A multi-branch and multi-node comparison calculation method based on knowledge graph

Abstract

The present invention discloses a multi-branch multi-node comparison calculation method based on a knowledge graph, which relates to the technical field of knowledge graphs and includes the following steps: collecting job data and job seeker data, and constructing a knowledge graph of positions and job seekers after preprocessing the data; analyzing job requirements based on the job data, and determining key skill nodes. The multi-branch multi-node comparison calculation method based on the knowledge graph can systematically integrate and analyze multi-dimensional information between job seekers and positions, including skills, experience, educational background, etc., by constructing a knowledge graph. The multi-branch multi-node comparison calculation method ensures that every detail is fully considered, thereby greatly improving the accuracy of matching. At the same time, the automated comparison process significantly shortens the time required for matching and improves processing efficiency, so that the recruiter can quickly locate the most suitable candidate, and the job seeker can also find a position that matches their own conditions more quickly.

Description

Technical Field

[0001] The present invention relates to the field of knowledge graph technology, and specifically to a multi-branch multi-node comparison and calculation method based on knowledge graph. Background Art

[0002] In job recruitment, different companies may use different terms and descriptions to describe the same position. The requirements for job seekers for different positions vary greatly, involving not only hard conditions such as professional skills and work experience, but also soft abilities such as personality traits, teamwork ability, and innovative thinking. Knowledge graph technology can display complex job requirements through data mining, information processing, knowledge measurement, and graphics, revealing the patterns of job requirements and providing practical and valuable reference for job recruitment.

[0003] Traditional recruitment methods make it difficult to comprehensively and accurately evaluate the match between job seekers and the job market, while knowledge graph technology can finely match job requirements with the various abilities of job seekers. Therefore, how to set multiple key skill nodes and evaluation indicators, and comprehensively evaluate the match degree of job seekers at each skill node through multi-branch and multi-node comparison calculation in the knowledge graph, is the problem we need to solve. To this end, a multi-branch and multi-node comparison calculation method based on knowledge graph is proposed. Summary of the invention

[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions: a multi-branch multi-node comparison calculation method based on knowledge graph, comprising the following steps:

[0005] Step 1: Collect job data and job seeker data, pre-process the data, and build a knowledge graph of job positions and job seekers;

[0006] Step 2: Analyze job requirements based on job data, determine key skill nodes, set corresponding evaluation indicators for each skill node, and establish a preset comparison node system, allowing users to establish a comparison node system for requirements in the knowledge graph, where the comparison nodes include single-select nodes, multiple-select nodes, and text nodes;

[0007] Step 3: Label the job data and job seeker data to train the node comparison model, extract key information from the text, and match it with the nodes in the knowledge graph;

[0008] Step 4: Use the knowledge graph and node comparison model to compare the job seeker data with the job position data, and calculate the matching degree of the job seeker at each skill node based on the comparison results;

[0009] Step 5. Based on the comparison results, provide personalized career development suggestions for job seekers, intuitively display the comparison results to users, sort job seekers in descending order of similarity, and recommend job seekers with high matching degrees to recruiters.

[0010] Preferably, in step 1, the process of constructing the job position and job seeker knowledge graph includes:

[0011] Collect job data and job seeker data, where job data includes job description, required skills, years of experience, education background, and job responsibilities; job seeker data includes job seekers’ education background, work experience, skills, certificates, and personal interests. Use web crawler technology to collect job data from recruitment websites, company websites, job posting platforms, and other channels, and collect job seeker data from resumes submitted by job seekers, online career profiles, and social media channels to ensure the comprehensiveness and timeliness of the data. In addition, the crawling of website data must comply with relevant laws and regulations and the website’s robots protocol to ensure that the collection process complies with the principle of privacy protection and avoid illegal data collection;

[0012] Pre-process the collected job data and job seeker data by using data cleaning to remove duplicate data, correct erroneous information, fill in missing values, compare key information in job data and job seeker data, remove completely duplicate and highly similar data, and standardize terms in data from different sources;

[0013] Perform entity recognition on job data and job seeker data to identify job entities, skill entities, and experience entities. Job entities include job title, company name, and work location; skill entities include skills required for the job and skills in the job seeker's resume; and experience entities include job seekers' work experience and years of experience required for the job. Analyze the relationships between entities, including the relationship between job and company, the relationship between job and skills, the matching degree between job seekers and job, and the mastery of skills by job seekers, such as "job seekers - have - skills", "job - require - skills", "job seekers - graduated from - educational institution", etc.

