An aigc-based person-post matching analysis system and method thereof

Abstract

This invention discloses a job-person matching analysis system and method based on AIGC, relating to the field of artificial intelligence technology. The method includes: acquiring job applicant resume data, job requirement data, and historical job-person matching data from target companies; converting the job applicant resume data and job requirement data into resume text and application text, respectively; performing local semantic feature analysis on the resume text and application text to obtain local resume feature vectors and local application feature vectors; substituting the historical job-person matching data into a graph neural network model to obtain job-person matching feature vectors; performing dynamic feature fusion on the local resume feature vectors and job-person matching feature vectors to obtain a final resume feature vector; calculating the similarity between the final resume feature vector and the local application feature vectors to obtain a job-person matching value; and comparing the job-person matching value with a preset job-person matching threshold to obtain a matching result. This invention improves the accuracy of job-person matching and the completeness of resume information.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence technology, specifically relating to a human-job matching analysis system and method based on AIGC. Background Technology

[0002] AIGC (Generative Artificial Intelligence) is responsible for transforming the knowledge and capabilities acquired from data, models, and computing resources in the field of artificial intelligence, ensuring that these capabilities can be effectively applied to business scenarios. AIGC uses a series of technical components to transform massive amounts of data and pre-trained models into deployable intelligent generation capabilities.

[0003] The prior art (publication number: CN120471595A) discloses a data analysis-based intelligent job matching model management method and system, including: collecting historical recruitment data to construct job seeker feature vectors and job feature vectors; extracting job seeker feature vectors and job feature vectors corresponding to job application results to form a historical job application feedback dataset, and calculating the unbiased feature matching degree to construct a biased feedback dataset; extracting unbiased feedback data based on the biased feedback dataset and the historical job application feedback dataset, and reconstructing and optimizing the historical job application feedback dataset; constructing an unbiased job matching dataset, and using the unbiased job matching dataset to train the intelligent job matching model; applying the trained intelligent job matching model to job recommendation, and collecting subsequent job application result data to optimize the model.

[0004] In the aforementioned patent, a human-job intelligent matching model is trained and optimized using job application data to improve the accuracy of the matching results. However, in practical applications, there are still defects in the correlation between data in the human-job matching task for enterprises, which may result in missing resume information and thus failure to meet the job requirements. Summary of the Invention

[0005] The purpose of this invention is to solve the problem that in practical applications, there are still defects in the correlation between data in the task of matching people with jobs in enterprises, which may result in missing resume information and thus failure to meet the job requirements. Therefore, this invention proposes a person-job matching analysis system and method based on AIGC.

[0006] In a first aspect of this invention, a job-person matching analysis method based on AIGC is first proposed, the method comprising:

[0007] Acquire job application data, job requirements data, and historical job-person matching data from the target company; convert the job application data and job requirements data into resume text and application text, respectively.

[0008] Local semantic features are performed on the resume text and the job application text respectively to obtain local resume feature vector and local job application feature vector;

[0009] Substitute the historical job matching data into the graph neural network model to obtain the job matching feature vector;

[0010] The final resume feature vector is obtained by dynamically fusing the local resume feature vector and the job matching feature vector.

[0011] The similarity between the final resume feature vector and the local job application feature vector is calculated to obtain the job matching value. The matching result is obtained by comparing the job matching value with the preset job matching threshold.

[0012] Optionally, the step of performing local semantic feature analysis on the resume text and job application text to obtain local resume features and local job application features respectively includes:

[0013] The resume text and the job application text are preprocessed separately to generate the resume input sequence and the job application input sequence, respectively. The sequence length is a preset length and they are the same.

[0014] The resume input sequence and the job application input sequence are sequentially input into the word embedding layer and the position embedding layer to generate the first initial embedding vector and the second initial embedding vector, respectively.

[0015] Linear transformations are performed on the first initial embedding vector and the second initial embedding vector to obtain the first query vector, the second query vector, the first key vector, the second key vector, the first value vector, and the second value vector, respectively.

