Knowledge graph-based human resource archive storage management system

Abstract

The present application relates to the technical field of human resource management, in particular to a human resource archive storage management system based on a knowledge graph, which comprises a post identification module, a path modeling module, a node linkage module, a fracture identification module and an access control module.In the present application, through the normalized mapping of post description and skill combination, the standardization processing capability of unstructured archive content is improved, a stage label system that conforms to the post responsibility process is constructed in combination with the jump frequency and span characteristics of the multi-stage path nodes in the task record, the cross correlation of the responsibility fields between the nodes is identified by using the tenure cycle division result, the path structure mapping sequence is dynamically adjusted, the change situation of the responsibility transmission relationship in the structure path is accurately identified, and in the continuous vacancy path identification, the responsibility chain interruption range is judged by the field group difference, and based on the sensitive field distribution characteristics and the role field coverage ratio in the access path, the access permission judgment and the permission checking control at the path level are realized.

Description

Technical Field

[0001] This invention relates to the field of human resource management technology, and in particular to a human resource file storage and management system based on knowledge graphs. Background Technology

[0002] The field of human resource management technology encompasses the systematic planning, organization, and coordination of personnel-related affairs within an organization, and the automated processing of these processes. It focuses on using information technology to manage the entire process of human resources, including recruitment, hiring, training, performance evaluation, allocation, motivation, and personnel file management. This includes human resource information collection, person-job relationship management, job responsibility allocation, employee resume recording, and access control. It is systematically built upon standardized data structures, rule-driven modeling methods, and dynamic correlation of person-job matching relationships. With the continuous development of management needs, it integrates mechanisms such as semantic recognition, relationship construction, and structured management to improve information integration and management efficiency. Among these, knowledge graph-based human resource management is particularly important. The Human Resources Archives Storage and Management System refers to the process of transforming unstructured information in human resources archives into structured graph data through semantic modeling and graph structure mapping, and achieving centralized and unified storage and management. This includes the identification of key entities in personnel files, attribute classification, semantic understanding of archive content, graph construction of employee information and organizational relationships, establishment of entity relationship paths, and query processing. Specifically, it involves using rule-driven natural language parsing to identify entities and attribute content in archive text, using preset triple generation logic to transform personnel information into graph nodes and relationship edges, and performing unified storage and retrieval management of information based on a graph database, thus completing the structured organization and graph representation of human resources archive information.

[0003] Traditional human resource record storage and management technologies rely on standardized modeling and graph mapping as their core construction methods. However, they lack the ability to provide detailed descriptions of changes in responsibilities at key nodes in multi-stage task recording and dynamic job transitions. They cannot effectively identify structural migration trends caused by overlapping responsibility fields during job path evolution, resulting in a lack of timely warnings when responsibilities are fragmented or processes are interrupted. In terms of path access management, they rely solely on static role field configurations and lack a mechanism for linking path structure with the distribution of sensitive fields. In scenarios involving high-frequency job changes or access penetration, they are prone to sensitive information leakage or blind spots in access configuration, affecting the accuracy and security of record access control. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a knowledge graph-based human resource record storage and management system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A knowledge graph-based human resource file storage and management system includes:

[0006] The job identification module calls employee file information, analyzes the combination of job description phrases and skill keywords, compares the differences in field order and word usage between the job phrases and the standard template, performs expression mapping between job terminology and the standard template, and generates job terminology normalization results.

[0007] Based on the job terminology normalization results, the path modeling module calculates the jump frequency and node span differences of path nodes in employee task records at multiple stages, analyzes the alignment of path connections with job responsibilities and processes, filters node reconstruction segments, establishes stage labels, and generates job cycle division results.

[0008] The node linkage module uses the job cycle division results to determine the intersection of responsibility fields between job nodes, filter fields with cross-path connection, adjust the mapping order, calculate the crossover of upstream and downstream responsibilities, construct structural change description tags, and generate path evolution feature information.

[0009] The fracture identification module combines the path evolution feature information to construct the job connection path, analyze the length of the continuous vacant job node group and the path ratio, determine the difference in the responsibility field of upstream and downstream nodes, delineate the scope of responsibility interruption, and generate structural risk identification results.

[0010] Based on the structural risk identification results, the access control module analyzes the distribution of sensitive fields in the access path, calls the role field group to compare the field matching coverage ratio, determines the path access permission status, and generates file access management results.

[0011] As a further aspect of the present invention, the job terminology normalization result includes standard job field, skill attribution label, and expression mapping structure; the tenure cycle division result includes stage behavior identifier, task span range, and path segment number; the path evolution feature information includes responsibility migration label, node connection order, and evolution path structure; the structural risk identification result includes vacant job connection segment, responsibility breakpoint location, and link interruption range; and the file access management result includes permission path label, sensitive field density, and access judgment level.

[0012] As a further aspect of the present invention, the job identification module includes:

[0013] The job description phrase extraction submodule obtains employee file information, extracts job description phrases and skill keywords, analyzes the combination of phrases and keywords, establishes a job skill dataset, and generates job phrase data.

[0014] The structure format comparison submodule compares the differences in field order and word usage between the job term data and the standard term template, based on the amount of job term data. It determines the order in which keywords appear in the template, filters fields that match the template structure, and generates the job structure difference quantity.

[0015] The terminology mapping adjustment submodule calls the job structure difference quantity, adjusts the original job phrase descriptions according to the combination matching results between jobs and skills, maps them to standard template expressions, and obtains job terminology normalization results.

[0016] As a further aspect of the present invention, the path modeling module includes:

[0017] The node frequency analysis submodule obtains the normalization result of the job terminology, counts the frequency of occurrence of path nodes in each stage of the employee task record and the number of stage nodes, analyzes the sequential distribution of nodes in each stage, establishes the node occurrence change range, and generates the stage jump change amount.

[0018] The task path evaluation submodule calls the stage jump change amount, compares the connection order between task path nodes and key nodes of the job standard task process, filters the continuity of node distribution and missing segments, judges the co-occurrence ratio of actual task nodes and standard path nodes, and obtains the critical path fit coefficient.

