Resume processing method and device, electronic equipment and storage medium

Abstract

The invention discloses a resume processing method and device, electronic equipment and a storage medium. The method comprises the steps that job hunting behavior information in a preset time period and recruitment information published in the preset time period are acquired, a heterogeneous graph is established, and the heterogeneous graph comprises the incidence relation among resume nodes, position nodes and function nodes; according to the heterogeneous graph and a target vector generation model, vector information of each node in the heterogeneous graph is determined, and the target vector generation model is obtained through training based on a training sample set generated by the heterogeneous graph; according to the vector information of the resume nodes and the vector information of the function nodes, determining a plurality of confidence resume nodes and first N confidence function nodes having the highest similarity with each confidence resume node; and determining function information matched with the resume corresponding to each confidence resume node according to the recruitment information and the first N confidence function nodes with the highest similarity between each confidence resume node and the recruitment information. According to the embodiment of the invention, the accuracy of matching positions for resumes can be improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a resume processing method, device, electronic device and storage medium. Background Art

[0002] With the popularization of the Internet and the increasing maturity of data processing technology, more and more convenience has been brought to users' lives and work. For example, in the recruitment industry, recruiters can post recruitment information on recruitment software and find the job seekers they need in the recruitment software, and job seekers can post resumes on the recruitment software and find suitable positions in the recruitment software.

[0003] In order to improve the user experience of recruitment software, recruitment software also provides the function of matching resumes with positions, so as to find positions that meet the job-seeking intentions and recommend them to job seekers. However, at present, when matching positions for job seekers' resumes, it is often limited to mining the content included in the resume itself, resulting in a large difference between the matching positions provided by the recruitment software and the job seekers' job-seeking intentions, which affects the job seekers' satisfaction with the recruitment software. Summary of the invention

[0004] In view of this, the embodiments of the present application provide a resume processing method, device, electronic device and storage medium, which can improve the accuracy of matching resumes to positions, further help improve the accuracy of matching resumes to positions, and facilitate more accurate generation of resume portraits of job seekers.

[0005] The embodiment of the present application provides a resume processing method, including: obtaining job-seeking behavior information within a preset time period and recruitment information released within a preset time period, establishing a heterogeneous graph, the heterogeneous graph including associations between resume nodes, position nodes and function nodes, and the recruitment information including function information corresponding to the positions one by one; determining vector information of each node in the heterogeneous graph according to the heterogeneous graph and a target vector generation model, the target vector generation model being trained based on a training sample set generated by the heterogeneous graph; determining multiple trusted resume nodes and the top N trusted function nodes with the highest similarity to each trusted resume node according to the vector information of the resume nodes and the vector information of the function nodes; determining function information matching the resume corresponding to each trusted resume node according to the top N trusted function nodes with the highest similarity between the recruitment information and each trusted resume node.

[0006] Optionally, according to the method of the embodiment of the present application, the job search behavior information includes the behavior of viewing positions and submitting resumes, and the recruitment information includes a position identification corresponding to each functional information; based on nodes of preset types, a heterogeneous graph is established, including: determining the relationship information between each resume node and the position node based on the behavior of viewing positions and submitting resumes; determining the relationship information between the position node and the functional node based on each position identification and the position description corresponding to each position identification; generating a heterogeneous graph based on the relationship information between each resume node and the position node, and the relationship information between each position node and the functional node.

[0007] Optionally, according to the method of the embodiment of the present application, the node vector information of the heterogeneous graph is determined according to the heterogeneous graph and the target vector generation model, including: generating a training sample set according to the heterogeneous graph, wherein the training sample set includes a walk sequence obtained by walking the heterogeneous graph according to different meta-paths; training a pre-constructed vector generation model according to the training sample set, and obtaining a target vector generation model after meeting preset training conditions; obtaining the node vector information of the heterogeneous graph through the target vector generation model, the node vector information including the vector information of the resume node and the vector information of the function node.

[0008] Optionally, according to the method of an embodiment of the present application, a training sample set is generated according to a heterogeneous graph, including: according to a preset meta-path, wandering the heterogeneous graph to obtain a wandering sequence set, the wandering sequence set including multiple wandering sequences corresponding to each preset meta-path; according to the order of nodes in each wandering sequence in the wandering sequence set and K nodes in the neighborhood of each node, a training sample set is generated, where K is a positive integer.

[0009] Optionally, according to the method of an embodiment of the present application, the training sample set includes multiple positive training samples and multiple negative training samples; according to the order of nodes in each walking sequence and the K nodes in the neighborhood of each node, a training sample set is generated, including: for each walking sequence, a current node is combined with each of the K nodes in the neighborhood of the current node to obtain a positive training sample, the current node is any node in the walking sequence, the current node in each positive training sample is an input sample node, and the other node is an output positive sample node; and, for the input sample node in each positive training sample, L nodes are randomly obtained from the multiple walking sequences included in the first walking sequence set as output negative sample nodes, and the L output negative sample nodes are respectively combined with the input sample nodes to obtain L negative training samples corresponding to each positive training sample.

[0010] Optionally, according to the method of the embodiment of the present application, multiple trusted resume nodes and the top N trusted functional nodes with the highest similarity to each trusted resume node are determined based on the vector information of the resume node and the vector information of the functional node, including: determining the node degree of each vector node and the node degree of each functional node based on the heterogeneous graph; obtaining the vector information of the first resume node and the vector information of the first functional node in the node vector information based on the node degree of each vector node and the node degree of each functional node, wherein the node degree corresponding to the first resume node is greater than a first threshold, and the node degree corresponding to the vector information of the first functional node is greater than a second threshold; determining the first similarity value between each first resume node and each first functional node based on the vector information of the first resume node and the vector information of the first functional node; determining each first resume node as a trusted resume node based on the first similarity value between each first resume node and each first functional node, and determining the first functional nodes corresponding to the largest top N first similarity values ​​as trusted functional nodes.

[0011] Optionally, according to the method of the embodiment of the present application, the function information that matches the resume corresponding to each trusted resume node is determined based on the top N trusted function nodes with the highest similarity between the recruitment information and each trusted resume node, including: for each trusted resume node, determining a target trusted function node among the N trusted function nodes, the first similarity value corresponding to the target trusted function node being greater than a third threshold; obtaining the function information corresponding to each target trusted function node in the recruitment information; for each trusted resume node, determining the function information corresponding to the target trusted function node as the function information that matches the resume corresponding to the trusted resume node.

[0012] Optionally, according to the method of the embodiment of the present application, before determining the target trusted function node among N trusted function nodes, the method further includes: for each resume corresponding to the trusted resume node, according to the job-seeking behavior information within a preset time period, determining a first function information set corresponding to the trusted resume node, the first function information set including a plurality of first function information, wherein the first function information is the function information viewed by the job seeker corresponding to the trusted resume node or the function information corresponding to the position applied for by the job seeker; for each trusted resume node, determining a second similarity value between each first function information in the corresponding first function information set and each first function node in the heterogeneous graph; for For each confident resume node, obtain the functional information corresponding to the maximum M second similarity values ​​corresponding to each first functional information in the first functional information set, and obtain the second functional information set, where M is a positive integer greater than N; for each confident resume node, determine Y confident functional nodes according to the preset node screening conditions and the second functional information set, where Y is a positive integer less than or equal to N, and the preset node screening conditions include: when the functional information corresponding to the confident functional node is not included in the second functional information set, screen out the confident functional node; determine the target confident functional node among the N confident functional nodes, including: determine the target confident functional node among the Y confident functional nodes.

