Label generation method, device, electronic device and storage medium

Abstract

The present application discloses a label generation method, device, electronic device and storage medium. The method includes: obtaining user information of a job seeker and job information of a job related to the job seeker, wherein the job related to the job seeker includes the job viewed by the job seeker and the job applied for by the job seeker; inputting the user information of the job seeker and the job information of the job related to the job seeker into a pre-trained target label generation model, and outputting the job seeker's demand label through the target label generation model; wherein the target label generation model includes an embedding layer, a feature recognition layer and an output layer; outputting the job seeker's demand label through the target label generation model, and outputting the demand intention corresponding to the demand label. The embodiment of the present application can flexibly determine labels for job seekers, which is conducive to improving the accuracy of subsequent matching positions.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a label generation 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 publish recruitment information in recruitment software and find the required job seekers in the recruitment software. Job seekers can publish resumes in recruitment software and find suitable positions in recruitment software. In order to better provide business services, recruitment software has introduced tags to provide personalized job matching for job seekers, so that job seekers can view more positions of interest when accessing recruitment software.

[0003] However, in the related art, job seekers' tags are usually obtained by extracting keywords from their resumes or are set by the users themselves. Therefore, job seekers' tags have low maintainability and are difficult to accurately describe their job search needs. Therefore, when matching positions for users based on these tags, the accuracy of the matching is directly affected, resulting in a poor user experience when using recruitment software. Summary of the invention

[0004] In view of this, the embodiments of the present application provide a tag generation method, device, electronic device and storage medium, which can flexibly determine tags for job seekers, which is conducive to improving the accuracy of subsequent job matching.

[0005] An embodiment of the present application provides a label generation method, including: obtaining user information of a job seeker and position information of positions related to the job seeker, wherein the positions related to the job seeker include positions for which the job seeker has viewed but not submitted a resume and positions for which the job seeker has viewed but has submitted a resume; inputting the user information of the job seeker and the position information of the positions related to the job seeker into a pre-trained target label generation model, and outputting the job seeker's demand label through the target label generation model; wherein the target label generation model includes an embedding layer, a feature recognition layer and an output layer; outputting the job seeker's demand label through the target label generation model, including: converting user information into a user feature vector and converting position information into a position feature vector through the embedding layer, determining a target feature that the job seeker is concerned about and a demand intention corresponding to the target feature based on the user feature vector and the position feature vector through the feature recognition layer, converting the target feature into a demand label through the output layer, outputting the demand label, and outputting the demand intention corresponding to the demand label.

[0006] Optionally, according to the method of an embodiment of the present application, the user feature vector includes a first feature vector and a second feature vector; converting user information into a user feature vector includes: identifying unstructured feature information in the user information, and structured feature information corresponding to a first preset attribute, wherein the user information includes at least any one of the following information: resume information, job communication information, job viewing behavior information, job submission behavior information, and the unstructured information is data without a predetermined format; determining the attribute value of the structured feature information corresponding to each first preset attribute to obtain a first feature vector; performing semantic understanding on the unstructured feature information, extracting text features, and converting the text features into a second feature vector.

[0007] Optionally, according to the method of an embodiment of the present application, converting the position information into a position feature vector includes: identifying structured feature information in the position information corresponding to a second preset attribute; determining the attribute value of the structured feature information corresponding to each second preset attribute, and obtaining the position feature vector.

[0008] Optionally, according to the method of the embodiment of the present application, the feature recognition layer includes: an attention layer and a fully connected layer, and the position feature vector includes a first position feature vector for viewing positions corresponding to unsubmitted resumes and a second position feature vector for viewing positions corresponding to submitted resumes; the target features that job seekers are concerned about and the demand intentions corresponding to the target features are determined through the feature recognition layer according to the user feature vector and the position feature vector, including: determining difference information through the attention layer according to the user feature vector, the first position feature vector and the second position feature vector, and generating a feature vector of the target feature based on the difference information; identifying the feature vector of the target feature through the fully connected layer, and determining the demand intention corresponding to the target feature.