[0016] The first query vector, the first key vector, and the first value vector are concatenated to obtain the first local semantic vector;

[0017] The second query vector, the second key vector, and the second value vector are concatenated to obtain the second local semantic vector;

[0018] Substituting the first local semantic vector and the second local semantic vector into the BERT model respectively, we obtain local resume features and local job application features.

[0019] Optionally, the working principle of the graph neural network model includes:

[0020] Construct an undirected resume graph based on historical job matching data and job application data;

[0021] An updated resume diagram is obtained by updating the undirected resume diagram in real time.

[0022] The feature vector obtained from the updated resume image is determined as the person-job matching feature vector;

[0023] The construction of the undirected resume graph is specifically as follows:

[0024] The job application data is divided into individual request resumes, and each request resume is used as the first node. All historically successful job recruitment resumes from the historical job matching data are used as the second node. The first node and the second node are combined to form a node set. The node set is initialized with local resume features to obtain initial features.

[0025] The initial features of any two nodes are matched with job features to obtain a job matching value. If the job matching value is greater than the preset matching value, an undirected edge is constructed between the two nodes; if the job matching value is less than or equal to the preset matching value, no undirected edge is constructed between the two nodes.

[0026] Obtain all edges and construct an undirected resume graph with the node set.

[0027] Optionally, the step of dynamically fusing the local resume feature vector and the job-person matching feature vector to obtain the final resume feature vector includes:

[0028] The local resume feature vector and the job matching feature vector are convolved to obtain the first local resume feature vector and the first job matching feature vector with C channels.

[0029] The first local resume feature vector and the first job matching feature vector are fused to obtain a first fused feature vector with C channels;

[0030] The first fused feature vector and the first local resume feature vector are concatenated to obtain a first concatenated feature vector with 2C channels;

[0031] The first fused feature vector and the first person-position matching feature vector are concatenated to obtain a second concatenated feature vector with 2C channels;

[0032] The first and second spliced ​​feature vectors are fused to obtain a second fused feature vector with 2C channels;

[0033] Convolution is performed on the second fused feature vector to obtain a first convolutional feature vector with C channels;

[0034] The final resume feature vector is obtained by adding the first convolutional feature vector, the first local resume feature vector, and the first job matching feature vector element by element.

[0035] Optionally, the step of calculating the similarity between the final resume feature vector and the local job application feature vector to obtain a job-person matching value, and comparing the job-person matching value with a preset job-person matching threshold to obtain a matching result includes:

[0036] If the job-person matching value is greater than the preset job-person matching threshold, the data of the currently qualified job applicants will be input to the backend.

[0037] If the job-person matching value is less than or equal to the preset job-person matching threshold, the current job application data will be removed.

[0038] All qualified job application resumes are integrated to obtain matching results.

[0039] The steps for calculating the person-job matching value are as follows:

[0040]

[0041] Where x represents the local job application feature vector, and y represents the final resume feature vector. This represents the magnitude of the local job application feature vector. This represents the magnitude of the final resume feature vector. This represents the person-job matching value.

[0042] In a second aspect of this invention, a job-person matching analysis system based on AIGC is proposed, the system comprising:

[0043] Job matching text module: Acquires job applicant resume data, job requirement data, and historical job matching data from the target company; converts the job applicant resume data and job requirement data into resume text and job application text, respectively.

[0044] Local semantic module: Local semantic features are applied to the resume text and job application text respectively to obtain local resume feature vector and local job application feature vector;

[0045] Graph Neural Network Module: Substitutes the historical job matching data into the graph neural network model to obtain job matching feature vectors;

[0046] Fusion module: Dynamically fuses local resume feature vectors and job-person matching feature vectors to obtain the final resume feature vector;

[0047] Matching Result Module: Calculates the similarity between the final resume feature vector and the local job application feature vector to obtain the job matching value, and compares the job matching value with the preset job matching threshold to obtain the matching result.