[0014] Define the graph architecture, design the nodes and edges of the knowledge graph, determine the entity types and relationship types, fill the preprocessed data into the knowledge graph to form nodes and edges, and use a graph database (such as Neo4j) to store the constructed knowledge graph;

[0015] Use natural language processing technology to link entities from different sources to the same entity, enhance the connectivity of the graph, extract new relationships from unstructured text through natural language processing technology, enrich the graph content, and fuse data from different sources to ensure the consistency and accuracy of the knowledge graph.

[0016] Preferably, in step 2, the process of setting a corresponding evaluation index for each skill node includes:

[0017] Classify and organize the collected job data, identify the commonalities and differences between different jobs, and analyze the skills, experience, and educational background requirements of each job;

[0018] Based on the results of the job demand analysis, combined with the importance and universality of skills, frequently occurring skill requirements are identified as key skill nodes. Key skill nodes are indispensable skills in job performance and have a certain degree of universality in different jobs. Evaluation indicators are set for each key skill node. The evaluation indicators include proficiency level, years of relevant project experience, certifications, training experience, etc. The proficiency levels are elementary, intermediate and advanced.

[0019] Establish a node preset system, allowing users to establish the required node system in the knowledge graph, compare it with other similar node systems, and calculate the matching results through node similarity. The comparison method is divided into single-graph comparison and multi-graph comparison. If the user establishes a single-select node, the node established in the next branch has the following characteristics: when the user clicks on one of the nodes in the same branch, the node will be selected and the other nodes will be unselected. If the user clicks on the same option node multiple times, the selected state will be switched. After the comparison starts, the system will record the path of the user's selected option;

[0020] If the user creates a check node, the next branch node has the following characteristics: when the user clicks one of the nodes in the same branch, the node will be selected, and the status of other nodes will not change. If the user clicks the same option node multiple times, the selected status will be switched. After the comparison starts, the system will record the path of the user's selected option;

[0021] If the user creates a node, the next branch can create multiple nodes;

[0022] If the user creates a text node, the next branch creates a node with the following characteristics: the user can only edit this node, and the system will record the path of the user's selected option after the comparison starts;

[0023] If the user creates a link node, the current node can edit a link address, and click the address to jump directly;

[0024] Conduct quantitative analysis based on the set evaluation indicators, calculate the skill evaluation index, and determine the skill level of each key skill node.

[0025] Preferably, the expression of the skill evaluation index is:

[0026] ;

[0027] in, is the skill assessment index, is the number of evaluation indicators, It is The weight of an evaluation indicator indicates the relative importance of the indicator in the overall evaluation. It is The actual score of the evaluation index is is the highest score obtained among all evaluation indicators. is an index used to adjust the nonlinear impact of the score and increase the weight of high-scoring indicators. Higher values ​​indicate higher skill levels.

[0028] Preferably, the single spectrum comparison process includes:

[0029] Compare the node paths and node keywords to see if they are consistent. Use different comparison methods according to different node types (single-select node, multiple-select node, text node). For single-select nodes: determine whether the node paths are consistent. For multiple-select nodes: determine whether all nodes under the same path are selected consistently. If they are inconsistent, calculate the weight based on the consistent nodes. For text nodes: use keyword matching to compare, calculate the comparison results, and display them.

[0030] The multi-map comparison process includes:

[0031] Start comparison: record the node system established by the user and generate a selection path. Extract path keywords from the selection path to generate a keyword path. The keyword path is a reverse path, that is, first record the last node of the path, and then progress to the root node layer by layer.

[0032] During the comparison: the keyword path generated by the node system established by the user is matched with all other keyword paths for similarity, that is, the last similarity is matched first, then layer by layer, and finally a similarity percentage value is generated. The similarity calculation method compares the similarity of two keywords through semantic analysis, that is, the reciprocal of the sum of the shortest paths from a node to other nodes. The calculation formula is: ;

[0033] After comparison: the matching node system will be returned to the user and displayed in descending order of similarity.

[0034] Preferably, in step 3, the process of extracting key information from the text and matching it with nodes in the knowledge graph includes:

[0035] Use word segmentation tools (such as jieba word segmentation, HanLP, etc.) to segment the text in the job data and job seeker data. The word segmentation results serve as the basis for subsequent extraction of key information. Extract key information such as skills, experience, and educational background from job descriptions and job seeker resumes as the main input of the node comparison model. The extracted key information is structured and stored in a structured data format (such as JSON, XML, etc.). The structured data facilitates subsequent matching with nodes in the knowledge graph.