[0009] Optionally, according to the method of the embodiment of the present application, the method also includes: in response to a job seeker visiting a recruitment platform, obtaining the job seeker's demand tag information, wherein the demand tag information includes the job seeker's demand tag and demand intention; adjusting the recall weight corresponding to the demand tag information in the position recall model, and determining the first matching position for the demand tag information based on the adjusted position recall model; and displaying the first matching position to the job seeker.

[0010] Optionally, according to the method of the embodiment of the present application, the label information also includes custom label information of the job seeker; according to the adjusted position recall model, the second matching position of the custom label information is determined; the first matching position is displayed to the job seeker, including: inputting the first matching position, the requirement label information, the second matching position and the custom label information into the sorting model, and through the sorting model, outputting the display order corresponding to the first matching position and the second matching position respectively, so that the priority display probability of the first matching position is greater than the priority display probability of the second matching position; based on the display order, the first matching position is displayed to the job seeker, and based on the display order, the second matching position is displayed to the job seeker.

[0011] Optionally, according to the method of an embodiment of the present application, before obtaining user information of a job seeker and position information of a position associated with the job seeker, the method includes: obtaining historical job search behavior information within a preset time period, the historical job search behavior information including multiple historical job seekers and associated positions of each historical job seeker; obtaining historical user information of each historical job seeker, historical position information of positions associated with the historical job seeker, and preset user tags of each historical job seeker; generating a training sample set based on the historical user information of each historical job seeker, historical position information of positions associated with the historical job seeker, and preset user tags of each historical job seeker; and training a pre-constructed label generation model based on the training sample set until preset training conditions are met to obtain a target label generation model, wherein the label generation model includes an embedding layer, a feature recognition layer, and an output layer.

[0012] An embodiment of the present application provides a label generation device, including: an acquisition module, used to acquire user information of a job seeker and position information of positions related to the job seeker, wherein the positions related to the job seeker include positions for which the job seeker has viewed but not submitted a resume and positions for which the job seeker has viewed but has submitted a resume; a processing module, used to input the user information of the job seeker and the position information of the positions related to the job seeker into a pre-trained target label generation model, and output the job seeker's demand label through the target label generation model; wherein the target label generation model includes an embedding layer, a feature recognition layer and an output layer; wherein the processing module outputs the job seeker's demand label through the target label generation model, including: converting the user information into a user feature vector through the embedding layer, and converting the position information into a position feature vector, determining the target feature that the job seeker is concerned about and the demand intention corresponding to the target feature according to the user feature vector and the position feature vector through the feature recognition layer, converting the target feature into a demand label through the output layer, outputting the demand label, and outputting the demand intention corresponding to the demand label.

[0013] 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.

[0014] 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.

[0015] 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.

[0016] According to the embodiment of the present application, for each job seeker, the user information of the job seeker and the position information of the job seeker's related positions can be obtained. The related positions, for example, the job seeker's related positions include the positions viewed by the job seeker and the positions submitted by the job seeker. By inputting the user information and the position information of the related positions into the pre-trained target label generation model, the user information is converted into a user feature vector through the embedding layer in the target label generation model, and the position information is converted into a position feature vector, so as to analyze the user information and the position information and the position information, and extract useful information from them. Next, the key information affecting the job seeker's job-seeking intention is found according to the user feature vector and the position feature vector through the feature recognition layer, that is, the target feature of the job seeker and its corresponding demand intention. Finally, the target feature is converted into a demand label through the output layer, and the demand label and demand intention are output. Based on this, by analyzing the user information of the job seeker and the position information of the related positions, a demand label that is more in line with the job seeker's job-seeking needs is automatically generated, which is conducive to improving the matching efficiency between the job seeker and the position, reducing the time cost of the job seeker to screen the position, and improving the user's job-seeking experience. For the recruiter, it can also find suitable candidates faster, reduce recruitment costs, and improve recruitment efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] 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.