[0048] Optionally, the local semantic module includes: a preprocessing module, an embedding module, a linear transformation module, a first concatenation module, a second concatenation module, and a BERT module.

[0049] The preprocessing module is used to preprocess the resume text and the job application text respectively to generate a resume input sequence and a job application input sequence, the sequence lengths being preset lengths and being the same.

[0050] The embedding module is used to input the resume input sequence and the job application input sequence into the word embedding layer and the position embedding layer respectively, and generate a first initial embedding vector and a second initial embedding vector.

[0051] The linear transformation module is used to perform linear transformations on the first initial embedding vector and the second initial embedding vector respectively, to obtain the first query vector, the second query vector, the first key vector, the second key vector, the first value vector, and the second value vector respectively;

[0052] The first concatenation module is used to concatenate the first query vector, the first key vector, and the first value vector to obtain a first local semantic vector;

[0053] The second concatenation module is used to concatenate the second query vector, the second key vector, and the second value vector to obtain the second local semantic vector;

[0054] The BERT module is used to substitute the first local semantic vector and the second local semantic vector into the BERT model to obtain local resume features and local job application features, respectively.

[0055] Optionally, the graph neural network module includes: an undirected resume graph module, a resume graph update module, and a job-person matching feature module.

[0056] The undirected resume graph module is used to construct an undirected resume graph based on historical job matching data and job application resume data.

[0057] The updated resume image module is used to update the resume image in real time based on the undirected resume image.

[0058] The person-job matching feature module is used to obtain the feature vector in the updated resume image and determine it as the person-job matching feature vector.

[0059] The construction of the undirected resume graph is specifically as follows:

[0060] The job application data is divided into individual request resumes, and each request resume is used as the first node. All historically successful job recruitment resumes from the historical job matching data are used as the second node. The first node and the second node are combined to form a node set. The node set is initialized with local resume features to obtain initial features.

[0061] The initial features of any two nodes are matched with job features to obtain a job matching value. If the job matching value is greater than the preset matching value, an undirected edge is constructed between the two nodes; if the job matching value is less than or equal to the preset matching value, no undirected edge is constructed between the two nodes.

[0062] Obtain all edges and construct an undirected resume graph with the node set.

[0063] Optionally, the fusion module includes: a first convolution module, a first feature fusion module, a first feature concatenation module, a second feature concatenation module, a second feature fusion module, a second convolution module, and an addition module.

[0064] The first convolution module is used to perform convolution operations on the local resume feature vector and the job matching feature vector to obtain a first local resume feature vector and a first job matching feature vector with C channels respectively;

[0065] The first feature fusion module is used to fuse the first local resume feature vector and the first job matching feature vector to obtain a first fused feature vector with C channels;

[0066] The first feature splicing module is used to splice the first fused feature vector and the first local resume feature vector to obtain a first spliced ​​feature vector with 2C channels;

[0067] The second feature splicing module is used to splice the first fused feature vector and the first person-job matching feature vector to obtain a second spliced ​​feature vector with 2C channels;

[0068] The second feature fusion module is used to fuse the first concatenated feature vector and the second concatenated feature vector to obtain a second fused feature vector with 2C channels;

[0069] The second convolution module is used to convolve the second fused feature vector to obtain a first convolution feature vector with C channels;

[0070] The addition module is used to add the first convolutional feature vector, the first local resume feature vector, and the first job matching feature vector element by element to obtain the final resume feature vector.

[0071] Optionally, the matching result module includes: a first judgment module, a second judgment module, and an integration module.

[0072] The first judgment module is used to input the current qualified job application resume data to the backend if the person-job matching value is greater than the preset person-job matching threshold.

[0073] The second judgment module is used to remove the current job application data if the job matching value is less than or equal to the preset job matching threshold.

[0074] The integration module is used to integrate all qualified job application resumes to obtain matching results.

[0075] The steps for calculating the person-job matching value are as follows:

[0076]

[0077] Where x represents the local job application feature vector, and y represents the final resume feature vector. This represents the magnitude of the local job application feature vector. This represents the magnitude of the final resume feature vector. This represents the person-job matching value.