[0036] Each node in the knowledge graph is labeled so that each node has a unique identifier and corresponding attribute description. These nodes will serve as the reference standard for the comparison model. The job data and job seeker data are matched and labeled with the nodes in the knowledge graph. The labeling results include the corresponding relationship between each data item and the node in the knowledge graph. The labeled data are integrated to obtain the node comparison data set, and the node comparison data set is divided into a training set and a test set.

[0037] Build a node comparison model based on the neural network model, use the relevant data in the training set to train the node comparison model, and use the test set to test and evaluate the trained node comparison model;

[0038] Apply the trained node comparison model to actual job matching and job applicant screening scenarios to ensure that the model can process real-time data and interact with the knowledge graph, and assist recruitment decisions and career planning through model prediction results.

[0039] Preferably, in step 4, the process of calculating the matching degree of the job seeker at each skill node includes:

[0040] Define node matching rules, including single-select node, multiple-select node and text node matching rules. When matching single-select nodes, only one most relevant node can be selected. When matching multiple-select nodes, multiple relevant nodes can be selected, but weight distribution needs to be considered. When matching text nodes, the similarity of keywords and phrases is used.

[0041] The node comparison model is applied to each node of the job data and the job seeker data, and the node comparison model is used to output the matching degree of each node. For each single-select node, the matching scores are added and multiplied by their weight to obtain the single-select node matching index. For multiple-select nodes, the matching scores of the multiple-select nodes are processed using a logical function, and the growth rate of the scores is adjusted according to the matching degree. The matching scores are added and multiplied by their weights to obtain the multiple-select node matching index. For text nodes, the square root of the matching score of each text node is accumulated and multiplied by its weight to obtain the text node matching index.

[0042] Combine the matching scores of all nodes to generate a matching coefficient, analyze the overall matching degree between job seekers and positions, and display the matching results;

[0043] According to the comparison matching coefficient, the overall matching degree between the job seeker and the position is divided into a high matching level, a medium matching level and a low matching level, and a corresponding matching threshold is set for each matching level;

[0044] Based on the comparison of matching coefficients and matching levels, the matching degree at each skill node is analyzed to determine the strengths and weaknesses of job seekers in specific skills.

[0045] Preferably, the expression of the comparison matching coefficient is:

[0046] ;

[0047] ;

[0048] ;

[0049] ;

[0050] in, is the comparison matching coefficient, is the single-select node matching index, is the check node matching index, is the text node matching index, is the weight of the single-choice node, is the weight of the check node, is the weight of the text node, is the number of single-select nodes, It is The matching score of a single-select node, is the number of checked nodes, It is The matching score of the selected nodes, It is The baseline response sensitivity of the selected nodes, It is The threshold parameter of the logistic regression of the selected nodes, is the number of text nodes, It is The matching score of the text nodes, , The value range is between 0 and 1. The higher the value, the better the match between the job seeker and the job.

[0051] Preferably, the plurality of matching levels correspond to the plurality of matching thresholds, the matching thresholds include an upper threshold and a lower threshold, and the corresponding relationship between the plurality of matching levels and the plurality of matching thresholds is:

[0052] High matching level ;

[0053] Medium Match Level ;

[0054] Low matchmaking level ;

[0055] in, is the matching coefficient, is the lower threshold corresponding to the high matching level and the upper threshold corresponding to the medium matching level, is the lower threshold corresponding to the medium matching level and the upper threshold corresponding to the low matching level, , .

[0056] Preferably, in step 5, the process of visually displaying the comparison result to the user includes:

[0057] We deeply analyze the matching degree of each job seeker at each skill node, confirm their strengths, provide specific improvement suggestions for their weaknesses, provide career development paths and job recommendations based on their skills and interests, provide short-term suggestions on how to optimize resumes and prepare for interviews based on their current job search goals, and provide career path planning suggestions based on their career interests and long-term goals, including skills that need to be improved and possible development directions;

[0058] According to the matching coefficient, the job seekers are sorted from high to low according to the matching degree with the position, and according to the needs of the employer and the matching degree of the job seekers, the job seekers with high matching degree are automatically screened out, and suitable candidates are recommended. The job seekers with the highest matching degree are placed at the top of the recommendation list for the employer to quickly view;

[0059] Create visual reports that show how well job seekers match positions, and use charts to represent the match level, showing specific match scores and recommendations for each skill node.