[0018] Figure 1 It is a schematic diagram of the system architecture of an embodiment of the present application.

[0019] Figure 2 It is a flowchart of the label generation method of an embodiment of the present application.

[0020] Figure 3 It is a flowchart of another label generation method according to an embodiment of the present application.

[0021] Figure 4 It is a structural block diagram of a label generation model provided in an embodiment of the present application.

[0022] Figure 5 It is a structural block diagram of the label generation device of an embodiment of the present application.

[0023] Figure 6 It is a schematic diagram of an electronic device used to implement the label generation method of an embodiment of the present application. DETAILED DESCRIPTION

[0024] 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.

[0025] 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).

[0026] 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.

[0027] 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 in 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 label generation method, apparatus, electronic device, computer-readable storage medium, and computer program product. Figure 1 A schematic diagram of a system architecture of an embodiment of the present application is shown. Figure 1As 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. For example, a job seeker may operate on the terminal device 102a. For example, the job seeker may open the terminal device 102, use the recruitment software, and perform operations such as viewing positions, posting resumes, and submitting resumes in the recruitment software. Optionally, the recruiter may also operate on another terminal device 102b, such as posting positions and position-related introductions in the recruitment software.

[0028] In an embodiment of the present application, the executor of the label generation 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.

[0029] As a specific example, after the terminal device 102a receives the instruction input by the job seeker, the request information including the instruction is sent 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 102a. The user instruction is completed through a series of data processing and information interaction. Based on these instructions, the server 104 can obtain job search behavior information. Based on the same logic as the job seeker using the terminal device 102a, the recruiter uses the terminal device 102b, which can also enable the server 104 to obtain job information. Based on these instructions, the server 104 can obtain but is not limited to the user information, resume, some job search behavior information and recruitment information of the job seeker during the job search process. Based on this, the server 104 can regularly obtain this information and analyze it to determine a label with higher accuracy and reliability for the job seeker, so as to improve the convenience of the job seeker using the recruitment software, and at the same time, it is also beneficial for the recruiter to find suitable employees.

[0030] Figure 2 A schematic diagram showing a flow chart of a label generation method according to an embodiment of the present application is shown. Figure 2 As shown, the tag generation method may include the following steps 201 to 202.

[0031] Step 201, obtaining user information of a job seeker and position information of positions related to the job seeker, wherein the positions related to the job seeker include positions for which the job seeker has viewed but not submitted resumes and positions for which the job seeker has viewed but submitted resumes.

[0032] Step 202 , input the user information of the job seeker and the position information of the positions related to the job seeker into a pre-trained target label generation model, and output the job seeker's requirement label through the target label generation model.

[0033] Among them, the target label generation model includes an embedding layer, a feature recognition layer and an output layer.

[0034] Figure 3 is a flow chart of another label generation method provided in an embodiment of the present application, and further combined with Figure 3 As shown, the target tag generation model is used to output the job seeker's requirement tags, and details may refer to steps 202a to 202c.

[0035] Step 202a, converting the user information into a user feature vector and converting the position information into a position feature vector through an embedding layer.

[0036] Step 202b, determining the target features that the job seeker is concerned about and the demand intentions corresponding to the target features through the feature recognition layer according to the user feature vector and the position feature vector.

[0037] Step 202c, convert the target features into demand labels through the output layer, output the demand labels, and output the demand intentions corresponding to the demand labels.

[0038] The above steps are described in detail below in conjunction with specific embodiments, as shown below.

[0039] In step 201, the preset time period may be the time interval between each execution of the tag generation method, or may be a required time length set according to application requirements. By timely analyzing user information and job information of job seekers' related positions, the reliability of job seekers' tags is improved, and reliable recommended positions are provided for job seekers.