[0078] The beneficial effects of this invention are:

[0079] This invention proposes a job matching analysis system and method based on AIGC. By leveraging a graph neural network model, it can effectively mine the complex nonlinear relationships between people and jobs in historical data, thereby improving the accuracy of job matching. The dynamic feature fusion method can enhance the representation ability of resume information and improve the completeness and richness of feature information. Attached Figure Description

[0080] The invention will now be further described with reference to the accompanying drawings.

[0081] Figure 1 A flowchart of a person-job matching analysis method based on AIGC provided in an embodiment of the present invention;

[0082] Figure 2 This is a framework diagram of a job matching analysis system based on AIGC provided in an embodiment of the present invention. Detailed Implementation

[0083] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and B can represent: A alone, A and B simultaneously, and B alone. Furthermore, descriptions involving "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" can explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0084] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0085] This invention provides a method for person-job matching analysis based on AIGC. See also... Figure 1 , Figure 1 A flowchart illustrating a person-job matching analysis method based on AIGC (Artificial Intelligence of Governance) provided in an embodiment of the present invention. The method includes the following steps:

[0086] Acquire job application data, job requirements data, and historical job-person matching data from target companies; convert job application data and job requirements data into resume text and application text.

[0087] Local semantic features are performed on the resume text and the job application text respectively to obtain local resume feature vector and local job application feature vector;

[0088] The historical job matching data is substituted into the graph neural network model to obtain the job matching feature vector.

[0089] The final resume feature vector is obtained by dynamically fusing the local resume feature vector and the job matching feature vector.

[0090] The similarity between the final resume feature vector and the local job application feature vector is calculated to obtain the job matching value. The matching result is obtained by comparing the job matching value with the preset job matching threshold.

[0091] The AIGC-based person-job matching analysis method provided in this invention can effectively obtain the complex nonlinear relationship between people and jobs in historical data through a graph neural network model, thereby improving the accuracy of person-job matching. Furthermore, it improves the completeness of resume information through dynamic feature fusion.

[0092] In one implementation, performing local semantic feature analysis on the resume text and job application text respectively to obtain local resume features and local job application features includes:

[0093] The resume text and the job application text are preprocessed separately to generate the resume input sequence and the job application input sequence, respectively. The sequence length is a preset length and they are the same.

[0094] The resume input sequence and the job application input sequence are sequentially input into the word embedding layer and the position embedding layer to generate the first initial embedding vector and the second initial embedding vector, respectively.

[0095] Linear transformations are performed on the first initial embedding vector and the second initial embedding vector to obtain the first query vector, the second query vector, the first key vector, the second key vector, the first value vector, and the second value vector, respectively.

[0096] The first query vector, the first key vector, and the first value vector are concatenated to obtain the first local semantic vector;

[0097] The second query vector, the second key vector, and the second value vector are concatenated to obtain the second local semantic vector;

[0098] Substituting the first local semantic vector and the second local semantic vector into the BERT model respectively, we obtain local resume features and local job application features.

[0099] In one implementation, the preprocessing steps include segmenting the resume text and job application text into words, removing stop words, and vectorizing the words to obtain the resume input sequence and job application input sequence; the preset sequence length is specifically 256, etc.; the embedded vector specifically includes lexical and positional feature information.

[0100] Specifically, the BERT model is a pre-trained language model, mainly used for language understanding tasks;

[0101] The formulas for partial resume features and partial job application features specifically include:

[0102]

[0103]

[0104] in, and The dimensions are the same as those of the embedding layer (768 dimensions). This represents the resume input sequence. This represents the input sequence for job applications. and Let represent local resume features and local job application features respectively, and Bert() represents the identifier symbol of the pre-trained language model.

