of groups, and the number of incomplete recognitions. Based on the calculated maximum value of the number of identical results, the number of groups, and the number of incomplete recognitions, the recognition result evaluation value is calculated, and the recognition result evaluation value is compared with the recognition result evaluation threshold to generate an accurate recognition signal or an inaccurate recognition signal. Therefore, when identifying an entity, multiple repetitions of recognition can be performed, and the results can be comprehensively analyzed, thereby effectively improving the accuracy and reliability of entity recognition, and providing strong support for the construction of knowledge graphs based on digital archive resources.

[0052] (2) When generating an accurate signal of the recognition result, the present invention first obtains the target recognition result, then performs primary classification and secondary classification on the target result, and defines the relationship between entities based on the two-level classification results to construct a knowledge graph, thereby optimizing the comprehensive, refined and structured management of electronic archives through primary classification and secondary classification, improving the utilization rate and management efficiency of archive resources, and providing users with more convenient and efficient retrieval and analysis services;

[0053] (3) When performing secondary classification, the present invention further analyzes the secondary classification standard (i.e., the creation time period of the electronic archives), obtains the comparison ratio values ​​of different time periods under each same theme, evaluates the comparison ratio values, generates a dense number signal or a sparse number signal or a moderate number signal, shortens the corresponding time period based on the generated dense number signal, and merges adjacent time periods based on the generated sparse number signal. Thus, by analyzing the ratio of the number of archives within the time period, the division of the time period can be dynamically adjusted according to the actual situation, making the management more flexible and efficient, and reducing the waste of resources or management difficulties caused by unreasonable division of time periods. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The present invention will be further described below in conjunction with the accompanying drawings.

[0055] Figure 1 It is a schematic diagram of the process of constructing a knowledge graph based on digital archive resources of the present invention;

[0056] Figure 2 It is a schematic diagram of the recognition result evaluation process in the knowledge graph construction method based on digital archive resources of the present invention;

[0057] Figure 3 It is a schematic diagram of the knowledge graph construction system based on digital archive resources of the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] Example 1

[0060] See also Figure 1 - Figure 2 As shown, the present invention is a method for constructing a knowledge graph based on digital archive resources, comprising the following steps:

[0061] Step 1: Convert paper archives into electronic documents, obtain digital archive resources, and extract key content;

[0062] In some embodiments, obtaining paper files of the enterprise and scanning and converting them into electronic files;

[0063] The process of converting paper archives into electronic archives includes scanning paper archives and preprocessing scanned images:

[0064] Exemplarily, the scanning process of paper archives includes: scanning the paper archives using professional scanning equipment (such as a high-speed scanner), starting a scanning program, scanning the paper archives page by page or in batches, and converting them into digital image formats, wherein the digital image formats include but are not limited to: JPEG;

[0065] Exemplarily, the process of preprocessing the image includes: performing denoising, image quality enhancement, automatic cropping and rotation on the image;

[0066] Obtain processed electronic files and extract key content;

[0067] Exemplarily, the process of extracting the key content includes: using OCR recognition software to open the pre-processed image file, the software recognizes the text in the image through image processing and pattern recognition technology, and converts it into an editable text format (such as TXT, Word, etc.), during the OCR recognition process, the software uses a trained recognition model to analyze and match the text features to recognize the text content in the image, and then extract the key information;

[0068] Among them, key information includes but is not limited to: the domain and subject of the electronic archive, the time information of the electronic archive;

[0069] Step 2: Based on the extracted key content, perform multiple recognition of entities in the same electronic archive text, obtain recognition result data, process it, and calculate the recognition result evaluation value;

[0070] In some embodiments, a preset number of recognition times is performed, and entity recognition is repeatedly performed on the extracted key content to obtain recognition results each time;

[0071] The preset number of identification times is set by those skilled in the art based on experience;

[0072] Traverse all recognition results, mark the same recognition results as the same result group, obtain the number of the same result group, and perform ratio processing on the number of the same result group and the recognition times to obtain the group number ratio;

[0073] In each identical result group, obtain the number of recognition results, mark it as the identical result number value, perform ratio processing on the identical result number value and the recognition times to obtain the identical result number percentage value, extract the maximum value of the identical result number percentage value in all identical result groups, and mark it as the identical result number percentage maximum value;