[0060] The present invention provides a multi-branch multi-node comparison calculation method based on knowledge graph. It has the following beneficial effects:

[0061] 1. This multi-branch and multi-node comparison and calculation method based on the knowledge graph can systematically integrate and analyze the multi-dimensional information between job seekers and positions, including skills, experience, educational background, etc., by constructing a knowledge graph. The multi-branch and multi-node comparison and calculation method ensures that every detail is fully considered, thereby greatly improving the matching accuracy. At the same time, the automated comparison process significantly shortens the time required for matching and improves processing efficiency, allowing recruiters to quickly locate the most suitable candidates and job seekers to find positions that match their own conditions more quickly.

[0062] 2. This multi-branch and multi-node comparison calculation method based on the knowledge graph can deeply analyze the strengths and weaknesses of job seekers according to the comparison results, and provide them with personalized career development suggestions. By identifying the strengths and weaknesses of job seekers in specific skills or fields, it can accurately recommend improvement paths and career planning directions, help job seekers clarify their own positioning, and develop practical career development plans. Through automated screening and precise recommendations, recruiters can quickly lock in high-quality candidates, reduce the time and cost of invalid interviews and screening, and at the same time, improve the matching degree, which will correspondingly increase the recruitment success rate and further reduce the overall cost of recruitment. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A method flow chart of a multi-branch multi-node comparison calculation method based on a knowledge graph of the present invention;

[0064] Figure 2 A flow chart for setting corresponding evaluation indicators for each skill node in the present invention;

[0065] Figure 3 The present invention is a flow chart for calculating the matching degree of job seekers at various skill nodes. DETAILED DESCRIPTION

[0066] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are provided for the purpose of illustration and description, and are not intended to be exhaustive or to limit the present invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention, and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

[0067] The first embodiment, as Figure 1 , Figure 2 As shown, the present invention provides a technical solution: a multi-branch multi-node comparison calculation method based on a knowledge graph, comprising the following steps:

[0068] Step 1: Collect job data and job seeker data, and after preprocessing the data, build a job and job seeker knowledge graph, collect job data and job seeker data, where job data includes job description, required skills, years of experience, educational background, and job responsibilities information, and job seeker data includes job seekers' educational background, work experience, skills, certificates, and personal interest information. Use web crawler technology to collect job data from recruitment websites, company websites, job posting platforms and other channels, and collect job seeker data from resumes submitted by job seekers, online career profiles, social media and other channels to ensure the comprehensiveness and timeliness of the data. In addition, for the crawling of website data, it is necessary to comply with relevant laws and regulations and the website's robots protocol to ensure that the collection process complies with the principle of privacy protection and avoid illegal data collection. Preprocess the collected job data and job seeker data, use data cleaning to remove duplicate data, correct erroneous information, fill in missing values, compare key information in job data and job seeker data, remove completely duplicate and highly similar data, and standardize the terms in data from different sources to conduct job data and job seeker data collection. Entity recognition of candidate data, identifying position entities, skill entities and experience entities, where position entities include position name, company name, and work location; skill entities include skills required for the position and skills in the job seeker's resume; and experience entities include the job seeker's work experience and the number of years of experience required for the position. The relationship between entities is analyzed, including the relationship between position and company, the relationship between position and skills, the match between job seekers and positions, and the mastery of skills by job seekers, such as "job seekers-possess-skills", "position-required-skills", "job seekers-graduated-from-educational institutions", etc. The graph architecture is defined, and the nodes and edges of the knowledge graph are designed. The entity type and relationship type are determined. The pre-processed data is filled into the knowledge graph to form nodes and edges. A graph database (such as Neo4j) is used to store the constructed knowledge graph. Natural language processing technology is used to link entities from different sources to the same entity to enhance the connectivity of the graph. New relationships are extracted from unstructured text through natural language processing technology to enrich the graph content. Data from different sources are integrated to ensure the consistency and accuracy of the knowledge graph.