[0013] An embodiment of the present application provides a resume processing device, which includes: an acquisition module, which is used to acquire job-seeking behavior information within a preset time period and recruitment information released within a preset time period, and establish a heterogeneous graph based on nodes of preset types, wherein the nodes of the preset types include resume nodes, position nodes and function nodes, and the recruitment information includes function information corresponding to the positions one by one; a processing module, which is used to determine the node vector information of the heterogeneous graph according to the heterogeneous graph and a target vector generation model, wherein the node vector information includes the vector information of the resume node and the vector information of the function node, and the target vector generation model is trained based on a training sample set generated by the heterogeneous graph; the processing module is also used to determine multiple trusted resume nodes and the top N trusted function nodes with the highest similarity to each trusted resume node according to the vector information of the resume node and the vector information of the function node; the processing module is also used to determine the function information matching the resume corresponding to each trusted resume node according to the top N trusted function nodes with the highest similarity between the recruitment information and each trusted resume node.

[0014] An embodiment of the present application provides an electronic device, which includes a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the steps of the above method are implemented.

[0015] An embodiment of the present application provides a computer-readable storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the steps of the above method are implemented.

[0016] An embodiment of the present application provides a computer program product, which includes computer program instructions. When the computer program instructions are executed by a processor, the steps of the above method are implemented.

[0017] According to the embodiment of the present application, a heterogeneous graph including resume nodes, position nodes and function nodes is established for the job-seeking behavior information and the published recruitment information obtained within a preset time period, so as to conveniently represent the complex relationship between nodes of each preset type, so as to find the function information with a higher matching degree with the resume. Next, the vector information of each node in the heterogeneous graph can be obtained through the target vector generation model, wherein, since the target vector generation model is obtained by training the training samples generated based on the heterogeneous graph, it is conducive to fully exploring the job-seeking intention of the resume corresponding to the resume node, and improving the accuracy and reliability of the vector information output by the model. Afterwards, the vector information of the resume node and the vector information of the function node are combined to find the confident resume node, wherein, since the confident resume node has more connection relationships with other nodes in the heterogeneous graph, it is conducive to improving the accuracy of the matching result of the function information for each confident resume node, and finally, according to the top N confident function nodes with the highest similarity between the recruitment information and each confident resume node, the function information matching the resume corresponding to each confident resume node is accurately determined, thereby, it is conducive to reducing the gap between the matching position provided by the recruitment software and the job-seeking intention of the job seeker, and improving the job seeker's satisfaction with the recruitment software. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solution of the embodiments of the present application, the following briefly introduces the drawings in the embodiments of the present application.

[0019] Figure 1 It is a schematic diagram of the system architecture of an embodiment of the present application.

[0020] Figure 2 It is a flowchart of the resume processing method of the embodiment of the present application.

[0021] Figure 3 It is a schematic diagram of a heterogeneous graph according to an embodiment of the present application.

[0022] Figure 4 It is a schematic diagram of the walking process of an embodiment of the present application.

[0023] Figure 5 It is a schematic diagram of obtaining a wandering sequence according to an embodiment of the present application.

[0024] Figure 6 It is a flowchart of another resume processing method of an embodiment of the present application.

[0025] Figure 7It is a structural block diagram of the resume processing device of an embodiment of the present application.

[0026] Figure 8 It is a schematic diagram of an electronic device used to implement the resume processing method of an embodiment of the present application. DETAILED DESCRIPTION

[0027] The principles and spirit of the present application will be described below with reference to several exemplary embodiments. It should be understood that the purpose of providing these embodiments is to make the principles and spirit of the present application clearer and more thorough, so that those skilled in the art can better understand and implement the principles and spirit of the present application. The exemplary embodiments provided herein are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments herein, all other embodiments obtained by ordinary technicians of the art without creative work are within the scope of protection of this application.

[0028] Embodiments of the present application relate to terminal devices and / or servers. Those skilled in the art will appreciate that the embodiments of the present application may be implemented as a system, apparatus, device, method, computer-readable storage medium, or computer program product. Therefore, the present disclosure may be specifically implemented as at least one of the following forms: complete hardware, complete software, or a combination of hardware and software. According to the embodiments of the present application, the present application requests protection for a resume processing method, apparatus, electronic device, and storage medium. Figure 1 A schematic diagram of a system architecture of an embodiment of the present application is shown. Figure 1 As shown, the system includes a terminal device 102 and a server 104. The terminal device 102 may include at least one of the following: a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart TV, various wearable devices, an augmented reality AR device, a virtual reality VR device, and the like. The terminal device 102 may use recruitment software. For example, a client may be installed on the terminal device 102. The client may be a client that specifically performs a specific function (such as an application app), or a client that is embedded with a variety of application applets (with different functions), or a client that logs in through a browser. Job seekers may perform operations on the terminal device 102. For example, job seekers may open the terminal device 102, use the recruitment software, view positions in the recruitment software, post resumes, and perform other operations. Optionally, the recruiter may also perform operations on the terminal device, for example, post positions and position-related introductions in the recruitment software, and perform other operations.

[0029] The server 104 can provide storage and retrieval functions, for example, it can store information such as resumes of job seekers and positions of recruiters. After the terminal device 102 receives the instructions input by the job seeker or the recruiter, it sends the request information including the instructions to the server 104. After receiving the request information, the server 104 performs corresponding processing, and then returns the processing result information to the terminal device 102. The user instructions are completed through a series of data processing and information interaction. Based on these instructions, the server 104 can obtain job search behavior information and recruitment information. Based on this, the server 104 can regularly obtain job search behavior information and recruitment information, analyze the job search behavior information and recruitment information, so as to provide a suitable position for the resume, so as to improve the convenience of job seekers using the recruitment software, and at the same time, it is also beneficial for the recruiter to find suitable employees.

[0030] It should be noted that the acquisition, storage, use, and processing of data in the embodiments of the present application are in compliance with the relevant provisions of national laws and regulations.

[0031] In this document, terms such as first, second, third, etc. are only used to distinguish one entity (or operation) from another entity (or operation), but not to require or imply any order or relationship between these entities (or operations).

[0032] Figure 2 A schematic diagram showing a resume processing method according to an embodiment of the present application is shown. Figure 2 As shown, the resume processing method may include the following steps 201 to 204.

[0033] Step 201, obtain job-seeking behavior information within a preset time period and recruitment information released within a preset time period, and establish a heterogeneous graph, which includes the association relationship between resume nodes, position nodes and function nodes, and the recruitment information includes function information corresponding to the position one by one.

[0034] Step 202, determining the vector information of each node in the heterogeneous graph according to the heterogeneous graph and the target vector generation model, wherein the target vector generation model is trained based on a training sample set generated based on the heterogeneous graph.

[0035] Step 203, based on the vector information of the resume node and the vector information of the function node, determine multiple trusted resume nodes and the top N trusted function nodes with the highest similarity to each trusted resume node.

[0036] Step 204, based on the top N trusted function nodes with the highest similarity between the recruitment information and each trusted resume node, determine the function information that matches the resume corresponding to each trusted resume node.

[0037] The above steps are described in detail below in conjunction with specific embodiments, as shown below.

[0038] In an embodiment of the present application, the executor of the resume processing method may be an electronic device with data processing capabilities, such as a server. Optionally, the server may be a physical server, a virtual server, or a cloud server. No specific restrictions are imposed on the form of the server.