[0074] Extract the number of incomplete identifications in the identification results, process the number of incomplete identifications and the number of identifications to obtain the ratio of incomplete identifications;

[0075] It should be explained that incomplete identification means that the entity that is expected to be identified is not completely and accurately identified, including but not limited to: entity omission, wrong identification, partial identification;

[0076] Substitute into the formula The recognition result evaluation value JG is calculated, where XT represents the maximum value of the number of identical results, ZB represents the number of groups, BW represents the number of incomplete recognitions, and a1, a2, and a3 are preset proportional coefficients;

[0077] It should be explained that the meaning reflected by the recognition result evaluation value is: the recognition result evaluation value is calculated by the maximum value of the proportion of the number of identical results, the proportion of the number of groups and the proportion of the number of incomplete recognitions, that is, the larger the maximum value of the proportion of the number of identical results, the higher the similarity of the recognition results, the larger the recognition result evaluation value, the higher the accuracy of the recognition result, that is, the smaller the proportion of the number of groups, the smaller the number of groups, the fewer different recognition results, the larger the recognition result evaluation value, the higher the accuracy of the recognition result, that is, the smaller the proportion of the number of incomplete results, the fewer the number of incomplete results, the larger the recognition result evaluation value, the higher the accuracy of the recognition result;

[0078] Step 3: Based on the obtained recognition result evaluation value, the recognition results of the same electronic file are evaluated;

[0079] In some embodiments, the recognition result evaluation value of each electronic profile is obtained, and the recognition result evaluation value is compared with a recognition result evaluation threshold, wherein the recognition result evaluation threshold is set by a person skilled in the art based on a summary of multiple historical experimental data;

[0080] If the recognition result evaluation value is greater than or equal to the recognition result evaluation threshold, it means that the entity recognition result of the electronic file is highly accurate, and an accurate recognition signal is generated;

[0081] If the recognition result evaluation value is less than the recognition result evaluation threshold, it means that the entity recognition result of the electronic record is less accurate, and an inaccurate recognition signal is generated;

[0082] Based on the inaccurate recognition signals generated above: check the content of the electronic archives that lead to inaccurate recognition, analyze text quality issues (such as ambiguity, confusing format), entity complexity (such as rare entities, polysemous words), check incompletely recognized entities: omitted, misrecognized or partially recognized entities, analyze their characteristics and patterns, and make corrections based on the inspection and analysis results;

[0083] The core technology of this embodiment is: by performing multiple repeated recognitions on the same archive to obtain multiple results, and analyzing the recognition result data, calculating the maximum value of the proportion of the number of identical results, the proportion of the number of groups, and the proportion of the number of incomplete recognitions, and based on the calculated maximum value of the proportion of the number of identical results, the proportion of the number of groups, and the proportion of the number of incomplete recognitions, calculating the recognition result evaluation value, comparing the recognition result evaluation value with the recognition result evaluation threshold, generating an accurate recognition signal or an inaccurate recognition signal, so that when identifying entities, multiple repeated recognitions can be performed and the results can be comprehensively analyzed, which effectively improves the accuracy and reliability of entity recognition and provides strong support for the construction of knowledge graphs based on digital archive resources.

[0084] Example 2

[0085] Based on the above embodiments, the present invention is a method for constructing a knowledge graph based on digital archive resources, which further includes the following steps:

[0086] Step 4: Based on the evaluation results, construct a knowledge graph of the electronic archives recognition results;

[0087] The process of constructing a knowledge graph includes:

[0088] Based on the generated accurate recognition signal, a target recognition result is obtained;

[0089] Based on the obtained target recognition results, the electronic files are classified into the first level;

[0090] Based on the first-level classification results, the electronic archives are classified into the second level;

[0091] Based on the primary and secondary classification results, define the relationship between entities and build a knowledge graph;

[0092] In some embodiments, the process of obtaining the target recognition result is:

[0093] Obtain the generated recognition accurate signal and the corresponding electronic file, obtain the same result group corresponding to the maximum number of same results in the electronic file, extract the recognition result corresponding to the same result group, and mark it as the target recognition result;

[0094] It should be explained that the maximum percentage of the number of identical results means that the result with the largest percentage of identical results among the recognition results obtained by multiple recognitions is the most accurate one among the recognition results, and this result is the target recognition result to be obtained;