[0069] Step 2: Analyze job requirements based on job data, determine key skill nodes, set corresponding evaluation indicators for each skill node, and establish a comparison node preset system, allowing users to establish a comparison node system for requirements in the knowledge graph, where comparison nodes include single-choice nodes, multiple-choice nodes, and text nodes. Classify and organize the collected job data, identify the commonalities and differences of different jobs, and analyze the requirements of each job for skills, experience, and educational background. According to the results of the job requirement analysis, combined with the importance and universality of skills, identify frequently occurring skill requirements as key skill nodes. Key skill nodes are indispensable skills in job performance and have certain universality in different jobs. Set evaluation indicators for each key skill node. The evaluation indicators include proficiency level, years of relevant project experience, certification certificates, training experience, etc. The proficiency levels are elementary, intermediate, and advanced. Establish a node preset system, allowing users to establish a node system for requirements in the knowledge graph, compare with other similar node systems, and calculate the matching results through node similarity. The comparison method is divided into single graph comparison and multi-graph comparison Comparison, if the user establishes a single-select node, the node established in the next branch has the following characteristics: if the user clicks on one of the nodes in the same branch, the node will be selected and the other nodes will be unselected. If the user clicks on the same option node multiple times, the selected state will be switched. After the comparison starts, the system will record the path of the user's selected option. If the user establishes a check node, the node established in the next branch has the following characteristics: if the user clicks on one of the nodes in the same branch, the node will be selected and the status of other nodes will not change. If the user clicks on the same option node multiple times, the selected state will be switched. After the comparison starts, the system will record the path of the user's selected option. If the user establishes a node, the next branch can establish multiple nodes. If the user establishes a text node, the node established in the next branch has the following characteristics: the user can only edit the node. After the comparison starts, the system will record the path of the user's selected option. If the user establishes a link node, the current node can edit a link address, and clicking the address can jump directly. Combined with the set evaluation indicators, quantitative analysis is performed, the skill evaluation index is calculated, and the skill level of each key skill node is determined;

[0070] Furthermore, the expression of the skill evaluation index is:

[0071] ;

[0072] in, is the skill assessment index, is the number of evaluation indicators, It is The weight of an evaluation indicator indicates the relative importance of the indicator in the overall evaluation. It is The actual score of the evaluation index is is the highest score obtained among all evaluation indicators. is an index used to adjust the nonlinear impact of the score and increase the weight of high-scoring indicators. The higher the value, the higher the skill level. A value close to 1 indicates excellent skills, while a value close to 0 indicates insufficient skills. Can be used to quickly assess whether job seekers meet the skill requirements of the position. Job seekers can use To identify your own strengths and weaknesses in skills, so as to carry out targeted career development and learning;

[0073] Furthermore, the process of single map comparison includes:

[0074] Compare the node paths and node keywords to see if they are consistent. Use different comparison methods according to different node types (single-select node, multiple-select node, text node). For single-select nodes: determine whether the node paths are consistent. For multiple-select nodes: determine whether all nodes under the same path are selected consistently. If they are inconsistent, calculate the weight based on the consistent nodes. For text nodes: use keyword matching to compare, calculate the comparison results, and display them.

[0075] The process of multi-map comparison includes:

[0076] Start comparison: record the node system established by the user and generate a selection path. Extract path keywords from the selection path to generate a keyword path. The keyword path is a reverse path, that is, first record the last node of the path, and then progress to the root node layer by layer.

[0077] During the comparison: the keyword path generated by the node system established by the user is matched with all other keyword paths for similarity, that is, the last similarity is matched first, then layer by layer, and finally a similarity percentage value is generated. The similarity calculation method compares the similarity of two keywords through semantic analysis, that is, the reciprocal of the sum of the shortest paths from a node to other nodes. The calculation formula is: ;

[0078] After comparison: the matching node systems are returned to the user and displayed in descending order of similarity;

[0079] Step 3: Mark the job data and job seeker data to train the node comparison model, extract key information from the text, and match it with the nodes in the knowledge graph. Use word segmentation tools (such as Jieba word segmentation, HanLP, etc.) to segment the text in the job data and job seeker data. The word segmentation results serve as the basis for subsequent extraction of key information. Extract key information such as skills, experience, and educational background from job descriptions and job seeker resumes as the main input to the node comparison model. Store the extracted key information in a structured format (such as JSON, XML, etc.). Structured data facilitates subsequent matching with nodes in the knowledge graph. Label each node in the knowledge graph so that each node has a unique identifier and Corresponding attribute descriptions, these nodes will serve as reference standards for the comparison model, match and annotate the job data and job seeker data with the nodes in the knowledge graph, and the annotation results contain the corresponding relationship between each data item and the node in the knowledge graph. The annotated data is integrated to obtain the node comparison data set, and the node comparison data set is divided into a training set and a test set. A node comparison model is built based on the neural network model, and the node comparison model is trained using the relevant data in the training set, and the trained node comparison model is tested and evaluated using the test set. The trained node comparison model is applied to actual job matching and job seeker screening scenarios to ensure that the model can process real-time data and interact with the knowledge graph, and assist recruitment decisions and career planning through model prediction results;

[0080] Step 4: Use the knowledge graph and node comparison model to compare the job seeker data with the job position data, and calculate the matching degree of the job seeker at each skill node based on the comparison results;

[0081] Step 5. Based on the comparison results, provide personalized career development suggestions for job seekers, intuitively display the comparison results to users, sort job seekers in descending order of similarity, and recommend job seekers with high matching degrees to recruiters.