[0040] In the embodiment of the present application, the user information of the job seeker may include the unique identification of the job seeker, and may also include some basic attribute information of the job seeker, and include the communication information between the user and the recruiter during the process of using the recruitment software. Among them, the basic attribute information may include but is not limited to any of the following information: place of residence, education, age, gender, skills, certificates, expected salary, expected working city, expected industry, expected company nature, expected company size, etc. This information may be provided by the job seeker when registering an account, or may be obtained from the user's resume.

[0041] The job seeker's related positions may include positions that the job seeker has viewed but not submitted resumes to and positions that the job seeker has viewed but has submitted resumes to. Optionally, the job seeker's related positions may be obtained from a service device of the recruitment software. For example, the job seeker's actions of viewing positions and submitting resumes in the recruitment software may be recorded in the service device.

[0042] Exemplarily, job information may include an introduction to the job, requirements for job seekers, etc. Job information includes but is not limited to any of the following information: job title, salary range, specific work location, working hours, nature of the company, company size, company benefits, skill requirements, job description, etc.

[0043] After obtaining the user information of the job seeker and the position information of the relevant positions, the next step involves step 202, in which the user information of the job seeker and the position information of the relevant positions of the job seeker are input into a pre-trained target label generation model, and the target label generation model is used to output the job seeker's demand label.

[0044] Specifically, the target label generation model may include an embedding layer, a feature recognition layer, and an input layer. Figure 3 As shown, the user information is converted into a user feature vector, and the position information is converted into a position feature vector through the embedding layer.

[0045] Exemplarily, the basic attribute information in the user information and the communication information between the user and the recruiter during the use of the recruitment software can be used to extract the user's features in multiple dimensions and convert the user's features into a user feature vector. Position information can include information such as an introduction to the position and requirements for job seekers, from which position features can be extracted and converted into a position feature vector. Through the embedding layer, for each job seeker, the semantic relationship between each user feature and each position feature can be extracted, and both the user feature vector and the position feature vector can be converted into a vector represented by a numerical value, which is conducive to finding the job seeker's career interests and job search preferences.

[0046] The user feature vector and position feature vector of each job seeker output by the embedding layer can be further processed by the feature recognition layer. Optionally, the feature recognition layer can use the dot product attention mechanism to find the feature vectors that the job seeker pays more attention to among all the user feature vectors and position feature vectors corresponding to the job seeker, and then determine the feature vectors corresponding to the viewed positions and the feature vectors corresponding to the positions applied for among the feature vectors that the job seeker pays more attention to, and then analyze the difference between the feature vectors corresponding to the viewed positions and the feature vectors corresponding to the applied positions to find the target features that the job seeker pays more attention to.

[0047] By analyzing the feature vectors corresponding to the positions of the job seeker's "view submitted" and "view unsubmitted", the difference between two similar jobs is found, which is the target feature that the job seeker is concerned about. Optionally, when the difference between the features of the two jobs is found, the difference calculation, similarity calculation, etc. can be combined for calculation.

[0048] The feature recognition layer can extract key information from the user feature vector and the position feature vector, identify key features that affect the user's job search, and help to more accurately understand the user's job search intentions, bringing convenience to job seekers.

[0049] Finally, the target features are converted into demand labels through the output layer, and the demand labels are output. Based on this, by analyzing the user information of job seekers and the position information of related positions, demand labels that are more in line with the job search needs of job seekers are automatically generated, which is conducive to improving the matching efficiency between job seekers and positions, reducing the time cost of job seekers to screen positions, and improving the user's job search experience. For recruiters, it can also find suitable candidates faster, reduce recruitment costs, and improve recruitment efficiency.

[0050] In some embodiments of the present application, a user feature vector includes a first feature vector and a second feature vector; specifically, converting user information into a user feature vector may include the following steps: identifying unstructured feature information in the user information, and structured feature information corresponding to a first preset attribute, wherein the user information includes at least any one of the following information: resume information, job communication information, job viewing behavior information, and job submission behavior information, and the unstructured information is data without a predetermined format; determining the attribute value of the structured feature information corresponding to each first preset attribute to obtain a first feature vector; performing semantic understanding on the unstructured feature information, extracting text features, and converting the text features into a second feature vector.