[0095] In some embodiments, the process of first-level classification of electronic files is:

[0096] Obtain target recognition results of all electronic archives, and classify the electronic archives into primary categories, wherein the primary category is the subject to which the electronic archives belong;

[0097] Extracting subject information of the electronic archives in the recognition results, and classifying the electronic archives according to the categories of the primary classification;

[0098] It should be noted that the subject of the field is determined according to the scenario in which the electronic archive is used;

[0099] For example, if applied to business management, the topics include: contracts, meeting minutes, financial statements, employee files, etc.;

[0100] In some embodiments, the process of secondary classification of electronic files is as follows:

[0101] After classifying the subject matter, the electronic archives are classified into a secondary classification, wherein the secondary classification is the time period in which the electronic archives were created;

[0102] Exemplary settings of the time period include, but are not limited to: one year, one quarter, one month;

[0103] Get the number of electronic archives in each category with the same subject, and mark them as the number of electronic archives with the same subject;

[0104] In the same subject category, obtain the creation timestamp of each electronic file and classify each electronic file according to the set time period;

[0105] In the same time period category, obtain the number of electronic archives in each category, mark it as the number of the same time period, compare the number of the same time period with the number of the same subject, and obtain the proportion of the same time period, mark it as the comparison proportion;

[0106] It should be explained that the meaning of the comparison ratio is: the comparison ratio is calculated by the number of electronic files in the same time period category. The number of electronic files in the same time period category is the number of the same time period. That is, the larger the number of the same time period, the more electronic files there are in the time period and the denser the number.

[0107] In each identical subject, all comparison ratio values ​​are obtained, and the comparison ratio values ​​are compared with a first comparison ratio threshold and a second comparison ratio threshold, respectively, wherein the first comparison ratio threshold and the second comparison ratio threshold are set by a person skilled in the art based on a summary of multiple historical experimental data, and the first comparison ratio threshold is less than the second comparison ratio threshold;

[0108] If the comparison ratio value is greater than or equal to the second comparison ratio threshold, it means that the number of electronic files in this time period is large, and a dense number signal is generated;

[0109] Based on the densely populated signals generated above, the corresponding secondary classification time period is obtained, and the current time period is shortened;

[0110] For example, if the current time period is one year, it is reduced to one quarter, and if the current time period is one month, it is reduced to one week;

[0111] If the comparison ratio value is less than or equal to the first comparison ratio threshold, it means that the number of electronic files in this time period is small, and a sparse number signal is generated;

[0112] Based on the sparse signal generated above, a corresponding secondary classification time period is obtained, marked as an analysis time period, and an adjacent time period of the analysis time period is obtained, wherein the adjacent time period includes a previous time period and a subsequent time period of the analysis time period;

[0113] Obtain the comparison ratio value of the time period before the analysis time period and the comparison ratio value of the time period after the analysis time period, and compare them;

[0114] If the two values ​​are different, the time period corresponding to the smaller comparison ratio value is extracted and marked as the time period to be merged;

[0115] If the two values ​​are the same, the time period corresponding to one of the comparison ratio values ​​is randomly extracted and marked as the time period to be merged;

[0116] Merge the analysis time period with the time period to be merged to obtain the integrated time period.

[0117] If the first comparison ratio threshold is less than the comparison ratio value and the second comparison ratio threshold is less than the comparison ratio value, it means that the number of electronic archives in the time period is medium, and a moderate number signal is generated;

[0118] In some embodiments, the process of defining the relationship between entities and constructing the knowledge graph is as follows:

[0119] Define the relationships between entities and use graph databases or graph visualization tools to build electronic archives into a knowledge graph;

[0120] Among them, the relationship between entities may include, but is not limited to: affiliation (e.g. a contract belongs to a certain employee);

[0121] The core technology of this embodiment is: when generating an accurate signal of the recognition result, first obtain the target recognition result, then perform primary classification and secondary classification on the target result, and define the relationship between entities based on the two-level classification results to build a knowledge graph, so that the comprehensive, refined and structured management of electronic archives can be optimized through primary classification and secondary classification, and the utilization rate and management efficiency of archive resources can be improved. At the same time, more convenient and efficient retrieval and analysis services can be provided to users, and combined retrieval can be performed according to the subject and time dimensions to quickly locate the required electronic archives;

[0122] At the same time, when conducting secondary classification, the secondary classification standard (i.e., the creation time period of the electronic archives) is further analyzed to obtain the comparison ratio values ​​of different time periods under each same theme, and the comparison ratio values ​​are evaluated to generate a dense number signal, a sparse number signal, or a moderate number signal. Based on the generated dense number signal, the corresponding time period is shortened, and based on the generated sparse number signal, adjacent time periods are merged. By analyzing the proportion of the number of archives in the time period, the division of time periods can be dynamically adjusted according to actual conditions, making management more flexible and efficient, and reducing resource waste or management difficulties caused by unreasonable division of time periods.