[0082] The second embodiment is based on the first embodiment. Figure 3 As shown, in step 4, the process of calculating the matching degree of job seekers at each skill node includes:

[0083] Define node matching rules, including matching rules for single-select nodes, multiple-select nodes, and text nodes. When matching single-select nodes, only one most relevant node can be selected. When matching multiple-select nodes, multiple relevant nodes can be selected, but weight distribution needs to be considered. When matching text nodes, based on the similarity of keywords and phrases, the node comparison model is applied to each node of job data and job seeker data, and the node comparison model is used to output the matching degree of each node. For each single-select node, the matching scores are added and multiplied by their weight to obtain the single-select node matching index. For multiple-select nodes, the matching scores of the multiple-select nodes are processed using logical functions, and the growth rate of the scores is adjusted according to the matching degree. The matching scores are added up and multiplied by their weights to obtain the matching index of the selected nodes. For text nodes, the square root of the matching scores of each text node is accumulated and multiplied by its weight to obtain the text node matching index. The matching scores of all nodes are combined to generate a comparison matching coefficient, analyze the overall matching degree between job seekers and positions, and display the matching results. According to the comparison matching coefficient, the overall matching degree between job seekers and positions is divided into high matching level, medium matching level and low matching level, and the corresponding matching threshold is divided for each matching level. Based on the comparison matching coefficient and matching level, the matching degree on each skill node is analyzed to determine the strengths and weaknesses of job seekers in specific skills.

[0084] Furthermore, the expression of the comparison matching coefficient is:

[0085] ;

[0086] ;

[0087] ;

[0088] ;

[0089] in, is the comparison matching coefficient, is the single-select node matching index, is the check node matching index, is the text node matching index, is the weight of the single-choice node, is the weight of the check node, is the weight of the text node, is the number of single-select nodes, It is The matching score of the single-select node (a value between 0 and 1, where 1 indicates a perfect match), is the number of checked nodes, It is The matching score of the selected nodes, It is The baseline response sensitivity of the selected node (controls the steepness of the S-curve), It is The threshold parameter of the logistic regression of the selected nodes, is the number of text nodes, It is The matching score of the text nodes (based on the score obtained by the text similarity algorithm), , The value range is between 0 and 1. The higher the value, the better the match between the job seeker and the position;

[0090] Furthermore, multiple matching levels correspond to multiple matching thresholds, and the matching thresholds include an upper threshold and a lower threshold. The corresponding relationship between the multiple matching levels and the multiple matching thresholds is:

[0091] High matching level ; The applicant's skills, experience and job requirements are highly matched, making him an ideal candidate, with a matching coefficient of more than Indicates that the applicant is a high priority candidate as they are a good match for the position in terms of most key skills and experience;

[0092] Medium Match Level The applicant matches the job requirements in most aspects, but may lack some key skills or experience. arrive In between means the applicant generally meets the job requirements but may need further assessment or training to meet all requirements;

[0093] Low matchmaking level ; There is a large gap between the job seeker and the job requirements in key skills and experience, and the job seeker may not be suitable for the position. The matching coefficient is lower than Indicates that the applicant does not match the job requirements in multiple key areas and may need to consider other more suitable positions or improve skills;

[0094] in, is the matching coefficient, is the lower threshold corresponding to the high matching level and the upper threshold corresponding to the medium matching level, is the lower threshold corresponding to the medium matching level and the upper threshold corresponding to the low matching level, , ;

[0095] In step 5, the process of visually displaying the comparison results to the user includes:

[0096] In-depth analysis of the matching degree of each job seeker at each skill node, and confirm their strengths to the job seekers, provide specific improvement suggestions for the job seekers' weaknesses, provide career development paths and job recommendation suggestions based on the job seekers' skills and interests, provide short-term suggestions on how to optimize resumes and prepare for interviews based on current job search goals, and provide career path planning suggestions based on job seekers' career interests and long-term goals, including skills that need to be improved, possible development directions, etc. According to the comparison matching coefficient, job seekers are sorted from high to low according to the matching degree with the position, and according to the needs of the recruiter and the matching degree of the job seekers, job seekers with high matching degrees are automatically screened out, and suitable candidates are recommended. The job seekers with the highest matching degree are placed at the top of the recommendation list for the recruiter to quickly view, create visual reports, show the matching degree of job seekers and positions, and use charts to represent the matching level, showing the specific matching scores and suggestions for each skill node.