[0105] In one implementation, preprocessing generates input sequences of consistent length for both types of text, avoiding extraction biases caused by messy formats and length differences, and ensuring consistency in feature extraction. Word embedding and positional embedding are combined to enrich the dimensions of feature representation, simultaneously capturing basic word semantics and sequence position information, providing high-quality initial input. Linear transformations and vector concatenation enhance feature discriminative power, filtering core semantic information, reducing redundancy, and improving the specificity of local semantic vectors. The BERT model is used to deeply mine the deep semantics of the text, accurately capturing the core meanings of both types of text, improving the accuracy and adaptability of feature extraction. Symmetrical processing logic ensures that the two types of features are in the same feature space, avoiding feature misalignment, reducing the difficulty of subsequent matching, and improving matching efficiency and accuracy.

[0106] In one implementation, the working principle of the graph neural network model includes:

[0107] Construct an undirected resume graph based on historical job matching data and job application data;

[0108] An updated resume image is obtained by updating the undirected resume image in real time.

[0109] The feature vectors obtained from the updated resume image are used to determine the person-job matching feature vectors.

[0110] The construction of an undirected resume graph is as follows:

[0111] The job application data is divided into individual request resumes, and each request resume is used as the first node. All historically successful job postings from the historical job matching data are used as the second node. The first and second nodes are combined to form a node set. Initial features are obtained by initializing the node set using local resume features.

[0112] The initial features of any two nodes are matched with job features to obtain a job matching value. If the job matching value is greater than the preset matching value, an undirected edge is constructed between the two nodes; if the job matching value is less than or equal to the preset matching value, no undirected edge is constructed between the two nodes.

[0113] Obtain all edges and construct an undirected resume graph with the node set;

[0114] In one implementation, real-time updates specifically include:

[0115] Each node in the undirected resume graph is updated using an update function to obtain each updated node, and the graph of each updated node is used as the undirected resume graph.

[0116] In one implementation, the specific formula for the update function of the gated graph neural network (GGNN) in the t-th round is:

[0117]

[0118]

[0119]

[0120]

[0121]

[0122] in, Let represent the i-th row of the adjacency matrix, and T be the transpose of the matrix. This represents the feature vector of all nodes in the previous round. This represents the aggregated neighborhood node feature vector of the i-th node in the t-th round. Let represent the update gate output of node i in round t; σ() represents the sigmoid activation function; This represents the feature vector of the i-th node in the previous round; This represents the reset gate output of the i-th node in round t; , Let represent the learnable weight matrix of the update gate in round t. , This represents the learnable weight matrix of the reset gate. Let represent the candidate feature vector of the i-th node in the t-th round; tanh() represents the hyperbolic tangent activation function. Represents element-wise multiplication. The final feature vector of the i-th node in the t-th iteration.

[0123] One implementation constructs an undirected resume graph by combining historical job-person matching data with job application data. A dual-node setup (requested resumes as the first node and historically successful job postings as the second node) deeply connects past effective experience with current job requirements. Undirected edges are filtered using job feature matching thresholds, effectively eliminating irrelevant node associations and improving the effectiveness of the graph structure. Real-time updates to the resume graph dynamically adapt to changes in job application and recruitment data, avoiding information lag, simplifying matching complexity, and improving matching efficiency.

[0124] In one implementation, the final resume feature vector is obtained by dynamically fusing the local resume feature vector and the job-person matching feature vector, including:

[0125] Convolution operations are performed on the local resume feature vector and the job matching feature vector to obtain the first local resume feature vector and the first job matching feature vector with C channels;

[0126] The first local resume feature vector and the first job matching feature vector are fused to obtain a first fused feature vector with C channels;

[0127] The first fused feature vector and the first local resume feature vector are concatenated to obtain a first concatenated feature vector with 2C channels;

[0128] The first fused feature vector and the first person-position matching feature vector are concatenated to obtain a second concatenated feature vector with 2C channels;

[0129] The first and second concatenated feature vectors are fused to obtain a second fused feature vector with 2C channels;

[0130] Convolution is performed on the second fused feature vector to obtain a first convolutional feature vector with C channels;

[0131] The final resume feature vector is obtained by adding the first convolutional feature vector, the first local resume feature vector, and the first job matching feature vector element by element.