[0123] Example 3

[0124] Based on the above examples, please refer to Figure 3 As shown, the present invention is a knowledge graph construction system based on digital archive resources, which specifically includes: an electronic archive acquisition module, a recognition result acquisition processing module, a recognition result evaluation module, a knowledge graph construction module and a cloud management and control platform;

[0125] Electronic archive acquisition module: convert paper archives into electronic documents, obtain digital archive resources, and extract key content;

[0126] Recognition result acquisition and processing module: Based on the extracted key content, multiple entities in the same electronic archive text are recognized, the recognition result data is obtained, processed, and the recognition result evaluation value is calculated;

[0127] Recognition result evaluation module: based on the obtained recognition result evaluation value, the recognition result of the same electronic file is evaluated;

[0128] Knowledge graph construction module: Based on the evaluation results, a knowledge graph is constructed for the recognition results of electronic archives.

[0129] The size of the above threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the cardinality set by the technicians in this field for each group of sample data. For example, in the actual process, there are many groups of group number percentage values, maximum number of identical results percentage values, and incomplete recognition percentage values. The group number percentage values, maximum number of identical results percentage values, and incomplete recognition percentage values ​​of many groups are processed to obtain the recognition result evaluation values ​​of the corresponding groups. The staff evaluates the recognition result accuracy of the electronic file based on the recognition result evaluation values ​​of so many groups, thereby obtaining a corresponding relationship between the recognition result evaluation value and the recognition result accuracy of the electronic file, and then derives and divides the recognition result evaluation value threshold according to the recognition result accuracy of the electronic file, thereby obtaining the recognition result evaluation value threshold, and compares the obtained recognition result evaluation value with the recognition result evaluation value threshold, that is, completes the recognition of the recognition result accuracy of the electronic file corresponding to the recognition result evaluation value.

[0130] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A knowledge graph construction method based on digital archive resources, characterized in that: The following steps are involved: Step 1: Convert paper archives into electronic documents, obtain digital archive resources, and extract the content; Step 2: Based on the extracted content, perform multiple recognition of entities in the same electronic archive text, obtain recognition result data, process it, and calculate the recognition result evaluation value; Step 3: Based on the obtained recognition result evaluation value, the recognition result of the same electronic file is evaluated, and the evaluation result includes generating a recognition accuracy signal; Step 4: Based on the generated accurate recognition signal, a knowledge graph is constructed for the recognition result of the electronic archive. The process of constructing the knowledge graph includes: Obtaining the target recognition result, the specific process is: obtaining the generation of accurate recognition signals and the corresponding electronic files, obtaining the same result group corresponding to the maximum number of the same results of the electronic files, extracting the recognition result corresponding to the same result group, and marking it as the target recognition result; Based on the obtained target recognition results, the electronic files are classified into the first level; The process of primary classification is: obtaining the target recognition results of all electronic archives, and classifying the electronic archives into primary categories, where the primary classification is the subject to which the electronic archives belong; Extracting subject information of the electronic archives in the recognition results, and classifying the electronic archives according to the categories of the primary classification; Based on the first-level classification results, the electronic archives are classified into the second level; The process of secondary classification is as follows: after classifying the subject matter, the electronic archives are classified into secondary categories, where the secondary classification is the time period when the electronic archives were created; Get the number of electronic archives in each category with the same subject, and mark them as the number of electronic archives with the same subject; In the same subject category, obtain the creation timestamp of each electronic file and classify each electronic file according to the set time period; Based on the results of primary and secondary classification, the relationships between entities are defined, and graph visualization tools are used to construct electronic archives into the form of knowledge graphs.