[0097] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field and related fields without creative work should fall within the scope of protection of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention are implemented according to the conventional means in the field unless otherwise specified and limited.

Claims

1. A multi-branch multi-node comparison calculation method based on knowledge graph, characterized in that: The following steps are involved: Step 1: Collect job data and job seeker data, and after preprocessing the data, construct a job and job seeker knowledge graph. In step 1, the process of constructing the job and job seeker knowledge graph includes: Collect job data and job seeker data, where job data includes job description, required skills, years of experience, education background, and job responsibilities, and job seeker data includes job seekers’ education background, work experience, skills, certificates, and personal interests; Pre-process the collected job data and job seeker data by using data cleaning to remove duplicate data, correct erroneous information, fill in missing values, compare key information in job data and job seeker data, remove completely duplicate and highly similar data, and standardize terms in data from different sources; Perform entity recognition on job data and job seeker data to identify job entities, skill entities, and experience entities. Job entities include job title, company name, and work location; skill entities include skills required for the job and skills in the job seeker's resume; and experience entities include the job seeker's work experience and the years of experience required for the job. Analyze the relationships between entities, including the relationship between the job and the company, the relationship between the job and skills, the match between the job seeker and the job, and the job seeker's mastery of skills; Define the graph architecture, design the nodes and edges of the knowledge graph, determine the entity types and relationship types, fill the preprocessed data into the knowledge graph to form nodes and edges, and use the graph database to store the constructed knowledge graph; Use natural language processing technology to link entities from different sources to the same entity, extract new relationships from unstructured text through natural language processing technology, enrich graph content, and fuse data from different sources; Step 2: Analyze job requirements based on job data, determine key skill nodes, set corresponding evaluation indicators for each skill node, and establish a preset comparison node system, allowing users to establish a comparison node system for requirements in the knowledge graph, where the comparison nodes include single-select nodes, multiple-select nodes, and text nodes; Step 3: Mark the job data and job seeker data to train the node comparison model, extract key information from the text, and match it with the nodes in the knowledge graph. In step 3, the process of extracting key information from the text and matching it with the nodes in the knowledge graph includes: Use word segmentation tools to segment text in job data and job seeker data, and extract key information such as skills, experience, and educational background from job descriptions and job seeker resumes. Use this information as the main input to the node comparison model, and store the extracted key information in a structured data format. Label each node in the knowledge graph so that each node has a unique identifier and corresponding attribute description, match and annotate the job data and job seeker data with the nodes in the knowledge graph, and the annotation results contain the corresponding relationship between each data item and the node in the knowledge graph. Integrate the annotated data to obtain a node comparison data set, and divide the node comparison data set into a training set and a test set. Build a node comparison model based on the neural network model, use the relevant data in the training set to train the node comparison model, and use the test set to test and evaluate the trained node comparison model; Apply the trained node comparison model to actual job matching and job seeker screening scenarios, and interact with the knowledge graph to assist recruitment decisions and career planning through model prediction results; Step 4: Use the knowledge graph and node comparison model to compare the job seeker data with the job position data, and calculate the matching degree of the job seeker at each skill node based on the comparison results. In step 4, the process of calculating the matching degree of the job seeker at each skill node includes: Define node matching rules, including matching rules for single-select nodes, multiple-select nodes, and text nodes; The node comparison model is applied to each node of the job data and the job seeker data, and the node comparison model is used to output the matching degree of each node. For each single-select node, the matching scores are added and multiplied by their weight to obtain the single-select node matching index. For multiple-select nodes, the matching scores of the multiple-select nodes are processed using a logical function, and the growth rate of the scores is adjusted according to the matching degree. The matching scores are added and multiplied by their weights to obtain the multiple-select node matching index. For text nodes, the square root of the matching score of each text node is accumulated and multiplied by its weight to obtain the text node matching index. Combine the matching scores of all nodes to generate a matching coefficient, analyze the overall matching degree between job seekers and positions, and display the matching results; According to the comparison matching coefficient, the overall matching degree between the job seeker and the position is divided into a high matching level, a medium matching level and a low matching level, and a corresponding matching threshold is set for each matching level; Based on the matching coefficient and matching level, analyze the matching degree at each skill node to determine the strengths and weaknesses of job seekers in specific skills; Step 5. Based on the comparison results, provide personalized career development suggestions for job seekers, intuitively display the comparison results to users, sort job seekers in descending order of similarity, and recommend job seekers with high matching degrees to recruiters.