[0019] The behavior segmentation construction submodule monitors the distribution of node reconstruction behavior in the time series based on the critical path fit coefficient, collects the cumulative frequency, number of segment nodes, node time interval and span of each behavior type in the periodic time period, calculates the node behavior change index, groups the behavior types and divides them into periodic segments, forms a structured job cycle segmentation identifier, and establishes the job cycle division result.

[0020] As a further aspect of the present invention, the node linkage module includes:

[0021] The responsibility intersection extraction submodule, based on the job cycle division results, calls the responsibility field set in the original job node and the transferred job node, calculates the overlap ratio between the number of matching content items of the responsibility field in the text expression and the semantic range expression vector of the field, constructs the intersection expression vector based on the coverage rate of the responsibility items and the content similarity score, obtains the overlapping expression set and judges the consistency of information structure, and obtains the job responsibility intersection ratio.

[0022] The cross-job mapping submodule filters field groups with path continuity markers based on the job responsibility intersection ratio, detects the original position labels of the target field group in the job path structure, adjusts the mapping order and judges the degree of overlap of field positions after remapping, extracts overlapping structure segments and counts the range of connected paths, and obtains the cross-node field connection density.

[0023] The path change identification submodule calculates the number of differences, the degree of change in connection order, and the difference in expression span between the binding responsibility fields of upstream and downstream nodes in the structural path based on the cross-node field connection density. It also calculates the responsibility expression order offset density in the job path, constructs structural change description tags between the reconstructed path structure and the original path structure, and generates path evolution feature information.

[0024] As a further aspect of the present invention, the fracture detection module includes:

[0025] The vacant post section identification submodule obtains the path evolution feature information, detects the continuous node groups without configured personnel binding fields in the post connection path, calculates the length of the node group, and compares it with the total path length to obtain the vacant post continuity ratio.

[0026] The responsibility field comparison submodule collects the responsibility field set of upstream and downstream connecting nodes of the vacant post section based on the vacant post continuity ratio, analyzes the number of field set items, judges the difference between the number and the responsibility configuration benchmark value, and obtains the responsibility configuration offset.

[0027] The fracture interval determination submodule, based on the responsibility configuration offset, combined with the node distribution and responsibility field coverage in the job connection path, filters out locations where there are fractures in the responsibility field coverage, marks the fracture segment range in the node sequence, and establishes structural risk identification results.

[0028] As a further aspect of the present invention, the access control module includes:

[0029] The cumulative statistics submodule obtains the structural risk identification results, analyzes the salary field, performance evaluation field and organizational function field in the access path, counts the cumulative number of occurrences of each type of field in the path, and obtains the total number of path-sensitive fields.

[0030] The role permission comparison submodule, based on the total number of path sensitive fields, calls the access level field group corresponding to the user role, compares the coverage ratio of the number of sensitive fields with the number of accessible fields in the role field set, and obtains the permission coverage ratio.

[0031] The path permission determination submodule determines the matching relationship between sensitive fields and role field sets in the path node based on the permission coverage ratio, marks the access permission status of the path, and establishes the file access management result.

[0032] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0033] In this invention, the standardization of unstructured file content is improved by normalizing the mapping between job descriptions and skill combinations. By combining the jump frequency and span characteristics of multi-stage path nodes in task records, a stage label system that fits the job responsibility process is constructed. The cross-relationship of responsibility fields between nodes is identified by the results of the job cycle division. The path structure mapping order is dynamically adjusted to accurately identify changes in the responsibility transmission relationship in the structure path. In the identification of continuous vacant positions, the scope of the responsibility chain interruption is judged by the differences in field groups. Based on the distribution characteristics of sensitive fields in the access path and the coverage ratio of role fields, path-level access permission judgment and permission verification control are realized. Attached Figure Description

[0034] Figure 1 This is a system flowchart of the present invention;

[0035] Figure 2 This is a flowchart of the job identification module of the present invention;

[0036] Figure 3 This is a flowchart of the path modeling module of the present invention;

[0037] Figure 4 This is a flowchart of the node linkage module of the present invention;

[0038] Figure 5 This is a flowchart of the fracture recognition module of the present invention;

[0039] Figure 6 This is a flowchart of the access control module of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0041] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0042] Please see Figure 1 The knowledge graph-based human resource record storage and management system includes:

[0043] The job identification module calls employee file information, analyzes the combination of job description phrases and skill keywords, compares the differences in field order and word usage between the job phrases and the standard template, performs expression mapping between job terminology and the standard template, and generates job terminology normalization results.

[0044] The path modeling module calculates the jump frequency and node span differences of path nodes in multiple stages in the employee task record based on the job terminology normalization results, analyzes the alignment of path connections with job responsibilities and processes, filters node reconstruction fragments, establishes stage labels, and generates job cycle division results.

[0045] The node linkage module uses the job cycle division results to determine the intersection of responsibility fields between job nodes, filter fields with cross-path connection, adjust the mapping order, calculate the crossover of upstream and downstream responsibilities, construct structural change description tags, and generate path evolution feature information.

[0046] The fracture identification module combines path evolution feature information to construct job connection paths, analyze the length of continuous vacant job node groups and path ratios, determine the differences in responsibility fields between upstream and downstream nodes, delineate the scope of responsibility interruption, and generate structural risk identification results.

[0047] Based on the structural risk identification results, the access control module analyzes the distribution of sensitive fields in the access path, calls the role field group to compare the field matching coverage ratio, determines the path access permission status, and generates file access management results.

[0048] The job terminology standardization results include standard job fields, skill attribution tags, and expression mapping structures. The job cycle division results include stage behavior identifiers, task span ranges, and path segment numbers. The path evolution characteristic information includes responsibility migration tags, node connection order, and evolution path structure. The structural risk identification results include vacant job connection segments, responsibility breakpoint locations, and link interruption ranges. The file access management results include permission path tags, sensitive field density, and access judgment level.