[0039] Regarding the above step 201, job-seeking behavior information and recruitment information can be obtained from the service device of the recruitment software respectively. For example, the job seeker's operation behaviors such as viewing positions and submitting resumes in the recruitment software can trigger the service device to provide the job seeker with feedback information corresponding to the operation behavior. The recruiter's operation behaviors such as posting positions and position-related introductions in the recruitment software can also trigger the service device to provide feedback information corresponding to the operation behavior.

[0040] For example, in recruitment information, positions and functional information are interrelated and correspond one to one, but different positions may correspond to different or the same functional information. Specifically, a position may represent a job role or title. For example, within the same company, there may be one or more positions with the same function. Functional information may include specific work content, such as the skills or abilities required for a position. Different positions may correspond to the same functional information. For example, when personnel changes, the position may change, but the function may not change. For another example, different companies offer two different positions, but the functions corresponding to the two positions may be the same.

[0041] After obtaining the job-seeking behavior information and recruitment information, the job-seeking behavior information and recruitment information are analyzed to find the functional information that matches the resume. Optionally, the preset time period can be used as the time interval for each execution of the resume processing method, and the preset time period can be set according to the application requirements. By timely analyzing the latest job-seeking behavior information and recruitment information, it is helpful to improve the reliability of matching the functional information for the resume.

[0042] A heterogeneous graph includes multiple nodes and multiple edges, wherein nodes with associated relationships can be connected through edges. For example, a job seeker can submit a resume to a desired position during the job search process, thereby finding nodes and positions with job search relationships through job search behavior information, and then establishing connections between different resume nodes and position nodes based on the job search behavior, wherein each resume can correspond to a resume node, and each position can correspond to a position node.

[0043] The heterogeneous graph also includes function nodes, where each function information corresponds to a function node. By analyzing the recruitment information, the association between positions and function information can be found. For example, different positions may correspond to different function information or the same function information. In the case where multiple positions correspond to the same function information, multiple positions corresponding to multiple position nodes can be connected to the same function node.

[0044] According to the embodiment of the present application, after the nodes with associated relationships are connected, the relationship between resume, position and function is conveniently expressed through a heterogeneous graph.

[0045] After the heterogeneous graph is established, next, step 202 is involved to determine the vector information of each node in the heterogeneous graph, that is, to determine the vector information of each resume node, the vector information of the position node and the vector information of the function node in the heterogeneous graph.

[0046] Optionally, the pre-trained model can output the vector information of each node in the heterogeneous graph.

[0047] In the embodiment of the present application, the target vector generation model is trained based on a training sample set generated based on a heterogeneous graph. It is understandable that the training sample set includes multiple training samples, wherein the training samples include input data and output data, and the input data and output data are nodes selected from the heterogeneous graph. Since the training samples of the model are derived from the heterogeneous graph, the target vector generation model can more fully learn the semantic relationship between each node, so that the vector information of the node has higher accuracy, which is conducive to finding more suitable functional information for the resume.

[0048] Next, in step 203 and step 204, optionally, the trusted resume nodes and trusted function nodes can be determined based on the relationship between the nodes and the edges. By determining the trusted nodes, noise can be reduced, the quality of the overall data set can be improved, and it is helpful to find more suitable function information for the resume.

[0049] According to the vector information of each resume node and the vector information of each trusted function node, the similarity between each trusted vector node and each trusted function node is calculated, and then the top N trusted function nodes with the highest similarity to each trusted resume node are found. Finally, the trusted resume nodes are matched with the trusted function nodes to obtain the function information that meets the job seeker's needs.

[0050] According to the embodiment of the present application, by constructing a heterogeneous graph including resume nodes, position nodes and function nodes, the complex relationship between job seekers and positions can be accurately and intuitively represented. The heterogeneous graph can also more easily capture the multi-dimensional correlation between different nodes, and conveniently represent the complex relationship between resume nodes and position nodes and function nodes. Secondly, by selecting trusted function nodes, the function information corresponding to each trusted resume node can be better matched with the recruitment information, reducing the mismatch rate of job seekers in the job search process and improving the user experience and satisfaction of the recruitment software.

[0051] In some embodiments of the present application, job search behavior information includes job viewing behavior and resume submission behavior, and recruitment information includes a job identification corresponding to each functional information. Based on this, based on the preset type of nodes, the establishment of a heterogeneous graph can specifically refer to the following steps: according to the job viewing behavior and resume submission behavior, determine the relationship information between each resume node and the job node; according to each job identification and the job description corresponding to each job identification, determine the relationship information between the job node and the functional node; according to the relationship information between each resume node and the job node, and the relationship information between each job node and the functional node, generate a heterogeneous graph.

[0052] For example, based on the job viewing behavior and resume submission behavior, it can be determined which resumes interact with which jobs. The relationship not only reflects the job seeker's submission history, but also reflects the job seeker's interest preference when selecting jobs. For example, the job seeker's interest in the job he submitted his resume to is higher than the interest in the job he only viewed.

[0053] Different positions may correspond to different functional information or the same functional information. In the case where multiple positions correspond to the same functional information, multiple position nodes corresponding to the multiple positions may be connected to the same functional node.

[0054] For example, Figure 3 This is a schematic diagram of a heterogeneous graph provided by an embodiment of the present application, which shows the job seeker's behavior of viewing positions and submitting resumes. The job seeker's resume (R) is associated with the position (J). The resume (R), position (J), and function (F) are used as nodes, and viewing and submitting are used as edges to construct a heterogeneous graph of resumes, positions, and functions. Figure 3 As shown, R1, R2, R3, R4, and R5 represent different resume nodes, J1, J2, J3, and J4 represent different positions, and F1, F2, and F3 represent different functional information.

[0055] In an embodiment of the present application, after establishing connections between nodes with associated relationships, the relationship between resumes, positions and functions is conveniently expressed through a heterogeneous graph, so as to subsequently improve the accuracy of mining job seekers' job-seeking intentions and make the matching results of resumes and functions more in line with actual job-seeking needs.

[0056] In some embodiments of the present application, the steps of determining the node vector information of the heterogeneous graph according to the heterogeneous graph and the target vector generation model may specifically refer to the following steps 301 to 303.

[0057] Step 301: Generate a training sample set according to a heterogeneous graph, wherein the training sample set includes a walk sequence determined by walking the heterogeneous graph according to different meta-paths.

[0058] Step 302, training the pre-built vector generation model according to the training sample set, and obtaining the target vector generation model after reaching the preset training conditions.

[0059] Step 303, obtaining node vector information of the heterogeneous graph through the target vector generation model, where the node vector information includes vector information of resume nodes and vector information of function nodes.

[0060] In the embodiment of the present application, the training samples in the training sample set are determined by a walk sequence obtained by walking a heterogeneous graph according to different meta paths.

[0061] The heterogeneous graph includes different types of nodes. For example, in the embodiment of the present application, it includes three types of nodes: resume nodes, position nodes, and function nodes, as well as the relationship between each node. A meta-path refers to a specific path pattern between different types of nodes and edges in a heterogeneous graph. For example, the path from a resume node to a position node and then to a function node constitutes a meta-path. By setting different meta-paths, multiple random walks can be performed in the heterogeneous graph to generate a walk sequence that reflects the complex relationship between nodes. Each walk sequence actually represents a node relationship under a certain association mode and can be used as training data for subsequent models. For example, a pre-constructed meta-path, such as: (1) resume → position → function → position → resume; (2) position → function → position; (3) resume → position → resume; (4) position → resume → position. Corresponding to each meta-path, multiple walk sequences are found by walking.