[0051] Exemplarily, the structured feature information in the user information is information corresponding to the first preset attribute, which is data in a predetermined format. The information corresponding to the first preset attribute can be directly converted into a digital vector, that is, a first feature vector. The first preset attribute is, for example, a specific place of residence, education, age, gender, certificate, expected salary, expected working city, expected industry, expected company nature, expected company size, etc., which are not listed here one by one.

[0052] User information may also include resume information, job communication information, job viewing behavior information, and job submission behavior information, including unstructured information, that is, data without a predetermined format. Unstructured information includes information involved in the process of job seekers using recruitment software, specifically, job seekers' self-evaluation in resume information, job communication information such as chat records during the communication process between job seekers and HR, job viewing behavior information and job submission behavior information. The length of time that job seekers stay on various interfaces when browsing jobs, etc., are not listed here one by one.

[0053] Optionally, a pre-trained model such as BERT (Bidirectional Encoder Representations from Transformers) is embedded in the embedding layer, so that the embedding layer can extract useful information from complex language patterns and relationships in unstructured feature information.

[0054] According to the embodiment of the present application, the embedding layer can convert user information into a digital vector that is convenient for the feature recognition layer to process. In particular, the unstructured information may include a large amount of useful information. The embedding layer can effectively extract the unstructured information and convert the useful information into a feature vector, that is, the second feature vector. Based on this, the feature recognition layer can analyze the user information of the job seeker and the job information of the relevant position, and automatically generate a demand label that better meets the job search needs of the job seeker.

[0055] In some embodiments of the present application, the position information is converted into a position feature vector, and specifically the following steps can be referred to: identifying structured feature information in the position information corresponding to a second preset attribute; determining the attribute value of each structured feature information corresponding to the second preset attribute, and obtaining the position feature vector.

[0056] Exemplarily, the structured feature information in the job information is information corresponding to the first preset attribute, which is data in a predetermined format. The information corresponding to the first preset attribute can be directly converted into a digital vector, that is, a first feature vector. The first preset attribute may include specific residence, education, age, gender, certificate, expected salary, expected working city, expected industry, expected company nature, expected company size, etc., which are not listed here one by one.

[0057] According to the embodiment of the present application, the embedding layer can convert user information into a digital vector that is convenient for the feature recognition layer to process, that is, a position feature vector. Based on this, the feature recognition layer can analyze the user information of the job seeker and the position information of the relevant position, and automatically generate a demand label that better meets the job seeker's job search needs.

[0058] In the embodiment of the present application, the feature recognition layer may specifically include an attention layer and a fully connected layer. For example, Figure 4 It is a structural diagram of a target label generation model provided in an embodiment of the present application.

[0059] In the embodiment of the present application, the relevant positions of the job seeker include the positions that the job seeker has viewed but not submitted resumes and the positions that the job seeker has viewed and submitted resumes. Accordingly, the position feature vector may include a first position feature vector corresponding to the position that has not submitted resumes and a second position feature vector corresponding to the position that has submitted resumes.

[0060] Based on the target label generation model provided in the embodiment of the present application, the target features that job seekers are concerned about are determined through the feature recognition layer according to the user feature vector and the position feature vector. Specifically, the following steps 301 and 302 may be included.

[0061] Step 301, determining difference information according to the user feature vector, the first position feature vector and the second position feature vector through an attention layer, and generating a feature vector of the target feature according to the difference information.

[0062] Step 302: Identify the feature vector of the target feature through a fully connected layer to determine the demand intention corresponding to the target feature.

[0063] Exemplarily, the feature recognition layer includes an attention layer. In the attention layer, the dot product attention mechanism can be used to determine difference information in the job feature vector of the job seeker, and then combined with the user feature vector to find a feature vector that the job seeker pays more attention to.