[0049] Please see Figure 2 The job identification module includes:

[0050] The job description phrase extraction submodule obtains employee file information, extracts job description phrases and skill keywords, analyzes the combination of phrases and keywords, establishes a job skill dataset, and generates job phrase data.

[0051] Obtaining employee file information refers to retrieving structured and unstructured fields from the personnel information system, including the employee's current job title, job description, work records over the years, and key performance indicators. The extraction of job description phrases is performed through a two-step process combining regular expressions and keyword matching. First, using the "job title" and "job description" fields as target fields, combinations of verbal and noun phrases appearing more than 5 times are extracted as candidate job description phrases. For example, from "responsible for developing sales plans and maintaining customer resources," the phrases "develop sales plans" and "maintain customer resources" are extracted and added to the database. The next step in extracting skill keywords involves using phrases like "required skills" and "job requirements." The noun and adjective combinations appearing in the "Request" and "Assessment Criteria" fields are used as target filtering objects. Based on this, an initial set of skill keywords is established. For example, if a job description mentions "data analysis ability" and "communication and coordination ability," these are segmented into "data analysis" and "communication and coordination," and their frequencies are recorded as 9 and 5, respectively. Then, using the job ID as the index field, a job-skill combination structure is constructed to generate a job skill dataset containing the correspondence between job description phrases and skill keywords. In this dataset, each job record is mapped to a specific skill set and descriptive phrase combination, forming a key-value structure mapping. The number of phrases extracted from all job descriptions is counted, totaling 1865. The following are some sample job information:

[0052] Table 1. Examples of Job-Related Phrases and Keywords Extracted

[0053] ;

[0054] As shown in Table 1, the frequency of descriptive phrases and keywords extracted from different job positions varies. These data serve as the basis for frequency threshold judgment and reordering operations in subsequent structural comparison and term mapping, ultimately generating 1865 job phrase data entries.

[0055] The structure format comparison submodule compares the differences in field order and word usage between the job terminology data and the standard terminology template, based on the amount of job terminology data. It determines the order in which keywords appear in the template, filters fields that match the template structure, and generates the job structure difference quantity.

[0056] Based on the amount of job terminology data, the structure of each job terminology was compared with the standard terminology template in terms of field order and vocabulary content. In the field order comparison, the three fields "verb + noun + modifier" from the standard template were selected as the reference order. The actual job terminology was split and reordered to determine the degree of deviation from the standard order. If a job description is "resource maintenance customer," and there is an order difference from the standard "maintenance + customer + resource," it is recorded as one deviation. The frequency of deviation field position differences was counted. If the deviation position is at the beginning, it is counted as one type of deviation; if the deviation position is at the end, it is classified into another type of deviation. In the field content difference comparison, the Jaccard similarity between the keywords appearing in each terminology and the standard template terminology was calculated. If the template terminology is "analyze sales data" and the job terminology is "statistically analyze sales information," then their keyword intersection is "sales," and their union is "analysis, sales, data, statistics, information," with a similarity of [value missing]. If the similarity is below the set threshold of 0.6, it is considered a mismatch. The order of keywords in the template is mapped position by position to generate a sequence difference. Using "make a plan to maintain customers" as the original phrase and "plan to make customer maintenance" as the standard phrase, the original sequence index is [1,2,3], and the standard index is [2,1,3]. The average displacement is calculated as follows: The system determines whether the value is greater than the structural offset baseline of 0.5. If it is, it is recorded as a high displacement difference item. During the screening and matching field process, the system extracts the words that are exactly the same as the standard template field from each job word group, records them as matching field items, and counts the total number of matches. The system iterates through all jobs and finally generates the job structure difference quantity. The total number of difference word groups is 783, accounting for 42% of the total number of job word groups.

[0057] The terminology mapping adjustment submodule calls the job structure difference quantity, adjusts the original job phrase descriptions based on the combination matching results between jobs and skills, maps them to standard template expressions, and obtains job terminology normalization results;

[0058] The system retrieves the 783 difference items marked in the job structure difference data, traces back to the corresponding job skill dataset records by job ID, and performs word group reordering and keyword adjustment operations on each job data record based on the comparison results of job description phrases and skill keyword combinations. In the order adjustment operation, the index order of each word group in the original job phrase is read, and the phrase fields are reordered according to the maximum overlap principle, referring to the template order with the highest matching degree in the standard template. For example, if the original phrase is "planning and organizing activities" and the corresponding template is "organizing and planning activities," then the original phrase is adjusted to "organizing and planning activities." In the keyword adjustment operation, the frequency of skill keywords is used as the primary order sorting weight, with a threshold frequency of 5. Words with a frequency below 5 are not included in the keyword adjustment target group. Keywords with a frequency of 5 or higher, totaling 412, were selected. These keywords were inserted, replaced, or deleted according to their corresponding positions in the job descriptions. Standard expression formats were then constructed within the adjusted job description phrases. During the normalization mapping process, the adjusted phrases were searched item by item using the standard job terminology database. The closest standard job expression was matched according to the principle of minimum edit distance. The maximum character distance threshold was set to 3. If the edit distance between the original description "maintain customer resources" and the standard expression "maintain customer resources" was 3 (character replacement + addition), the mapping condition was met, and the mapping result was "maintain customer resources". A total of 1865 job terminology normalization results were obtained, of which 783 were adjusted items and 1082 were items that did not require adjustment.

[0059] Please see Figure 3 The path modeling module includes:

[0060] The node frequency analysis submodule obtains the job terminology normalization results, counts the frequency of occurrence of path nodes in each stage of employee task records and the number of stage nodes, analyzes the sequential distribution of nodes in each stage, establishes the node occurrence change range, and generates stage jump change amount.