[0062] In some embodiments, the pre-built vector generation model includes but is not limited to supporting multiple graph neural network model frameworks (Deep Graph Library, DGL), PyG, NeuGraph and other model frameworks for construction. Optionally, in the embodiment of the present application, a pre-built vector generation model based on the DGL model framework is used to perform graph embedding on the relationship between nodes in the heterogeneous graph.

[0063] For example, the pre-built vector generation model can be a model developed based on the Skip-gram model, where the Skip-gram model is a prediction model in natural language processing that can be used to learn continuous vector representations of words. After training, the pre-built vector generation model can predict the context of a given word based on the given word.

[0064] The preset training condition may be that the function value of the loss function approaches a preset value, or that the training cycle (Epoch) of the model reaches a preset number of times, etc. The specific preset training condition may be set according to actual training needs.

[0065] After the model training is completed, the target vector generation model is obtained. The target vector output model can be used to obtain the node vector information of the heterogeneous graph, which can include the vector information of the resume node and the vector information of the function node, and can also include the vector information of other nodes, such as the vector information of the position node.

[0066] According to the embodiment of the present application, by using a heterogeneous graph generated based on job-seeking behavior information and recruitment information to generate training samples and train a vector generation model, it is beneficial to deeply explore the potential relationship between job seekers' job-seeking intentions and job information, thereby improving the accuracy of matching results. By setting different meta-paths, it is convenient to flexibly adjust the roaming strategy according to the needs of actual recruitment scenarios to meet personalized matching needs. The node vector generation method based on heterogeneous graphs and target vector generation models can effectively mine useful information from job-seeking behavior information and recruitment information, and convert it into a vectorized representation to support subsequent matching processes.

[0067] In some embodiments of the present application, according to a heterogeneous graph, the step of generating a training sample set can refer to the following steps: according to a preset meta-path, wandering the heterogeneous graph, obtaining a wandering sequence set, the wandering sequence set including multiple wandering sequences corresponding to each preset meta-path; according to the order of nodes in each wandering sequence in the wandering sequence set and the K nodes in the neighborhood of each node, generating a training sample set, K is a positive integer.

[0068] For example, Figure 4 is a schematic diagram of a walking process provided in an embodiment of the present application, combined with Figure 4 As shown. Combined Figure 4 As shown in (a), based on the meta-path (1) resume → position → function → position → resume, we can get the walk sequence: R2-J2-F1-J3-R5. Figure 4 As shown in (b), based on the meta-path (2) position → function → position, we can get the walking sequence: J2-F1-J3. Corresponding to each meta-path, after walking the heterogeneous graph, we can get multiple walking sequences. Here, we will not list each walking sequence corresponding to each meta-path one by one.

[0069] In some embodiments, the connection lines between nodes in the heterogeneous graph are undirected edges. The walking direction can be randomly selected at any node.

[0070] Optionally, in the process of walking on a heterogeneous graph, walking parameters can be pre-set, such as walking depth and the number of times each node in the walking process is used as the starting point of the walking. Continuing with meta-path 1 as an example, if the walking depth is 1, the obtained walking sequence can be R2-J2-F1-J3-R5. If the walking depth is 2, based on the walking sequence with a depth of 1, its sequence end R5 can be used as the starting point, and the walking in the heterogeneous graph continues according to meta-path 1. The obtained walking sequence can be R2-J2-F1-J3-R5-J2-F1-J2-R3. Optionally, the walking depth and the number of times each node in the walking process is used as the starting point of the walking can be set according to application requirements, and no specific restrictions are made here.

[0071] Optionally, in the embodiment of the present application, the weight between the resume node and the position node can also be set. Figure 5 As shown in the figure, V represents the job seeker's viewing behavior, A represents the job seeker's application behavior, and the number after the colon represents the weight of the edge. The larger the weight, the greater the probability that the node will be selected during the walk. The advantage of this is that it is conducive to fully mining the intention information of the job seeker's behavior and converting it into a vectorized representation to support the subsequent matching process.

[0072] In each walk sequence, assuming that a node represents a resume node, the K neighboring nodes of the node may include the position node or function node that has interacted with it. By using the sequential relationship between these nodes and their neighboring nodes as training samples, the relationship between resumes and positions and functions can be further analyzed, and training data can be provided for subsequent vector generation models. The value of K can be adjusted according to actual needs. Generally, the larger K is, the more neighboring nodes around the node can participate in the training, thereby generating richer training samples.

[0073] According to the embodiment of the present application, the meta-path-based random walk can capture the multi-dimensional associations between different types of nodes in the heterogeneous graph, so that the training sample set is not limited to the specific relationship of a certain type of node, but can reflect information in multiple dimensions such as resume, position and function. Based on this, it is helpful to better mine the potential matching relationship between job seekers and positions, so that the subsequent vector generation model can more accurately characterize the matching degree between resume and function.

[0074] In some embodiments of the present application, the training sample set includes multiple positive training samples and multiple negative training samples. Optionally, the training sample set is generated according to the order of nodes in each walk sequence and the K nodes in the neighborhood of each node, and the positive training samples and negative training samples can be obtained by referring to the following steps.

[0075] Specifically, for each walking sequence, the current node is combined with each of the K nodes in the neighborhood of the current node to obtain a positive training sample, the current node is any node in the walking sequence, the current node in each positive training sample is the input sample node, and the other node is the output positive sample node.

[0076] For each input sample node in the positive training sample, L nodes are randomly obtained from multiple walking sequences included in the first walking sequence set as output negative sample nodes, and the L output negative sample nodes are respectively combined with the input sample nodes to obtain L negative training samples corresponding to each positive training sample.

[0077] For example, taking the walk sequence R2-J2-F1-J3-R5 as an example, R2, J2, F1, J3 and R5 will be used as the center point, and combined with each of the K nodes in its neighborhood. Taking K=2 as an example, R2 is the center point, and the neighborhood nodes of R2 are J2 and F1 respectively. Thus, two positive samples can be obtained: R2-J2 and R2-F1. Among them, R2 is the input sample node in the positive training sample.

[0078] J2 is the center point, and its neighboring nodes are R2, F1, and J3. Thus, three positive samples can be obtained: J2-R2, J2-F1, and J2-J3. Among them, J2 is the input sample node in the positive training sample.

[0079] F1 is the center point, and its neighboring nodes are R2, F1, J3 and R5, thus, four positive samples can be obtained: F1-R2, F1-J2, F1-J3, F1-R5. Among them, F1 is the input sample node in the positive training sample.

[0080] Based on the same positive sample generation logic, three positive samples can be obtained when J3 is the center point, and two positive samples can be obtained when R5 is the center point. They are not listed here one by one.

[0081] Optionally, there may be duplicate samples in the generated samples, for example, a sample consisting of a resume node and a job information node. The more times this sample appears, the more interested the job seeker is in the job information corresponding to the resume node, which is conducive to improving the reliability of mining the jobs that job seekers are interested in.

[0082] Usually, there are many nodes in a heterogeneous graph. In order to speed up model training, negative samples can be generated by negative sampling. Taking meta-path (1) as an example, R1~R5 are used as starting points respectively. According to the preset depth, the walking paths are obtained, and a total of 5 paths are obtained. Among all the nodes included in the 5 paths, nodes are selected to generate negative samples. For example, for the positive sample F1-R2, 5 nodes are randomly selected from all R, J, and F as negative samples according to probability: F1-F5, F1-R7, F1-R3, F1-J4, and F1-J1.