[0132] Specifically, C represents a value greater than 0;

[0133] One implementation method achieves deep coupling and information enhancement of local resume features and job matching features through multi-stage dynamic feature fusion and residual connections. First, convolution unifies the feature channel dimensions to ensure feature scale alignment. Then, step-by-step fusion and cross-concatenation fully exploit the complementarity between resume semantic information and job matching correlation information, avoiding single-feature dominance or loss of key information. Secondary fusion and convolution further refine highly recognizable features. Finally, element-wise addition forms a residual structure, preserving original effective features while mitigating gradient vanishing and improving feature learning stability.

[0134] In one implementation, a similarity calculation is performed on the final resume feature vector and the local job application feature vector to obtain a job-person matching value. The matching result is obtained by comparing the job-person matching value with a preset job-person matching threshold, including:

[0135] If the job-person matching value is greater than the preset job-person matching threshold, the data of the currently qualified job applicants will be input to the backend.

[0136] If the job-person matching value is less than or equal to the preset job-person matching threshold, the current job application data will be removed.

[0137] All qualified job application resumes are integrated to obtain matching results.

[0138] In one implementation, the steps for calculating the person-job matching value are as follows:

[0139]

[0140] Where x represents the local job application feature vector, and y represents the final resume feature vector. This represents the magnitude of the local job application feature vector. This represents the magnitude of the final resume feature vector. This represents the person-job matching value.

[0141] Based on the same inventive concept, this invention also provides a method for a person-job matching system based on AIGC. See also Figure 2 , Figure 2 A framework diagram of a person-job matching analysis system based on AIGC provided for embodiments of the present invention includes:

[0142] Job matching text module: Acquires job applicant resume data, job requirement data, and historical job matching data from target companies; converts job applicant resume data and job requirement data into resume text and application text;

[0143] Local semantic module: Local semantic features are applied to the resume text and job application text respectively to obtain local resume feature vector and local job application feature vector;

[0144] Graph Neural Network Module: Substitutes historical job matching data into the graph neural network model to obtain job matching feature vectors;

[0145] Fusion module: Dynamically fuses local resume feature vectors and job-person matching feature vectors to obtain the final resume feature vector;

[0146] Matching Results Module: Calculates the similarity between the final resume feature vector and the local job application feature vector to obtain the job matching value, and obtains the matching result by comparing the job matching value with the preset job matching threshold.

[0147] The AIGC-based person-job matching analysis system provided in this invention can effectively obtain the complex nonlinear relationship between people and jobs in historical data through a graph neural network model, thereby improving the accuracy of person-job matching. Through dynamic feature fusion, it also improves the completeness of resume information.

[0148] The foregoing has described one embodiment of the present invention in detail, but this content is merely a preferred embodiment and should not be considered as limiting the scope of the present invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the scope of the claims of this invention.

Claims

1. A person-job matching analysis method based on AIGC, characterized in that, The method includes: Acquire job application data, job requirements data, and historical job-person matching data from the target company; convert the job application data and job requirements data into resume text and application text, respectively. Local semantic features are performed on the resume text and the job application text respectively to obtain local resume feature vector and local job application feature vector; Substitute the historical job matching data into the graph neural network model to obtain the job matching feature vector; The final resume feature vector is obtained by dynamically fusing the local resume feature vector and the job matching feature vector. The similarity between the final resume feature vector and the local job application feature vector is calculated to obtain the job matching value. The matching result is obtained by comparing the job matching value with the preset job matching threshold.