[0061] Extract the standardized job identifier code and related skill fields corresponding to each employee at different time periods from the processed normalized job terminology database, and establish a mapping table with task records. To statistically analyze the frequency of path nodes in each stage of an employee's task record, the employment period is divided into fixed windows, with a time period length of 30 days. Each employee's task record is divided into several stages based on timestamps. For example, employee A recorded five task nodes in period 1: "data collection, data cleaning, modeling training, result analysis, and report writing." The frequency of each node in this stage is statistically analyzed. Assuming "data collection" appears 4 times and "modeling training" appears 1 time, the frequency array is [4,1,1,1,1], and the number of nodes recorded in this stage is 5. To analyze the node distribution order of each stage, the task nodes are sorted in ascending order by timestamp and compared with the node order in the standard job task flow. If a node is ranked 2nd in the standard flow but 4th in the actual record, its order offset value is 2. Traverse all nodes to obtain the total sequence offset within the stage, and establish the position change range of each node in different stages. For example, "Report Writing" appears last in stage 1 and second to last in stage 2, so its change range is [5,4]. After recording the set of position change ranges of all nodes in the whole cycle, combine the difference in the order of occurrence and the difference in frequency of all nodes in two adjacent stages, sum them, and normalize them to finally obtain the stage jump change between each stage. If the jump change between stage 1 and stage 2 is 11, then it is recorded as the change intensity value of stage pair (1→2).

[0062] The task path evaluation submodule calls the jump change amount, compares the connection order between task path nodes and key nodes of the job standard task process, filters the continuity of node distribution and missing segments, judges the co-occurrence ratio of actual task nodes and standard path nodes, and obtains the critical path fit coefficient.

[0063] Using the aforementioned stage transition changes, the task path nodes for each employee are compared item by item with the key nodes of the corresponding position's standard task flow. When comparing the order, a numbered index list is created for the standard flow nodes; for example, "Data Collection-Cleaning-Modeling-Analysis-Reporting" corresponds to numbers [1,2,3,4,5]. The actual path nodes are numbered according to the task order as [1,3,2,4,5]. Each node is compared sequentially to see if it follows the standard order dependency, and the continuity of the matching segments is statistically analyzed. If the order of numbers is reversed or jumps by more than two positions, it is determined to be a discontinuous segment. Based on this, the node numbers of the skipped segments are selected and included in the discontinuous node set. For example, if node [2,3] jumps to a standard discontinuous interval, it is recorded as one skipped segment. Then, the co-occurrence ratio between the actual task nodes and the standard path nodes is determined. Assuming the standard flow contains 8 key task nodes, and an employee completes 5 of them in one stage, the co-occurrence ratio is... A multi-stage averaging method is used, averaging the co-occurrence ratios of each stage throughout the entire lifecycle as the critical path fit coefficient. If an employee's co-occurrence ratios across the three stages are 0.625, 0.75, and 0.5, then their fit coefficient is... Finally, the fit coefficient between the employee's task path and the standard job task process is recorded.

[0064] The behavior segmentation construction submodule monitors the distribution of node reconstruction behaviors in the time series based on the critical path fit coefficient, collecting the cumulative frequency, number of segmented nodes, node time interval, and span for each behavior type within a periodic time period, using the formula:

[0065] ;

[0066] Calculate the node behavior change index, group the behavior types and divide them into periodic segments to form a structured employment period segmentation identifier and establish the employment period division results;

[0067] in, For the first The frequency normalization value of each path node is obtained by counting the number of times the node appears in the task records within the term of the node and dividing by the maximum node frequency in the same term. For the first The normalized value of the number of path nodes within a period segment is obtained by counting the number of nodes within that period segment and dividing by the maximum number of nodes in the entire period. For the first The normalized time interval between each node and its predecessor is obtained by calculating the difference in timestamps of the nodes and dividing it by the maximum time difference over the entire period. For the first The normalized span value of the path to which each node belongs within this period is obtained by calculating the span of the node index in the path and dividing it by the maximum span of the entire period. The total number of behavior nodes within the current behavior cycle segment is obtained by directly counting the number of nodes within this segment. This is a node behavior change index, representing the degree of change in the behavior sequence within a period, used for subsequent division of behavior stages. This is the index number of the behavior node within the periodic segment;

[0068] Based on the critical path fit coefficient, employee task path stages with a fit value below 0.6 are designated as key monitoring targets. In these stages, node restructuring behavior is identified, referring to actions such as reversed order, merging, disassembling, or replacement of the actual path node sequence compared to the standard process. Restructuring behavior is determined by comparing the order of each node with the timestamp interval. Timestamps corresponding to all nodes are collected in the time series, and the frequency of each behavior type is statistically analyzed by stage. For example, if an employee's "node merging" occurs 6 times and "node replacement" occurs 2 times within a 3-month period, then their behavior type frequency array is [6,2]. For each behavior period, the number of nodes in the segment, the average time interval between nodes, and the span are recorded. The number of nodes is directly calculated as the total number of task nodes within the behavior segment. The time interval is calculated by averaging the differences in node timestamps, and the span is the maximum value minus the minimum value of the node index. Subsequently, the above parameters are normalized and substituted into the following formula:

[0069] ;

[0070] In the formula, Indicates the first The normalized frequency value of a node within a period is obtained by dividing the number of times the node appears in the periodic task by the maximum frequency in the same period. For the first The normalized value of the number of path nodes in each period segment is obtained by dividing the number of nodes in that period segment by the maximum number of nodes in the entire period. For the first The normalized time interval between each node and its predecessor is calculated by dividing the timestamp difference by the maximum time difference. This is the normalized value of the path span of the node within this period; This represents the total number of nodes within the current action segment.

[0071] Suppose a certain behavior segment has 3 nodes, and record their frequency, time interval and path span as shown in the table below.