[0083] It can be seen that in the process of selecting negative samples, the frequency of node occurrence is used as the probability of being selected as a negative sample. That is, the higher the frequency of occurrence of the node, the greater the probability of being selected as a negative sample. The advantage of this setting is to strengthen the training of high-frequency nodes and avoid the model being biased by high-frequency nodes.

[0084] It is understandable that the number of negative samples corresponding to each positive sample is set according to actual application requirements and performance, and no specific limitation is made here.

[0085] According to the embodiment of the present application, the node vector generation method based on the heterogeneous graph and the target vector generation model can effectively mine useful information from job search behavior information and recruitment information, and convert it into a vectorized representation to support the subsequent matching process.

[0086] In some embodiments of the present application, multiple trusted resume nodes and the top N trusted functional nodes with the highest similarity to each trusted resume node are determined based on the resume node vector information and the functional node vector information. For details, please refer to steps 401 to 404.

[0087] Step 401, determining the node degree of each vector node and the node degree of each functional node according to the heterogeneous graph.

[0088] Step 402, according to the node degree of each vector node and the node degree of each functional node, obtain the vector information of the first resume node and the vector information of the first functional node in the node vector information, wherein the node degree corresponding to the first resume node is greater than the first threshold, and the node degree corresponding to the vector information of the first functional node is greater than the second threshold.

[0089] Step 403: Determine a first similarity value between each first resume node and each first function node according to the vector information of the first resume node and the vector information of the first function node.

[0090] Step 404, according to the first similarity value between each first resume node and each first function node, each first resume node is determined as a trusted resume node, and the first function nodes corresponding to the largest first N similarity values ​​are determined as trusted function nodes.

[0091] Specifically, the degree of a node refers to the number of edges directly connected to a node. In the embodiment of the present application, the edges in the heterogeneous graph are directional edges, and the degree of a node refers to the total number of edges connected to the node. For example, if a node is connected to three other nodes, then its degree is 3.

[0092] Optionally, in an embodiment of the present application, the first threshold and the second threshold may be the same or different, and may be set specifically according to the application scenario, and no specific limitation is made here. For example, if the first threshold and the second threshold are both 9, then in all the vector information, only the vector information of nodes greater than 10 is retained, and the vector information of nodes less than or equal to 9 is filtered out, so that the vector information of the first resume node is obtained in the node vector information. Optionally, the first resume node can be directly determined as a trusted resume node.

[0093] After determining the first resume node and the first functional node, the first similarity value of each first resume node and each first functional node can be determined according to the vector information of the first resume node and the vector information of the first functional node. Optionally, the similarity between nodes can be measured by vector distance. When performing similarity calculation, the cosine distance between two vectors is calculated, and the smaller the distance, the higher the similarity.

[0094] Finally, each first resume node can be determined as a trusted resume node based on the first similarity value between each first resume node and each first functional node, and the first functional nodes corresponding to the largest first N similarity values ​​can be determined as trusted functional nodes.

[0095] In some embodiments, in the process of calculating similarity, the calculation speed can be improved based on the Annoy search and matching algorithm and the Faiss search algorithm. Taking the Annoy search and matching algorithm as an example, the Annoy vector search index of each first function node can be used, and for each first resume node vector information, the Annoy cosine distance metric is used to recall the closest preset number of function information, and obtain the similarity score of each function.

[0096] In some embodiments, optionally, in order to further improve the accuracy of the matching results, optionally, for each resume, the N trusted function nodes that match it can be thresholded again. Specifically, it can involve the above step 404, according to the top N trusted function nodes with the highest similarity between the recruitment information and each trusted resume node, determine the function information that matches the resume corresponding to each trusted resume node, and refer to the following steps 501 to 503.

[0097] Step 501, for each confidence resume node, determine a target confidence function node among N confidence function nodes, and the first similarity value corresponding to the target confidence function node is greater than a third threshold;

[0098] Step 502, obtaining the function information corresponding to each target trusted function node in the recruitment information;

[0099] Step 503, for each trusted resume node, determine the function information corresponding to the target trusted function node as the function information that matches the resume corresponding to the trusted resume node.

[0100] According to the embodiment of the present application, by performing a truncation of a higher fixed threshold according to the similarity score, the relevance between the function and the resume can be effectively improved, thereby improving the accuracy of the matching result.

[0101] In some embodiments of the present application, in order to reduce the interference of irrelevant job information and improve the accuracy of the matching results, optionally, for each resume, the jobs corresponding to the target trusted job nodes that match it are filtered. Specifically, before executing step 501 and determining the target trusted job node among N trusted job nodes, the following steps 501(a) to 501(d) may be referred to.

[0102] Step 501(a), for each resume corresponding to a trusted resume node, determine the first functional information set corresponding to the trusted resume node based on the job-seeking behavior information within a preset time period, the first functional information set includes multiple first functional information, wherein the first functional information is the functional information that the job seeker corresponding to the trusted resume node has viewed or the functional information corresponding to the position the job seeker has applied for.

[0103] Step 501(b), for each trusted resume node, determine a second similarity value between each first function information in the corresponding first function information set and each first function node in the heterogeneous graph;

[0104] Step 501(c), for each trusted resume node, obtain the functional information corresponding to the largest M second similarity values ​​corresponding to each first functional information in the first functional information set, to obtain a second functional information set, where M is a positive integer greater than N;

[0105] Step 501(d), for each trusted resume node, according to a preset node screening condition and the second function information set, determine Y trusted function nodes, where Y is a positive integer less than or equal to N, and the preset node screening condition includes: if the function information corresponding to the trusted function node is not included in the second function information set, screen out the trusted function node;

[0106] Based on this, step 501 determines a target trusted functional node among N trusted functional nodes, including: determining a target trusted functional node among Y trusted functional nodes.

[0107] Specifically, each trusted resume node corresponds to a resume, thereby finding the positions that the job seeker has viewed and applied for based on the resume, and obtaining the functional information corresponding to these positions, thereby obtaining a first functional information set.

[0108] After obtaining the first function information set, in order to accurately expand the scope of each confidence resume node corresponding to the first function information set, find the function information that the job seeker may be interested in. Through step 501 (b) and step 501 (c), the second similarity value between each first function information in the first function information set and each first function node in the heterogeneous graph can be calculated. Afterwards, for each first function information in the first function information set, the second function information set corresponding to the confidence resume node is obtained through the function information corresponding to the largest M second similarity values. For example, each confidence resume node corresponds to the first function information set including 10 first function information, and for each first function information, the function information corresponding to the largest M second similarity values ​​is found. Taking the value of M as 30 as an example, the function information corresponding to the largest 30 second similarity values ​​is found for each first function information. Based on this, each confidence resume node corresponds to the first function information set including 10 first function information, and the obtained second function information set can include 300 function information.

[0109] Based on this, the first functional information set is effectively expanded, which helps to find functional information that job seekers may be interested in.

[0110] Next, for each resume, the functions corresponding to the target trusted function nodes that match it can be filtered according to the second function information set. Exemplarily, the trusted function node can be filtered out when the function information corresponding to the trusted function node is not included in the second function information set.

[0111] After the screening and filtering are completed, the target confidence function node is determined among the N confidence function nodes, which may specifically include: determining the target confidence function node among the Y confidence function nodes.