2. The method for constructing a knowledge graph based on digital archive resources according to claim 1, characterized in that: The process of obtaining the recognition result evaluation value is as follows: Get the maximum value of the percentage of identical results, the percentage of groups, and the percentage of incomplete identifications; Substitute into the formula , the recognition result evaluation value JG is calculated, where XT represents the maximum value of the number of identical results, ZB represents the number of groups, BW represents the number of incomplete recognitions, and a1, a2, and a3 are preset proportional coefficients.

3. The method for constructing a knowledge graph based on digital archive resources according to claim 2, characterized in that: The process of obtaining the maximum value of the number of identical results, the number of groups, and the number of incomplete identifications is as follows: Preset the number of recognition times, repeatedly perform entity recognition on the extracted key content, and obtain the recognition results each time; Traverse all recognition results, mark the same recognition results as the same result group, obtain the number of the same result group, and perform ratio processing on the number of the same result group and the recognition times to obtain the group number ratio; In each identical result group, obtain the number of recognition results, mark it as the identical result number value, perform ratio processing on the identical result number value and the recognition times to obtain the identical result number percentage value, extract the maximum value of the identical result number percentage value in all identical result groups, and mark it as the identical result number percentage maximum value; The number of incomplete identifications in the identification results is extracted, and the number of incomplete identifications is processed with the number of identifications to obtain the ratio of incomplete identifications.

4. The method for constructing a knowledge graph based on digital archive resources according to claim 1, characterized in that: The process of generating an accurate identification signal is: Obtaining a recognition result evaluation value of each electronic file, and comparing the recognition result evaluation value with a recognition result evaluation threshold; If the recognition result evaluation value is greater than or equal to the recognition result evaluation threshold, a recognition accuracy signal is generated.

5. The method for constructing a knowledge graph based on digital archive resources according to claim 1, characterized in that: The process of secondary classification of electronic archives also includes: In each identical topic, all comparison ratio values ​​are obtained, and the comparison ratio values ​​are compared with a first comparison ratio threshold and a second comparison ratio threshold respectively; If the comparison ratio value is greater than or equal to the second comparison ratio threshold, a dense number signal is generated; Based on the densely populated signals generated above, the corresponding secondary classification time period is obtained, and the current time period is shortened; If the comparison ratio value is less than or equal to the first comparison ratio threshold, a sparse number signal is generated; Based on the sparse signal generated above, a corresponding secondary classification time period is obtained, marked as an analysis time period, and an adjacent time period of the analysis time period is obtained, wherein the adjacent time period includes a previous time period and a subsequent time period of the analysis time period; Obtain the comparison ratio value of the time period before the analysis time period and the comparison ratio value of the time period after the analysis time period, and compare them; If the two values ​​are different, the time period corresponding to the smaller comparison ratio value is extracted and marked as the time period to be merged; If the two values ​​are the same, the time period corresponding to one of the comparison ratio values ​​is randomly extracted and marked as the time period to be merged; Merge the analysis time period with the time period to be merged to obtain the integrated time period; If the first comparison ratio threshold is less than the comparison ratio value and the second comparison ratio threshold is less than the comparison ratio value, a moderate number of signals is generated.

6. The method for constructing a knowledge graph based on digital archive resources according to claim 5 is characterized in that: The process of obtaining the comparison ratio value is: In the same time period category, obtain the number of electronic archives in each category, mark it as the number of the same time period, compare the number of the same time period with the number of the same subject, and obtain the proportion of the same time period, which is marked as the comparison proportion.

7. A knowledge graph construction system based on digital archive resources, the system implementing the method according to any one of claims 1 to 6, characterized in that: Specifically include: Electronic archive acquisition module, recognition result acquisition and processing module, recognition result evaluation module, knowledge graph construction module and cloud management and control platform; Electronic archive acquisition module: convert paper archives into electronic documents, obtain digital archive resources, and extract the content; Recognition result acquisition and processing module: Based on the extracted content, multiple entities in the same electronic archive text are recognized, the recognition result data is obtained, processed, and the recognition result evaluation value is calculated; Recognition result evaluation module: based on the obtained recognition result evaluation value, the recognition result of the same electronic file is evaluated, and the evaluation result includes generating a recognition accuracy signal; Knowledge graph construction module: Based on the generated recognition accuracy signals, the knowledge graph of the recognition results of electronic archives is constructed.

Images