[0072] Table 2 Behavior segmentation normalization parameter table

[0073] ;

[0074] As shown in Table 2, the original frequency, time difference, and span values ​​of each node were used to calculate the normalized parameters in subsequent calculations. The normalized parameters are as follows: , , , ; , , ; , , Substituting into the formula, we get:

[0075] ;

[0076] The results of the calculations are as follows: The first item is: The second item is: The third item is: ;

[0077] ;

[0078] The node behavior change index is an indicator used to measure the degree of change in node tasks within an employee's job behavior path over a specific time period. Specifically, within a structured knowledge graph path, the higher the density, the more compact the time distribution, and the more concentrated the span of task behavior nodes within a given time period, the larger the node behavior change index value for that period. This indicates that the employee has undergone concentrated task restructuring or responsibility switching behavior within that period. Conversely, if the node frequency is dispersed, the path span is wide, and the behavior span is weak, the index value is smaller, representing strong behavioral stability within the period. This parameter is directly used to construct the time profile of the employment cycle and is a key data foundation for subsequent node linkage identification and path evolution determination. It can be used to dynamically reveal employee behavioral rhythms and job change potential. The results show that the node behavior change index... Since the threshold for behavioral stage variation is set at 0.3, stages below this value are classified as "stable stages," while those above 0.6 are classified as "stages of drastic change." Therefore, this period is determined to be a "moderate change stage." Based on this, a segmentation identifier for the tenure cycle is established, and the cycle is recorded in the personnel file as "Stage B (Moderate Restructuring)."

[0079] Please see Figure 4 The node linkage module includes:

[0080] The responsibility intersection extraction submodule calls the responsibility field set in the original position node and the transferred position node according to the job cycle division result, calculates the overlap ratio between the number of matching content items of the responsibility field in the text expression and the semantic range expression vector of the field, constructs the intersection expression vector according to the coverage of responsibility items and the content similarity score, obtains the overlapping expression set and judges the consistency of information structure, and obtains the job responsibility intersection ratio.

[0081] Based on the job tenure cycle, the system retrieves the employee's original and new job nodes within a specific cycle and extracts their corresponding responsibility field sets. When extracting fields, the "core responsibility segment" field in the job description and responsibility record table is used as the target field. Independent responsibility phrases are extracted, such as "formulate annual budget," "review supplier contracts," and "monitor project progress," and these are standardized into a set of field sequences. Then, a matching operation between responsibility fields is performed. First, the number of overlapping responsibility field text expressions in the original and new positions is calculated. This is done by calculating the longest common substring at the character level for each phrase. For example, if the original position contains the field "manage budget expenditures" and the new position contains the field "budget expenditure approval," their common substring is "budget expenditures," which is counted as one match. After repeating this operation, assuming a total of 12 matched fields and a total of 20 fields, the field matching rate is... The semantic vector representation model is further invoked to construct semantic vectors for all responsibility items. Semantic similarity is calculated using the cosine similarity between the vectors. For example, the cosine similarity between "cost analysis" and "budget control" is 0.82. An approximation threshold of 0.75 is set, and the number of field pairs exceeding this value is 8. Therefore, the semantic similarity coverage rate is [percentage missing]. The weighted responsibility intersection expression score is calculated by weighting the matching rate and the approximate coverage rate at a set weight ratio of 1:2. Based on this, an intersection expression vector is constructed. Then, the original job expression structure and the overlapping expression structure are aligned by field order. If the field order structure is completely consistent, it is determined that the information structure is consistent; otherwise, it is marked as structural difference. Finally, the job responsibility intersection ratio is calculated to be 46.67%.

[0082] The cross-job mapping submodule filters field groups with path continuity markers based on the job responsibility intersection ratio, detects the original position labels of the target field group in the job path structure, adjusts the mapping order and judges the degree of overlap of field positions after remapping, extracts overlapping structure segments and counts the range of connected paths, and obtains the cross-node field connection density.

[0083] Based on the above job responsibility intersection ratio results, records with a ratio higher than 40% are selected as job pairs with a mappable basis. All responsibility fields are extracted from these job pairs, and field groups with path continuity are marked, meaning these fields appear consecutively two or more times in the task flow path. For example, if the "Project Review" field appears consecutively three times in multiple task stage paths, this field group is marked as a continuous field group. The target job path structure is located for these field groups, identifying the position label of their first appearance in the incoming job structure. For example, if "Project Review" first appears at node 6 in the incoming path node, it is recorded as the original position index 6. Next, a field remapping and sorting operation is performed, aligning the original path field order with the incoming path structure. The field index order is adjusted according to the shortest jump path principle, and the continuity of the adjusted field indexes is determined. If the new sorting of the three fields is [6,7,8], it is recorded as a complete overlap; if the sorting is [6,9,11], it is recorded as a partial jump overlap. Extract overlapping structural segments and count the number of paths connected to all field nodes in these segments. For example, if the "Approval Process Segment" has 4 paths connecting to it, then the number of connection paths is 4. If the segment has 3 fields, then the connection density is... The final cross-node field connection density is 1.33. This value will be used in subsequent steps to calculate the degree of change in the responsibility path.

[0084] The path change identification submodule calculates the number of differences, the degree of change in connection order, and the differences in expression span between upstream and downstream nodes in the structural path based on the cross-node field connection density, using the following formula:

[0085] ;

[0086] Calculate the offset density of the order of responsibility expression in the job path, construct structural change description labels between the reconstructed path structure and the original path structure, and generate path evolution feature information;

[0087] in, This indicates the density of the order offset of responsibility expressions in the job path, used to measure the density of the differences in responsibility field expressions between the original node and the transferred node relative to the span of the path structure. For the first The normalized order value of the responsibility field in the upstream node of the original position is obtained by obtaining the sort number of the field in the path node list and dividing that number by the total number of path nodes for the position. For the first The normalized order value of the "Responsibility" field in the downstream nodes of the transferred position is obtained by acquiring the new sort number of the field in the reorganized path structure and dividing that number by the number of nodes in the reorganized path. For the first The normalized span value of the mapped field in the path is obtained by calculating the index difference between the start and end positions of the node containing the field, and dividing it by the total length of the nodes in the entire path. The maximum normalized ordinal value in the mapping field is obtained by dividing the maximum sort number of the selected field across all structural node paths by the total number of nodes. The minimum normalized ordinal value in the mapping field is obtained by dividing the minimum sort number of the selected field across all structural node paths by the total number of nodes. This represents the total number of responsibility fields identified in the path node structure, indicating the number of types of responsibility fields. This is the index of the field currently being processed within the set of responsibility fields, used to locate the upstream and downstream order of the corresponding field in the structure. The total number of entries in the path structure to be processed represents the number of path segments whose span is being calculated. The path being processed is indexed and used to extract the normalized value of the span of nodes in the path;