[0064] The attention layer combines the global information of the input data to find features that may arouse the user's interest, where the global information is the user feature vector, the first position feature vector, and the second position feature vector. In one example, the attention layer can analyze the salary difference and overtime difference between the positions for which resumes have been submitted and the positions for which resumes have not been submitted, while there is no difference in company type and commuting distance.

[0065] For example, the salary level of the positions for which resumes have been submitted is higher than that of the positions for which resumes have not been submitted, so it is determined that the user may pay more attention to the feature of salary; for another example, the positions for which resumes have been submitted have information about no overtime, while the positions for which resumes have not been submitted have information about needing overtime, so it is determined that the user may pay more attention to the feature of whether to work overtime, while there is no difference in the type of enterprise and the commuting distance, so it is determined that the user does not pay much attention to the two features of type of enterprise and commuting distance. Therefore, the attention layer can determine the features that the user pays more attention to, that is, the target features.

[0066] The feature vector of the target feature output by the attention layer can be a high-dimensional vector, which can include information about the target feature itself and information about the contextual features of the target feature.

[0067] The feature vector of the target feature can represent the distribution of job seekers’ focus in the input content. For example, “salary” is given a high weight in both undelivered and delivered positions, and “company type” attracts attention in some undelivered positions.

[0068] Based on the feature vector of the target feature output by the attention layer, these features are further integrated to clarify the target features of job seekers and the demand intentions that job seekers are concerned about corresponding to the target features. Demand intentions include "need high salary", "no overtime", "no need to be close to home", etc.

[0069] Based on the embodiments of the present application, the feature recognition layer can find key information that affects the job-seeking intention of job seekers based on user feature vectors and position feature vectors, that is, the target features of job seekers. This can improve the degree of fit between job seekers and labels determined for them, which is conducive to subsequently improving the accuracy of matching between job seekers and positions, thereby improving user satisfaction with the matching results.

[0070] As a specific example, the user information of job seeker A includes the following: when using the recruitment software, "senior Java engineer" is used as the search keyword. The user information of job seeker A may also include the following information: education: master's degree, residence: E Road, D District, Shanghai, age: 35, gender: male, expected work city: Shanghai, expected salary: 30k-50k, skills: solr, elasticsearch, lucene, java, hadoop, kafka, redis, MongoDB. The behavior information of job seeker A may include the behavior: viewed and submitted position 1, viewed unsubmitted position 2.

[0071] The features that job seekers pay more attention to in Position 1 include: Position title: Senior Java Engineer, Salary 28K, Work Location: Road B, District A, Shanghai, Five social insurances and one housing fund, Skills required to be familiar with elasticsearch, lucene, java, Company nature: State-owned enterprise, Company scale 500-1000 people; The features that job seekers pay more attention to in Position 2 include: Information of Position 2: Position title: Java back-end engineer, Salary 35K, Work Location: Near Road D subway station, District C, Shanghai, Five social insurances and one housing fund, Skills required to be familiar with elasticsearch, lucene, springboot, mybatis, Company nature: Private enterprise, Company scale 500-1000 people.

[0072] Through the attention layer and the fully connected layer included in the feature recognition layer, the feature vector that affects the job seeker's decision can be found among all the user feature vectors and position feature vectors corresponding to the job seeker. For example, by extracting the position information of the user's "view submitted resume" and "view unsubmitted resume", the difference between the two position feature vectors is obtained, and the feature vector that affects the job seeker's decision is found. Among them, the features corresponding to the feature vector of the job seeker's decision are also the target features that the job seeker pays attention to.

[0073] It can be understood that the above features are all digital vectors when the feature recognition layer performs data processing, and the above text description is only for ease of understanding.

[0074] According to the embodiment of the present application, after the feature recognition layer finds the target feature that the job seeker is concerned about, the output layer can then convert the target feature into a demand tag, output the demand tag, and output the demand intention corresponding to the demand tag. Based on this, by analyzing the user information of the job seeker and the position information of the relevant position, a demand tag that better meets the job seeker's job search needs is automatically generated.