[0112] It can be understood that, for each resume, if the N trust function nodes corresponding to the resume are not screened out, then N is equal to Y, that is, all N trust function nodes are confirmed as target trust function nodes; if there are trust function nodes that are screened out among the N trust function nodes corresponding to the resume, then Y is less than N.

[0113] According to the embodiment of the present application, by performing relevance filtering, similarity truncation and other processing on the N trusted function nodes corresponding to each resume, it is helpful to reduce the gap between the matching positions provided by the recruitment software and the job seekers' job-seeking intentions, and improve the job seekers' satisfaction with the recruitment software.

[0114] The implementation methods and advantages of the embodiments of the present application are described above through multiple embodiments. In order to clearly introduce the embodiments of the present application, the specific processing process of the embodiments of the present application is described in detail below with reference to specific examples.

[0115] Figure 6 is a flow chart of another resume processing method provided in an embodiment of the present application, combined with Figure 6 As shown, the solution may include the following steps 601 to 610.

[0116] Step 601, obtaining job search behavior data and recruitment information, and constructing a heterogeneous graph.

[0117] Step 602, obtaining the meta-path.

[0118] Step 603, traverse the resume nodes, position nodes and function nodes in the heterogeneous graph, walk in the heterogeneous graph according to the meta-path, and generate multiple walk sequences corresponding to each meta-path.

[0119] Step 604: Sample each walking sequence to generate multiple positive samples and multiple negative samples associated with each positive sample.

[0120] Step 605: construct a vector generation model.

[0121] Step 606: training the vector generation model according to the positive training samples and the negative training samples.

[0122] Step 607: After training for multiple cycles, if the Loss of the vector generation model no longer decreases, stop training to obtain the target vector generation model.

[0123] After stopping the training, the model parameters of the current model are saved to obtain the target vector generation model.

[0124] Step 608, input the resume node, position node and function node in the heterogeneous graph into the target vector generation model to obtain the vector information corresponding to the resume node, position node and function node respectively.

[0125] Step 609, the vector information corresponding to the resume node, position node and function node is respectively used to determine the functions similar to the resume using the Annoy search and matching algorithm.

[0126] In this step, optionally, nodes in the heterogeneous graph with node degrees greater than or equal to 10 may be retained according to the node degrees. That is, nodes in the heterogeneous graph with node degrees less than 10 are filtered out according to the node degrees, wherein the filtered nodes are untrusted nodes.

[0127] Next, the Annoy vector retrieval index of the function node is constructed. For the vector information of each confidence resume node, the Annoy cosine distance metric is used to obtain the similarity value between the confidence resume node and each function.

[0128] According to the similarity value between the confident resume node and each function, functions similar to the resume can be found. Optionally, for each resume corresponding to the confident resume node, 10 functions with the largest similarity values ​​can be selected as functions similar to the resume.

[0129] Step 610, using correlation filtering and similarity truncation processing methods to generate functions that match the resume.

[0130] In this step, optionally, for each resume corresponding to a trusted resume node, the positions that the job seeker has viewed and applied for based on the resume can be found, and the first functional information set can be obtained by obtaining the functional information corresponding to these positions. After obtaining the first functional information set, the similarity of each first functional information to each functional information in the heterogeneous graph can be calculated based on the vector information of each first functional information, that is, the second similarity value. After that, the second functional information set corresponding to the trusted resume node is obtained by obtaining the functional information corresponding to the largest M second similarity values ​​corresponding to each first functional information. Based on this, a node of functional information that the job seeker may be interested in can be obtained. For example, for each first functional information, the functional information corresponding to the first 30 second similarity values ​​can be selected as the recalled functional information, and the recalled functional information is added to the second functional information set for filtering the functions similar to the resume determined based on step 609.

[0131] After the filtering is completed, according to the similarity score, for each resume corresponding to the confident resume node, the remaining matching functions are cut off by threshold to improve the relevance between the functions and the resume.

[0132] Optionally, in order to ensure that each resume corresponding to a trusted resume node has matching functional information, after the threshold stage, if the number of functions matched by the resume is less than a preset number, the condition that the number of functional information matched by the resume reaches a preset number can be preferentially satisfied. Optionally, the preset number can be 3.

[0133] According to the embodiment of the present application, by establishing a heterogeneous graph including resume nodes, position nodes and function nodes, the complex relationship between nodes of various preset types is conveniently represented, so as to find the function information with a higher degree of matching with the resume. Next, since the target vector generation model is obtained by training the training samples generated based on the heterogeneous graph, it is conducive to fully exploring the job-seeking intention of the resume corresponding to the resume node, and improving the accuracy and reliability of the vector information output by the model. Afterwards, the vector information of the resume node and the vector information of the function node are combined to find the confident resume node, wherein, since the confident resume node has more connection relationships with other nodes in the heterogeneous graph, it is conducive to improving the accuracy of the matching result of the function information for each confident resume node, and finally, according to the top N confident function nodes with the highest similarity between the recruitment information and each of the confident resume nodes, the function information matching the resume corresponding to each confident resume node is accurately determined, thereby helping to reduce the gap between the matching position provided by the recruitment software and the job-seeking intention of the job seeker, and improve the job seeker's satisfaction with the recruitment software.

[0134] Corresponding to the method embodiment of the present application, the present application also provides a resume processing device, such as Figure 7 As shown, the resume processing device 700 includes: an acquisition module 701 and a processing module 702.

[0135] The acquisition module 701 is used to obtain job-seeking behavior information within a preset time period and recruitment information released within a preset time period, and establish a heterogeneous graph based on preset types of nodes. The preset types of nodes include resume nodes, position nodes and function nodes. The recruitment information includes function information that corresponds one to one with the position.

[0136] Processing module 702 is used to determine the node vector information of the heterogeneous graph according to the heterogeneous graph and the target vector generation model, the node vector information includes the vector information of the resume node and the vector information of the function node, and the target vector generation model is trained based on the training sample set generated by the heterogeneous graph.

[0137] The processing module 702 is also used to determine multiple trusted resume nodes and the top N trusted functional nodes with the highest similarity to each trusted resume node based on the vector information of the resume node and the vector information of the functional node.

[0138] The processing module 702 is also used to determine the function information that matches the resume corresponding to each confident resume node based on the top N confident function nodes with the highest similarity between the recruitment information and each confident resume node.

[0139] In some embodiments of the present application, optionally, the job-seeking behavior information includes job-viewing behavior and resume-submitting behavior, and the recruitment information includes a job identification corresponding to each functional information.

[0140] The processing module 702 is also used to determine the relationship information between each resume node and the position node according to the position viewing behavior and the resume submission behavior.

[0141] The processing module 702 is further used to determine the relationship information between the position node and the function node according to each position identifier and the position description corresponding to each position identifier.

[0142] The processing module 702 is further used to generate a heterogeneous graph according to the relationship information between each resume node and the position node, and the relationship information between each position node and the function node.

[0143] In some embodiments of the present application, optionally, the processing module 702 is further used to generate a training sample set according to the heterogeneous graph, wherein the training sample set includes a walk sequence determined by walking the heterogeneous graph according to different meta-paths.

[0144] The processing module 702 is further used to train the pre-constructed vector generation model according to the training sample set, and obtain the target vector generation model after the preset training conditions are met.

[0145] The acquisition module 701 is also used to acquire the node vector information of the heterogeneous graph through the target vector generation model, and the node vector information includes the vector information of the resume node and the vector information of the function node.