[0088] Based on the cross-node field connection density, calculate the differences between the responsibility fields bound to upstream nodes in the original job path and the responsibility fields bound to downstream nodes in the new job path. First, calculate the normalized sort number of all responsibility fields in the original and new paths. For example, if a responsibility field "Report Review" ranks 5th in the original path and has a total path length of 10, then its... If its position in the new path is 2nd and the number of nodes in the recombined path is 8, then For all After performing this operation on the responsibility field, the squared difference in expression order is calculated and summed. For example, if there are 3 fields with upstream / downstream normalized values ​​of [0.5, 0.6, 0.75] and [0.25, 0.4, 0.5] respectively, then the sum of squared offsets is: The normalized sorting and difference data of the fields are shown in the table below:

[0089] Table 3. Example Table of Responsibility Path Structure Restructuring Parameters

[0090] ;

[0091] As shown in Table 3, the order of expression of the responsibility fields changed little between the original path and the new path, and the squared normalized difference of each field remained between 0.04 and 0.0625, indicating that the job restructuring was a local structural adjustment rather than a complete rearrangement.

[0092] Next, we calculate the denominator of the path span. Assume there are 3 participating path segments, with mapping field spans of 2, 3, and 2 in each path, and a total path node length of 8. Then, the normalized span is... The summation result is Simultaneously, selecting the maximum normalized position value as 0.75 and the minimum value as 0.25 among all fields, the span difference is expressed as follows: Substitute the values ​​into the formula to calculate the offset density of the job path responsibility expression order:

[0093] ;

[0094] The "Responsibility Expression Order Offset Density" in the job path refers to the intensity of the offset in the expression order of responsibility fields between the original and transferred positions within the structural span of the responsibility path. This parameter measures the relative relationship between the concentration of expression position offset during responsibility structure adjustment and the path's carrying capacity. A larger value indicates a concentrated and significant change in the expression order of responsibility fields, occurring within a path with limited structural carrying capacity, suggesting a risk of structural overlap and compression in the path reconstruction area. This indicator serves as an important basis for identifying job connection risks and assessing path adaptability, supporting dynamic equilibrium analysis between the structural carrying capacity of nodes and expression disturbances during job changes, and possessing practical data discrimination capabilities. This value characterizes the reconstruction density of the path's responsibility expression order during job changes. Since the offset density threshold is set at 0.2, values ​​below this indicate that the responsibility expression structure remains continuous, thus the degree of job path reconstruction is low, and the change is labeled as "mild path reorganization," recorded in the job evolution characteristic information table.

[0095] Please see Figure 5 The fracture detection module includes:

[0096] The vacant post identification submodule acquires path evolution feature information, detects continuous node groups without configured personnel binding fields in the post connection path, calculates the length of the node group, and compares it with the total path length to obtain the vacant post continuity ratio.

[0097] Obtaining path evolution characteristic information involves reading the path structure change records and job connection link tags generated by the previous module, and determining the personnel configuration status of the node segments marked for reconstructed paths. The determination method involves checking whether each job node in each path segment is bound to a specific employee ID or personnel identification code. If two or more job nodes are consecutively unbound, the system records them as an "empty job node group." Taking a typical path as an example, containing the node sequence [N1, N2, N3, N4, N5, N6, N7], where N2, N3, and N4 do not have personnel binding fields configured, while the remaining nodes are bound to personnel fields, the system will mark N2-N4 as an empty job node group with a length of 3. The ratio of this empty job node group length to the total number of nodes in the path is calculated to obtain the empty job continuity ratio. For example, if the total path length is 7 and the empty job node group length is 3, then the empty job continuity ratio is... If the ratio exceeds the set threshold of 0.4 for continuous vacancy, the node group is identified as a "continuous vacancy segment", its position range in the path is recorded as N2-N4, and it is added to the job structure risk candidate segment set for further verification.

[0098] The responsibility field comparison submodule collects the responsibility field set of upstream and downstream connecting nodes of the vacant section based on the vacancy continuity ratio, analyzes the number of field set items, judges the difference between the number and the responsibility configuration benchmark value, and obtains the responsibility configuration offset.

[0099] After confirming a node segment where the consecutive vacancy rate exceeds a threshold, the system extracts a set of responsibility fields from the upstream and downstream nodes of that vacancy segment. Specifically, it locates the preceding and following nodes of the vacancy segment and reads their job responsibility configuration field sets. For example, node N1 contains responsibility items [budget preparation, cost verification], and node N5 contains responsibility items [reimbursement approval, supplier collaboration]. These are merged to form a set of 4 upstream and downstream responsibility fields. The system then reads the baseline value of the responsibility configuration corresponding to this job path, i.e., the number of responsibility items that this path type should cover under a complete configuration. For example, a certain type of procurement path should cover 6 core responsibility items. The system compares the number of items in the upstream and downstream node sets of the vacancy segment (4 items here) with the baseline value of 6 items, finding that 2 responsibility items are missing and the configuration offset is [missing value]. If the offset exceeds the set offset threshold of 0.25, it is determined that the upstream and downstream responsibilities of the vacant post segment are incomplete, posing a risk of responsibility gaps. The table below shows the field differences for this node segment:

[0100] Table 4 Comparison of Upstream and Downstream Responsibility Fields and Configuration Benchmarks for Vacant Positions

[0101] ;

[0102] As shown in Table 4, the set of responsibility fields for this vacant post section is significantly missing compared to the baseline value, and the responsibility configuration offset reaches 33.3%. This section is marked as "responsibility coverage abnormal" and enters the fracture structure identification process.