2. The AIGC-based person-job matching analysis method according to claim 1, characterized in that, The step of performing local semantic feature analysis on the resume text and job application text respectively to obtain local resume features and local job application features includes: The resume text and the job application text are preprocessed separately to generate the resume input sequence and the job application input sequence, respectively. The sequence length is a preset length and they are the same. The resume input sequence and the job application input sequence are sequentially input into the word embedding layer and the position embedding layer to generate the first initial embedding vector and the second initial embedding vector, respectively. Linear transformations are performed on the first initial embedding vector and the second initial embedding vector to obtain the first query vector, the second query vector, the first key vector, the second key vector, the first value vector, and the second value vector, respectively. The first query vector, the first key vector, and the first value vector are concatenated to obtain the first local semantic vector; The second query vector, the second key vector, and the second value vector are concatenated to obtain the second local semantic vector; Substituting the first local semantic vector and the second local semantic vector into the BERT model respectively, we obtain local resume features and local job application features.

3. The AIGC-based person-job matching analysis method according to claim 1, characterized in that, The working principle of the graph neural network model includes: Construct an undirected resume graph based on historical job matching data and job application data; An updated resume diagram is obtained by updating the undirected resume diagram in real time. The feature vector obtained from the updated resume image is determined as the person-job matching feature vector; The construction of the undirected resume graph is specifically as follows: The job application data is divided into individual request resumes, and each request resume is used as the first node. All historically successful job recruitment resumes from the historical job matching data are used as the second node. The first node and the second node are combined to form a node set. The node set is initialized with local resume features to obtain initial features. The initial features of any two nodes are matched with job features to obtain a job matching value. If the job matching value is greater than the preset matching value, an undirected edge is constructed between the two nodes; if the job matching value is less than or equal to the preset matching value, no undirected edge is constructed between the two nodes. Obtain all edges and construct an undirected resume graph with the node set.

4. The AIGC-based person-job matching analysis method according to claim 1, characterized in that, The process of dynamically fusing local resume feature vectors and job-person matching feature vectors to obtain the final resume feature vector includes: The local resume feature vector and the job matching feature vector are convolved to obtain the first local resume feature vector and the first job matching feature vector with C channels. The first local resume feature vector and the first job matching feature vector are fused to obtain a first fused feature vector with C channels; The first fused feature vector and the first local resume feature vector are concatenated to obtain a first concatenated feature vector with 2C channels; The first fused feature vector and the first person-position matching feature vector are concatenated to obtain a second concatenated feature vector with 2C channels; The first and second spliced ​​feature vectors are fused to obtain a second fused feature vector with 2C channels; Convolution is performed on the second fused feature vector to obtain a first convolutional feature vector with C channels; The final resume feature vector is obtained by adding the first convolutional feature vector, the first local resume feature vector, and the first job matching feature vector element by element.

5. The AIGC-based person-job matching analysis method according to claim 1, characterized in that, The step of calculating the similarity between the final resume feature vector and the local job application feature vector to obtain a job-person matching value, and comparing the job-person matching value with a preset job-person matching threshold to obtain a matching result, includes: If the job-person matching value is greater than the preset job-person matching threshold, the data of the currently qualified job applicants will be input to the backend. If the job-person matching value is less than or equal to the preset job-person matching threshold, the current job application data will be removed. All qualified job application resumes are integrated to obtain matching results; The steps for calculating the person-job matching value are as follows: Where x represents the local job application feature vector, and y represents the final resume feature vector. This represents the magnitude of the local job application feature vector. This represents the magnitude of the final resume feature vector. This represents the person-job matching value.

6. A person-job matching analysis system based on AIGC, characterized in that, The system includes: Job matching text module: Acquires job applicant resume data, job requirement data, and historical job matching data from the target company; converts the job applicant resume data and job requirement data into resume text and job application text, respectively. Local semantic module: Local semantic features are applied to the resume text and job application text respectively to obtain local resume feature vector and local job application feature vector; Graph Neural Network Module: Substitutes the historical job matching data into the graph neural network model to obtain job matching feature vectors; Fusion module: Dynamically fuses local resume feature vectors and job-person matching feature vectors to obtain the final resume feature vector; Matching Result Module: Calculates the similarity between the final resume feature vector and the local job application feature vector to obtain the job matching value, and compares the job matching value with the preset job matching threshold to obtain the matching result.