[0146] In some embodiments of the present application, optionally, the processing module 702 is further configured to walk the heterogeneous graph according to a preset meta-path to obtain a walking sequence set, where the walking sequence set includes a plurality of walking sequences corresponding to each preset meta-path.

[0147] The processing module 702 is further configured to generate a training sample set according to the order of nodes in each walking sequence in the walking sequence set and K nodes in the neighborhood of each node, where K is a positive integer.

[0148] In some embodiments of the present application, optionally, the processing module 702 is further configured to train the sample set to include a plurality of positive training samples and a plurality of negative training samples.

[0149] The processing module 702 is further used to generate a training sample set according to the order of nodes in each walking sequence and the K nodes in the neighborhood of each node, including:

[0150] The processing module 702 is further used to combine the current node with each of the K nodes in the neighborhood of the current node for each walking sequence to obtain a positive training sample, the current node is any node in the walking sequence, the current node in each positive training sample is an input sample node, and the other node is an output positive sample node. And, the processing module 702 is also used to randomly obtain L nodes as output negative sample nodes from the multiple walking sequences included in the first walking sequence set for the input sample node in each positive training sample, and respectively combine the L output negative sample nodes with the input sample nodes to obtain L negative training samples corresponding to each positive training sample.

[0151] In some embodiments of the present application, optionally, the processing module 702 is further used to determine the node degree of each vector node and the node degree of each functional node according to the heterogeneous graph.

[0152] The processing module 702 is also used to obtain the vector information of the first resume node and the vector information of the first functional node in the node vector information according to the node degree of each vector node and the node degree of each functional node, wherein the node degree corresponding to the first resume node is greater than the first threshold, and the node degree corresponding to the vector information of the first functional node is greater than the second threshold.

[0153] The processing module 702 is further used to determine a first similarity value between each first resume node and each first function node according to the vector information of the first resume node and the vector information of the first function node.

[0154] The processing module 702 is also used to determine each first resume node as a trusted resume node according to the first similarity value between each first resume node and each first function node, and to determine the first function nodes corresponding to the largest first N first similarity values ​​as trusted function nodes.

[0155] In some embodiments of the present application, optionally, the processing module 702 is also used to determine a target confident functional node among N confident functional nodes for each confident resume node, and the first similarity value corresponding to the target confident functional node is greater than a third threshold.

[0156] The processing module 702 is further used to obtain the function information corresponding to each target trusted function node in the recruitment information.

[0157] The processing module 702 is also used to determine, for each trusted resume node, the function information corresponding to the target trusted function node as the function information that matches the resume corresponding to the trusted resume node.

[0158] In some embodiments of the present application, optionally, the processing module 702 is also used to determine, for each resume corresponding to the trusted resume node, a first functional information set corresponding to the trusted resume node based on job-seeking behavior information within a preset time period, the first functional information set including multiple first functional information, wherein the first functional information is the functional information viewed by the job seeker corresponding to the trusted resume node or the functional information corresponding to the position applied for by the job seeker.

[0159] The processing module 702 is further used to determine, for each trusted resume node, a second similarity value between each first function information in the corresponding first function information set and each first function node in the heterogeneous graph.

[0160] The acquisition module 701 is also used to obtain, for each trusted resume node, the functional information corresponding to the maximum M second similarity values ​​corresponding to each first functional information in the first functional information set, to obtain the second functional information set, where M is a positive integer greater than N.

[0161] The processing module 702 is also used to determine Y trusted functional nodes for each trusted resume node according to a preset node screening condition and a second functional information set, where Y is a positive integer less than or equal to N. The preset node screening condition includes: when the functional information corresponding to the trusted functional node is not included in the second functional information set, the trusted functional node is screened out.

[0162] The processing module 702 is further configured to determine a target trusted function node among the Y trusted function nodes.

[0163] It can be understood that the resume processing device of the embodiment of the present application can correspond to the execution entity of the resume processing method provided in the embodiment of the present application. The specific details of the operation and / or function of each module / unit of the resume processing device can be found in the description of the corresponding parts of the resume processing method provided in the above-mentioned embodiment of the present application. For the sake of brevity, they will not be repeated here.

[0164] The electronic device in the embodiment of the present application may be a user terminal device, a server, other computing devices, or a cloud server. Figure 8 A schematic diagram of the hardware structure of an electronic device according to an embodiment of the present application is shown. The electronic device may include a processor 801 and a memory 802 storing computer program instructions. When the processor 801 executes the computer program instructions, the process or function of any of the above-mentioned embodiments of the method is implemented.

[0165] Specifically, the processor 801 may include a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application. The memory 802 may include a large-capacity memory for data or instructions. For example, the memory 802 may be at least one of the following: a hard disk drive (HDD), a read-only memory (ROM), a random access memory (RAM), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a tape, a universal serial bus (USB) drive, or other physical / tangible memory storage devices. For another example, the memory 802 may include a removable or non-removable (or fixed) medium. For another example, the memory 802 may be inside or outside the integrated gateway disaster recovery device. The memory 802 may be a non-volatile solid-state memory. In other words, the memory 802 generally includes a tangible (non-transitory) computer-readable storage medium (such as a memory device) encoded with computer-executable instructions, and when the software is executed (such as by one or more processors), the operations described in the method of the embodiments of the present application may be performed. The processor 801 implements the process or function of any method in the above embodiments by reading and executing the computer program instructions stored in the memory 802.

[0166] In one example, Figure 8 The electronic device shown may also include a communication interface 803 and a bus 810. Among them, the processor 801, the memory 802, and the communication interface 803 are connected through the bus 810 and complete the communication between each other. The communication interface 803 is mainly used to realize the communication between each module, device, unit and / or device in the embodiment of the present application. The bus 810 includes hardware, software or both, and can couple the components of the online data traffic billing device to each other. For example, the bus may include at least one of the following: an accelerated graphics port (AGP) or other graphics bus, an enhanced industrial standard architecture (EISA) bus, a front-end bus (FSB), a hypertransport (HT) interconnect, an industrial standard architecture (ISA) bus, an infinite bandwidth interconnect, a low pin count (LPC) bus, a memory bus, a micro channel architecture (MCA) bus, a peripheral component interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a serial advanced technology attachment (SATA) bus, a video electronics standard association local (VLB) bus or other suitable buses. The bus 810 may include one or more buses. Although the embodiments of the present application describe or illustrate a specific bus, the embodiments of the present application may consider any suitable bus or interconnection method.

[0167] In combination with the method in the above embodiments, an embodiment of the present application also provides a computer-readable storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the process or function of any method in the above embodiments is implemented.

[0168] In addition, an embodiment of the present application further provides a computer program product, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the process or function of any one of the methods in the above embodiments is implemented.

[0169] The above exemplarily describes the flowcharts and / or block diagrams of the methods, devices, systems and computer program products of the embodiments of the present application, and describes the relevant various aspects. It should be understood that each box or combination thereof in the flowchart and / or block diagram can be implemented by computer program instructions, or by dedicated hardware that performs a specified function or action, or by a combination of dedicated hardware and computer instructions. For example, these computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer or other programmable data processing device to form a machine that enables these instructions executed by such a processor to enable the implementation of the functions / actions specified in each box or combination thereof in the flowchart and / or block diagram. Such a processor can be a general-purpose processor, a special-purpose processor, a special application processor or a field programmable logic circuit.