[0103] The fracture interval determination submodule, based on the responsibility configuration offset, combined with the node distribution and responsibility field coverage in the job connection path, filters out locations where there are fractures in the responsibility field coverage, marks the fracture segment range in the node sequence, and establishes structural risk identification results.

[0104] Based on the calculated responsibility configuration offset, the system performs further structural traversal and field coverage analysis on the job connection path. First, it checks the distribution of responsibility fields at each node in the path to identify any instances where responsibility fields cannot be continuously transmitted from upstream to downstream due to node structural breaks. Specifically, the system sequentially breaks down the job path into a sequence of nodes and marks the responsibility fields of each node with field coverage. If a field is absent from any node, it is determined that the responsibility field is in a broken state within that path. Combining the location labels of vacant job segments with the mapping results of responsibility fields, the system marks the broken sections of the node sequence. For example, in path nodes N1-N7, if the responsibility field "Project Approval" is completely missing between N2-N4 and no inherited field appears in upstream or downstream nodes, this responsibility field is marked as a broken item. Based on this, the system marks nodes N2-N4 as the broken section range and establishes structural risk identification results, including the path ID, the start and end node numbers of the broken section, a list of missing responsibility items, and the severity level of the broken section. If a path contains more than two broken fields and the length of the broken segment exceeds 30% of the total number of nodes, the risk level is determined to be "medium-high risk" and recorded in the structure early warning module for reference in operation and maintenance decisions.

[0105] Please see Figure 6 The access control module includes:

[0106] The cumulative statistics submodule obtains the structural risk identification results, analyzes the salary field, performance evaluation field and organizational function field in the access path, counts the cumulative number of occurrences of each type of field in the path, and obtains the total number of path-sensitive fields;

[0107] After obtaining the structural risk identification results, the system locates all job path node groups with broken sections or abnormal responsibility coverage, and identifies these paths as key areas for access control, initiating the sensitive field identification process. Fields involved in the access paths are categorized into three types based on content type: salary fields, performance evaluation fields, and organizational function fields. The system sequentially traverses all nodes in these paths that are bound to file records or task fields, extracting the name and field code of each type of field. For example, path P101 involves fields such as "basic salary," "quarterly performance evaluation level," "department affiliation," "salary adjustment coefficient," "job level," and "year-end bonus budget." Through field type mapping rules, the system identifies 3 salary fields, 2 performance evaluation fields, and 1 organizational function field. After counting field types at the path level, the system obtains the cumulative occurrence count of each type of field in the path and summarizes the total number of sensitive fields for the path. For example, if the salary field has a cumulative occurrence count of 4, the performance evaluation field 3, and the organizational function field 2 in a certain path, then the total number of sensitive fields for the path is [missing information]. The distribution structure of each type of field is recorded as follows:

[0108] Table 5 Distribution of Path-Sensitive Field Types

[0109] ;

[0110] As shown in Table 5, the sensitive fields in path P101 are relatively concentrated, especially the salary field, which appears frequently. It is necessary to compare them with the user's access permissions to determine whether the user has the corresponding field visibility permission.

[0111] The role permission comparison submodule, based on the total number of path-sensitive fields, calls the access level field group corresponding to the user role, compares the coverage ratio of the number of sensitive fields with the number of accessible fields in the role field set, and obtains the permission coverage ratio.

[0112] After obtaining the total number of path-sensitive fields, the system retrieves the current user's role type and extracts the corresponding field access level group from the access control table. Taking "Level 1 Management" as an example, its role access field group includes 2 accessible salary fields, 1 performance evaluation field, and 2 organizational function fields. The system matches by field type and counts the number of overlaps between the current role's accessible field set and path-sensitive fields. Taking path P101 as an example, it has 9 sensitive fields, of which 4 overlap with the role's field set: "Basic Salary," "Year-End Bonus Budget," "Department Affiliation," and "Job Level." The system calculates the permission coverage ratio as follows: If the coverage ratio is lower than the set threshold of 0.6, the system will mark the current role's field access capability for the path as "insufficient permission coverage" and record the ratio and field details in the permission matching report as the basis for determining the path permission status.

[0113] The path permission determination submodule determines the matching relationship between sensitive fields in the path node and the set of role fields based on the permission coverage ratio, marks the access permission status of the path, and establishes the file access management results.

[0114] Based on the permission coverage ratio, the system enters the access permission status determination process. The determination logic is as follows: if the permission coverage ratio is greater than or equal to 0.8, it is considered "fully permitted"; if it is between 0.6 and 0.8, it is marked as "restricted access"; if it is less than 0.6, it is marked as "access denied". Taking path P101 as an example, the user role corresponds to 4 accessible fields and a total of 9 sensitive fields. The calculated coverage ratio is 0.444, which is less than 0.6, meeting the "access denied" rule. Based on this, the system generates a path access permission status of "denied" and generates an access audit record. The record includes the user ID, access path number, list of sensitive fields, list of accessible fields, and the final judgment status. The system writes this permission status to the file access control master table and updates the path control policy to ensure that subsequent access by the user to the denied fields in this path will trigger an access denial or de-identification prompt mechanism.