7. The AIGC-based person-job matching analysis system according to claim 6, characterized in that, The local semantic module includes: a preprocessing module, an embedding module, a linear transformation module, a first concatenation module, a second concatenation module, and a BERT module. The preprocessing module is used to preprocess the resume text and the job application text respectively to generate a resume input sequence and a job application input sequence, the sequence lengths being preset lengths and being the same. The embedding module is used to input the resume input sequence and the job application input sequence into the word embedding layer and the position embedding layer respectively, and generate a first initial embedding vector and a second initial embedding vector. The linear transformation module is used to perform linear transformations on the first initial embedding vector and the second initial embedding vector respectively, to obtain the first query vector, the second query vector, the first key vector, the second key vector, the first value vector, and the second value vector respectively; The first concatenation module is used to concatenate the first query vector, the first key vector, and the first value vector to obtain a first local semantic vector; The second concatenation module is used to concatenate the second query vector, the second key vector, and the second value vector to obtain the second local semantic vector; The BERT module is used to substitute the first local semantic vector and the second local semantic vector into the BERT model to obtain local resume features and local job application features, respectively.

8. The AIGC-based person-job matching analysis system according to claim 6, characterized in that, The graph neural network module includes: an undirected resume graph module, a resume graph update module, and a job-person matching feature module. The undirected resume graph module is used to construct an undirected resume graph based on historical job matching data and job application resume data. The updated resume image module is used to update the resume image in real time based on the undirected resume image. The person-job matching feature module is used to obtain the feature vector in the updated resume image and determine it as the person-job matching feature vector. The construction of the undirected resume graph is specifically as follows: The job application data is divided into individual request resumes, and each request resume is used as the first node. All historically successful job recruitment resumes from the historical job matching data are used as the second node. The first node and the second node are combined to form a node set. The node set is initialized with local resume features to obtain initial features. The initial features of any two nodes are matched with job features to obtain a job matching value. If the job matching value is greater than the preset matching value, an undirected edge is constructed between the two nodes; if the job matching value is less than or equal to the preset matching value, no undirected edge is constructed between the two nodes. Obtain all edges and construct an undirected resume graph with the node set.

9. The AIGC-based person-job matching analysis system according to claim 6, characterized in that, The fusion module includes: a first convolution module, a first feature fusion module, a first feature concatenation module, a second feature concatenation module, a second feature fusion module, a second convolution module, and an addition module. The first convolution module is used to perform convolution operations on the local resume feature vector and the job matching feature vector to obtain a first local resume feature vector and a first job matching feature vector with C channels respectively; The first feature fusion module is used to fuse the first local resume feature vector and the first job matching feature vector to obtain a first fused feature vector with C channels; The first feature splicing module is used to splice the first fused feature vector and the first local resume feature vector to obtain a first spliced ​​feature vector with 2C channels; The second feature splicing module is used to splice the first fused feature vector and the first person-job matching feature vector to obtain a second spliced ​​feature vector with 2C channels; The second feature fusion module is used to fuse the first concatenated feature vector and the second concatenated feature vector to obtain a second fused feature vector with 2C channels; The second convolution module is used to convolve the second fused feature vector to obtain a first convolution feature vector with C channels; The addition module is used to add the first convolutional feature vector, the first local resume feature vector, and the first job matching feature vector element by element to obtain the final resume feature vector.

10. The AIGC-based person-job matching analysis system according to claim 6, characterized in that, The matching result module includes: a first judgment module, a second judgment module, and an integration module. The first judgment module is used to input the current qualified job application resume data to the backend if the person-job matching value is greater than the preset person-job matching threshold. The second judgment module is used to remove the current job application data if the job matching value is less than or equal to the preset job matching threshold. The integration module is used to integrate all qualified job application resumes to obtain matching results; The steps for calculating the person-job matching value are as follows: Where x represents the local job application feature vector, and y represents the final resume feature vector. This represents the magnitude of the local job application feature vector. This represents the magnitude of the final resume feature vector. This represents the person-job matching value.

Images