[0075] In some embodiments of the present application, the target label generation model is trained based on the historical job-seeking behavior information collected by the recruitment software within a preset time period. Specifically, before obtaining the user information of the job seeker and the position information of the position associated with the job seeker, the target label generation model can be obtained based on the training method provided in the following steps 401 to 404.

[0076] Step 401 : obtaining historical job-seeking behavior information within a preset time period, wherein the historical job-seeking behavior information includes a plurality of historical job-seeking users and associated positions of each historical job-seeking user.

[0077] Step 402 , obtaining the historical user information of each historical job-seeking user, the historical position information of the positions associated with the historical job-seeking user, and the preset user tag of each historical job-seeking user.

[0078] Step 403 : Generate a training sample set based on the historical user information of each historical job-seeking user, the historical position information of the positions associated with the historical job-seeking user, and the preset user label of each historical job-seeking user.

[0079] Step 404, based on the training sample set, the pre-built label generation model is trained until the preset training conditions are met to obtain a target label generation model, wherein the label generation model includes an embedding layer, a feature recognition layer and an output layer.

[0080] Optionally, during the training process, a cross entropy loss function can be used to optimize the model parameters.

[0081] According to the embodiment of the present application, job seekers are labeled through a trained model to obtain their personal preferences. In the recall and partial order stages, job information that meets the focus is prioritized to improve the application-to-distribution ratio and user experience.

[0082] In some embodiments of the present application, based on the job seeker's demand tags determined in the embodiments of the present application, the recall strategy can be adjusted based on the demand tags. For example, in response to the job seeker visiting a recruitment platform, the job seeker's demand tag information is obtained, wherein the demand tag information includes the job seeker's demand tags; the recall weight corresponding to the demand tag information in the position recall model is adjusted, and the first matching position for the demand tag information is determined based on the adjusted position recall model; and the first matching position is displayed to the job seeker.

[0083] Based on the job-seeking user's requirement tags determined in the embodiments of the present application, for the recalled first matching position, the display order of the first matching position among all matching positions can be adjusted based on the requirement tags.

[0084] In some embodiments, the tag information also includes custom tag information that can be used by job seekers, such as being close to home, not working overtime, etc. Based on this, when adjusting the display order of the first matching position among all matching positions, the following steps can be specifically referred to: determining the second matching position of the custom tag information according to the adjusted position recall model; displaying the first matching position to the job seeker, including: inputting the first matching position, the demand tag information, the second matching position and the custom tag information into the sorting model, and outputting the display order corresponding to the first matching position and the second matching position respectively through the sorting model, so that the priority display probability of the first matching position is greater than the priority display probability of the second matching position; displaying the first matching position to the job seeker based on the display order, and displaying the second matching position to the job seeker based on the display order.

[0085] According to the embodiments of the present application, by analyzing the user information of job seekers and the position information of related positions, demand tags that are more in line with the job search needs of job seekers are automatically generated, which is conducive to subsequently improving the matching efficiency between job seekers and positions, reducing the time cost of job seekers in screening positions, and improving the user's job search experience. For recruiters, it is also possible to find suitable candidates more quickly, reduce recruitment costs, and improve recruitment efficiency.

[0086] The above describes the implementation methods and advantages of the embodiments of the present application through multiple embodiments. The following describes the specific processing process of the embodiments of the present application in detail with reference to specific examples.

[0087] Corresponding to the method embodiment of the present application, the present application also provides a label generation device, such as Figure 5As shown, the label generating device 500 includes an acquisition module 501 and a processing module 502 .

[0088] The acquisition module 501 is used to acquire the user information of the job seeker and the position information of the positions related to the job seeker, wherein the positions related to the job seeker include the positions that the job seeker has viewed but not submitted resumes to and the positions that the job seeker has viewed but has submitted resumes to.

[0089] The processing module 502 is used to input the user information of the job seeker and the position information of the job seeker's related positions into a pre-trained target label generation model, and output the job seeker's requirement label through the target label generation model.