[0115] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A knowledge graph-based human resource record storage and management system, characterized in that, The system includes: The job identification module calls employee file information, analyzes the combination of job description phrases and skill keywords, compares the differences in field order and word usage between the job phrases and the standard template, performs expression mapping between job terminology and the standard template, and generates job terminology normalization results. Based on the job terminology normalization results, the path modeling module calculates the jump frequency and node span differences of path nodes in employee task records at multiple stages, analyzes the alignment of path connections with job responsibilities and processes, filters node reconstruction segments, establishes stage labels, and generates job cycle division results. The reconstructed segment is a segment in the actual path node sequence that has a reversed, merged, disassembled or replaced order with the standard process. By comparing the order of each node with the timestamp interval, it is determined whether a reconstruction has occurred. The establishment phase label specifically involves calculating the node behavior change index, grouping behavior types, and dividing them into periodic segments; The job term division results include stage behavior identifiers, task span ranges, and path segment numbers; The node linkage module uses the job cycle division results to determine the intersection of responsibility fields between the original job node and the transferred job node, selects the field group with cross-path connection continuity markers as the target field group and detects its original position label in the transferred job path structure to adjust the mapping order, calculates the responsibility overlap between the original job and the transferred job, constructs structural change description labels, and generates path evolution feature information. The path evolution feature information includes responsibility migration labels, node connection order, and evolution path structure; The fracture identification module combines the path evolution feature information to construct the job connection path, analyze the length of the continuous vacant job node group and the path ratio, determine the difference in responsibility fields of upstream and downstream nodes, delineate the scope of responsibility interruption, and generate structural risk identification results.

2. The knowledge graph-based human resource file storage and management system according to claim 1, characterized in that, The job terminology normalization result includes standard job fields, skill attribution tags, and expression mapping structure. The structural risk identification result includes vacant job connection segments, responsibility breakpoint locations, and link interruption ranges.

3. The knowledge graph-based human resource file storage and management system according to claim 1, characterized in that, The job identification module includes: The job description phrase extraction submodule obtains employee file information, extracts job description phrases and skill keywords, analyzes the combination of phrases and keywords, establishes a job skill dataset, and generates job phrase data. The structure format comparison submodule compares the differences in field order and word usage between the job term data and the standard term template, based on the amount of job term data. It determines the order in which keywords appear in the template, filters fields that match the template structure, and generates the job structure difference quantity. The terminology mapping adjustment submodule calls the job structure difference quantity, adjusts the original job phrase descriptions according to the combination matching results between jobs and skills, maps them to standard template expressions, and obtains job terminology normalization results.

4. The knowledge graph-based human resource file storage and management system according to claim 3, characterized in that, The path modeling module includes: The node frequency analysis submodule obtains the normalization result of the job terminology, counts the frequency of occurrence of path nodes in each stage of the employee task record and the number of stage nodes, analyzes the sequential distribution of nodes in each stage, establishes the node occurrence change range, and generates the stage jump change amount. The task path evaluation submodule calls the stage jump change amount, compares the connection order between task path nodes and key nodes of the job standard task process, filters the continuity of node distribution and missing segments, judges the co-occurrence ratio of actual task nodes and standard path nodes, and obtains the critical path fit coefficient. The behavior segmentation construction submodule monitors the distribution of node reconstruction behavior in the time series based on the critical path fit coefficient, collects the cumulative frequency, number of segment nodes, node time interval and span of each behavior type in the periodic time period, calculates the node behavior change index, groups the behavior types and divides them into periodic segments, forms a structured job cycle segmentation identifier, and establishes the job cycle division result.

5. The knowledge graph-based human resource file storage and management system according to claim 4, characterized in that, The node linkage module includes: The responsibility intersection extraction submodule, based on the job cycle division results, calls the responsibility field set in the original job node and the transferred job node, calculates the overlap ratio between the number of matching content items of the responsibility field in the text expression and the semantic range expression vector of the field, constructs the intersection expression vector based on the coverage rate of the responsibility items and the content similarity score, obtains the overlapping expression set and judges the consistency of information structure, and obtains the job responsibility intersection ratio. The cross-job mapping submodule, based on the job responsibility intersection ratio, selects field groups with cross-path connection continuity markers as target field groups, detects the original position labels of the target field groups in the imported job path structure, adjusts the mapping order and judges the degree of overlap of field positions after remapping, extracts overlapping structure segments and counts the range of connected paths, and obtains the cross-node field connection density. The path change identification submodule calculates the number of differences, the degree of change in connection order, and the difference in expression span between the binding responsibility fields of upstream and downstream nodes in the structural path based on the cross-node field connection density. It also calculates the responsibility expression order offset density in the job path, constructs structural change description tags between the reconstructed path structure and the original path structure, and generates path evolution feature information.

6. The knowledge graph-based human resource file storage and management system according to claim 5, characterized in that, The fracture detection module includes: The vacant post section identification submodule obtains the path evolution feature information, detects the continuous node groups without configured personnel binding fields in the post connection path, calculates the length of the node group, and compares it with the total path length to obtain the vacant post continuity ratio. The responsibility field comparison submodule collects the responsibility field set of upstream and downstream connecting nodes of the vacant post section based on the vacant post continuity ratio, analyzes the number of field set items, judges the difference between the number and the responsibility configuration benchmark value, and obtains the responsibility configuration offset. The fracture interval determination submodule, based on the responsibility configuration offset, combined with the node distribution and responsibility field coverage in the job connection path, filters out locations where there are fractures in the responsibility field coverage, marks the fracture segment range in the node sequence, and establishes structural risk identification results.

7. The knowledge graph-based human resource file storage and management system according to claim 1, characterized in that, The system also includes: Based on the structural risk identification results, the access control module analyzes the distribution of sensitive fields in the access path, calls the role field group to compare the field matching coverage ratio, determines the path access permission status, and generates file access management results. The file access management results include permission path labels, sensitive field density, and access judgment level.

8. The knowledge graph-based human resource file storage and management system according to claim 7, characterized in that, The access control module includes: The cumulative statistics submodule obtains the structural risk identification results, analyzes the salary field, performance evaluation field and organizational function field in the access path, counts the cumulative number of occurrences of each type of field in the path, and obtains the total number of path-sensitive fields. The role permission comparison submodule, based on the total number of path sensitive fields, calls the access level field group corresponding to the user role, compares the coverage ratio of the total number of sensitive fields with the number of accessible fields in the role field set, and obtains the permission coverage ratio. The path permission determination submodule determines the matching relationship between sensitive fields and role field sets in the path node based on the permission coverage ratio, marks the access permission status of the path, and establishes the file access management result.

Images