[0170] The functional blocks shown in the structured block diagram of the embodiment of the present application can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application-specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc.; when implemented in software, it is a program or code segment used to perform the required task. The program or code segment can be stored in a memory, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0171] It should be noted that the present application is not limited to the specific configurations and processes described above or shown in the figures. The above is only a specific implementation mode of the present application. Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process of the described system, device, module or unit can refer to the corresponding process in the method embodiment without further description. It should be understood that the scope of protection of the present application is not limited to this. Any technician familiar with the technical field can think of various equivalent modifications or substitutions within the technical scope disclosed in the present application, and these modifications or substitutions should be included in the scope of protection of the present application.

Claims

1. A resume processing method, characterized in that: include: Obtaining job-seeking behavior information within a preset time period and recruitment information released within the preset time period, and establishing a heterogeneous graph, wherein the heterogeneous graph includes associations among resume nodes, position nodes, and function nodes, and the recruitment information includes function information corresponding to positions one by one; Determine vector information of each node in the heterogeneous graph according to the heterogeneous graph and a target vector generation model, wherein the target vector generation model is trained based on a training sample set generated by the heterogeneous graph; Determine multiple trusted resume nodes and top N trusted functional nodes with the highest similarity to each of the trusted resume nodes based on the vector information of the resume node and the vector information of the functional node; Based on the top N trusted function nodes with the highest similarity between the recruitment information and each trusted resume node, determine the function information that matches the resume corresponding to each trusted resume node.

2. The method according to claim 1, characterized in that The job-seeking behavior information includes the behavior of checking positions and submitting resumes, and the recruitment information includes the position identification corresponding to each functional information; The step of establishing a heterogeneous graph based on nodes of a preset type includes: Determine the relationship information between each resume node and the position node according to the job viewing behavior and the resume submission behavior; According to each position ID and the position description corresponding to each position ID, determine the relationship information between the position node and the function node; The heterogeneous graph is generated according to the relationship information between each resume node and the position node, and the relationship information between each position node and the function node.

3. The method according to claim 1, characterized in that The step of determining the node vector information of the heterogeneous graph according to the heterogeneous graph and the target vector generation model includes: Generate a training sample set according to the heterogeneous graph, wherein the training sample set includes a walk sequence determined by walking the heterogeneous graph according to different meta-paths; The pre-constructed vector generation model is trained according to the training sample set, and after the preset training conditions are met, a target vector generation model is obtained; The target vector generation model is used to obtain node vector information of the heterogeneous graph, wherein the node vector information includes vector information of resume nodes and vector information of function nodes.

4. The method according to claim 3, characterized in that Generating a training sample set according to the heterogeneous graph includes: According to a preset meta-path, wandering the heterogeneous graph, obtaining a wandering sequence set, wherein the wandering sequence set includes a plurality of wandering sequences corresponding to each of the preset meta-paths; The training sample set is generated according to the order of nodes in each of the walking sequences in the walking sequence set and K nodes in the neighborhood of each node, where K is a positive integer.

5. The method according to claim 4, characterized in that The training sample set includes a plurality of positive training samples and a plurality of negative training samples; Generating the training sample set according to the order of nodes in each of the walking sequences and the K nodes in the neighborhood of each node includes: For each of the walking sequences, a current node is combined with each of the K nodes in the neighborhood of the current node to obtain a positive training sample, wherein the current node is any node in the walking sequence, and the current node in each of the positive training samples is an input sample node, and another node is an output positive sample node; as well as, For each input sample node in the positive training sample, L nodes are randomly obtained from multiple walking sequences included in the first walking sequence set as output negative sample nodes, and the L output negative sample nodes are respectively combined with the input sample nodes to obtain L negative training samples corresponding to each positive training sample.

6. The method according to claim 1, characterized in that Determining a plurality of trusted resume nodes and the top N trusted functional nodes with the highest similarity to each of the trusted resume nodes according to the vector information of the resume node and the vector information of the functional node includes: According to the heterogeneous graph, determining the node degree of each vector node and the node degree of each functional node; According to the node degree of each of the vector nodes and the node degree of each of the functional nodes, vector information of a first resume node and vector information of a first functional node are obtained in the node vector information, wherein the node degree corresponding to the first resume node is greater than a first threshold, and the node degree corresponding to the vector information of the first functional node is greater than a second threshold; Determine a first similarity value between each of the first resume nodes and each of the first function nodes according to the vector information of the first resume node and the vector information of the first function node; According to the first similarity value between each of the first resume nodes and each of the first function nodes, each of the first resume nodes is determined as the trusted resume node, and the first function nodes corresponding to the largest first N similarity values ​​are determined as trusted function nodes.

7. The method according to claim 6, characterized in that: Determining the function information matching the resume corresponding to each of the confident resume nodes according to the top N confident function nodes with the highest similarity between the recruitment information and each of the confident resume nodes includes: For each of the confident resume nodes, determine a target confident function node among the N confident function nodes, wherein the first similarity value corresponding to the target confident function node is greater than a third threshold; Acquire the function information corresponding to each of the target trusted function nodes in the recruitment information; For each of the trusted resume nodes, the functional information corresponding to the target trusted functional node is determined as functional information that matches the resume corresponding to the trusted resume node.

8. The method according to claim 7, characterized in that: Before determining the target trusted functional node among the N trusted functional nodes, the method further includes: For each resume corresponding to the trusted resume node, determine the first function information set corresponding to the trusted resume node according to the job-seeking behavior information within the preset time period, wherein the first function information set includes a plurality of first function information, wherein the first function information is the function information viewed by the job seeker corresponding to the trusted resume node or the function information corresponding to the position applied for by the job seeker; For each of the trusted resume nodes, determine a second similarity value between each first function information in the corresponding first function information set and each first function node in the heterogeneous graph; For each of the trusted resume nodes, obtain the functional information corresponding to the largest M second similarity values ​​corresponding to each first functional information in the first functional information set, to obtain a second functional information set, where M is a positive integer greater than N; For each of the trusted resume nodes, according to a preset node screening condition and the second function information set, Y trusted function nodes are determined, where Y is a positive integer less than or equal to N, and the preset node screening condition includes: if the function information corresponding to the trusted function node is not included in the second function information set, the trusted function node is screened out; The step of determining a target trusted function node among the N trusted function nodes comprises: A target trusted function node is determined among the Y trusted function nodes.

9. A resume processing device, characterized in that: include: An acquisition module, used to acquire job-seeking behavior information within a preset time period and recruitment information released within the preset time period, and to establish a heterogeneous graph based on nodes of preset types, wherein the nodes of the preset types include resume nodes, position nodes, and function nodes, and the recruitment information includes function information corresponding to each position one by one; A processing module, used to determine node vector information of the heterogeneous graph according to the heterogeneous graph and a target vector generation model, wherein the node vector information includes vector information of resume nodes and vector information of function nodes, and the target vector generation model is trained based on a training sample set generated by the heterogeneous graph; The processing module is further used to determine multiple trusted resume nodes and the top N trusted functional nodes with the highest similarity to each of the trusted resume nodes based on the vector information of the resume node and the vector information of the functional node; The processing module is also used to determine the function information that matches the resume corresponding to each of the trusted resume nodes based on the recruitment information and the top N trusted function nodes with the highest similarity to each of the trusted resume nodes.

10. An electronic device, characterized in that: The electronic device comprises: a processor and a memory storing computer program instructions; when the electronic device executes the computer program instructions, the method according to any one of claims 1 to 8 is implemented.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the method according to any one of claims 1 to 8 is implemented.

12. A computer program product, characterized in that It comprises computer program instructions, which, when executed by a processor, implement the method according to any one of claims 1 to 8.

Images