2. According to claim 1, a multi-branch multi-node comparison calculation method based on knowledge graph is characterized in that: In step 2, the process of setting corresponding evaluation indicators for each skill node includes: Classify and organize the collected job data, identify the commonalities and differences between different jobs, and analyze the skills, experience, and educational background requirements of each job; Based on the results of the job requirements analysis, combined with the importance and prevalence of skills, frequently occurring skill requirements are identified as key skill nodes, and evaluation indicators are set for each key skill node; Establish a node preset system, allowing users to establish the required node system in the knowledge graph, compare it with other similar node systems, and calculate the matching results through node similarity. The comparison method is divided into single graph comparison and multi-graph comparison; Conduct quantitative analysis based on the set evaluation indicators, calculate the skill evaluation index, and determine the skill level of each key skill node.

3. The multi-branch multi-node comparison calculation method based on knowledge graph according to claim 2 is characterized in that: The expression of the skill evaluation index is: ； in, is the skill assessment index, is the number of evaluation indicators, It is The weight of the evaluation index, It is The actual score of the evaluation index is is the highest score obtained among all evaluation indicators. is an exponent used to adjust the nonlinear effects of the score, Higher values ​​indicate higher skill levels.

4. The multi-branch multi-node comparison calculation method based on knowledge graph according to claim 3 is characterized in that: The process of single spectrum comparison includes: Compare the node paths and node keywords to see if they are consistent. Use different comparison methods according to different node types. For single-select nodes: determine whether the node paths are consistent. For multiple-select nodes: determine whether all nodes under the same path are selected consistently. If not, calculate the weight based on the consistent nodes. For text nodes: use keyword matching to compare, calculate the comparison results, and display them. The multi-map comparison process includes: Start comparison: record the node system established by the user and generate a selection path. Extract path keywords from the selection path to generate a keyword path. The keyword path is a reverse path, that is, first record the last node of the path, and then progress to the root node layer by layer. During the comparison: the keyword path generated by the node system established by the user is matched with all other keyword paths for similarity, that is, the last similarity is matched first, then layer by layer, and finally a similarity percentage value is generated. The similarity calculation method compares the similarity of two keywords through semantic analysis, that is, the reciprocal of the sum of the shortest paths from a node to other nodes. The calculation formula is: ; After comparison: the matching node system will be returned to the user and displayed in descending order of similarity.

5. According to claim 4, a multi-branch multi-node comparison calculation method based on knowledge graph is characterized in that: The expression of the comparison matching coefficient is: ； ； ； ； in, is the comparison matching coefficient, is the single-select node matching index, is the check node matching index, is the text node matching index, is the weight of the single-choice node, is the weight of the check node, is the weight of the text node, is the number of single-select nodes, It is The matching score of a single-select node, is the number of checked nodes, It is The matching score of the selected nodes, It is The baseline response sensitivity of the selected nodes, It is The threshold parameter of the logistic regression of the selected nodes, is the number of text nodes, It is The matching score of the text nodes, , The value range is between 0 and 1. The higher the value, the better the match between the job seeker and the job.

6. The multi-branch multi-node comparison calculation method based on knowledge graph according to claim 5, characterized in that: The plurality of matching levels correspond to the plurality of matching thresholds, the matching thresholds include an upper threshold and a lower threshold, and the corresponding relationship between the plurality of matching levels and the plurality of matching thresholds is: High matching level ; Medium Match Level ; Low matchmaking level ; in, is the matching coefficient, is the lower threshold corresponding to the high matching level and the upper threshold corresponding to the medium matching level, is the lower threshold corresponding to the medium matching level and the upper threshold corresponding to the low matching level, , .

7. The multi-branch multi-node comparison calculation method based on knowledge graph according to claim 6 is characterized in that: In step 5, the process of visually displaying the comparison results to the user includes: In-depth analysis of each job seeker's matching degree at each skill node, confirming their strengths to the job seeker, and providing specific improvement suggestions for their weaknesses; According to the matching coefficient, the job seekers are sorted from high to low according to the matching degree with the position, and the job seekers with high matching degree are automatically screened out according to the needs of the employer and the matching degree of the job seekers, and the job seekers with the highest matching degree are placed at the top of the recommendation list; Create visual reports that show how well job seekers match positions, and use charts to represent the match level, showing specific match scores and recommendations for each skill node.

Images