[0090] Among them, the target label generation model includes an embedding layer, a feature recognition layer and an output layer.

[0091] Among them, the processing module 502 outputs the job seeker's demand label through the target label generation model, including: converting user information into a user feature vector through the embedding layer, and converting position information into a position feature vector, determining the target features that the job seeker is concerned about and the demand intentions corresponding to the target features based on the user feature vector and the position feature vector through the feature recognition layer, converting the target features into demand labels through the output layer, and outputting the demand labels, and outputting the demand intentions corresponding to the demand labels.

[0092] In some embodiments of the present application, optionally, the user feature vector includes a first feature vector and a second feature vector.

[0093] The processing module 502 is also used to identify unstructured feature information in the user information and structured feature information corresponding to the first preset attribute, wherein the user information includes at least any one of the following information: resume information, job communication information, job viewing behavior information, and job submission behavior information, and the unstructured information is data without a predetermined format.

[0094] The processing module 502 is further used to determine the attribute value of the structured feature information corresponding to each first preset attribute to obtain a first feature vector.

[0095] The processing module 502 is further used to perform semantic understanding on the unstructured feature information, extract text features, and convert the text features into a second feature vector.

[0096] In some embodiments of the present application, optionally, the processing module 502 is further configured to identify structured feature information corresponding to the second preset attribute in the position information.

[0097] The processing module 502 is further used to determine the attribute value of the structured feature information corresponding to each second preset attribute to obtain a position feature vector.

[0098] In some embodiments of the present application, optionally, the feature recognition layer includes: an attention layer and a fully connected layer, and the position feature vector includes a first position feature vector for viewing positions corresponding to unsubmitted resumes and a second position feature vector for viewing positions corresponding to submitted resumes.

[0099] The processing module 502 is also used to determine difference information according to the user feature vector, the first position feature vector and the second position feature vector through the attention layer, and generate a feature vector of the target feature according to the difference information.

[0100] The processing module 502 is further used to identify the feature vector of the target feature through the fully connected layer, and determine the demand intention corresponding to the target feature.

[0101] In some embodiments of the present application, optionally, the acquisition module 501 is further used to obtain the job seeker's demand tag information in response to the job seeker visiting the recruitment platform, wherein the demand tag information includes the job seeker's demand tag and demand intention.

[0102] The processing module 502 is further used to adjust the recall weight corresponding to the bid requirement tag information in the position recall model, and determine the first matching position of the requirement tag information according to the adjusted position recall model.

[0103] The processing module 502 is further configured to display the first matching position to the job seeker.

[0104] In some embodiments of the present application, optionally, the tag information also includes customized tag information of the job seeker.

[0105] The processing module 502 is further used to determine a second matching position for the custom tag information according to the adjusted position recall model.

[0106] Processing module 502 is also used to input the first matching position, requirement label information, the second matching position and custom label information into the sorting model, and output the display order corresponding to the first matching position and the second matching position respectively through the sorting model, so that the priority display probability of the first matching position is greater than the priority display probability of the second matching position.

[0107] The processing module 502 is further configured to display the first matching position to the job seeker based on the display order, and to display the second matching position to the job seeker based on the display order.

[0108] In some embodiments of the present application, optionally, the acquisition module 501 is further used to acquire historical job-seeking behavior information within a preset time period, where the historical job-seeking behavior information includes multiple historical job-seeking users and associated positions of each historical job-seeking user.

[0109] The acquisition module 501 is further used to acquire the historical user information of each historical job-seeking user, the historical position information of the positions associated with the historical job-seeking user, and the preset user tag of each historical job-seeking user.

[0110] The processing module 502 is further used to generate a training sample set according to the historical user information of each historical job-seeking user, the historical position information of the positions associated with the historical job-seeking user, and the preset user label of each historical job-seeking user.

[0111] The processing module 502 is also used to train the pre-built label generation model based on the training sample set until the preset training conditions are met to obtain the target label generation model, wherein the label generation model includes an embedding layer, a feature recognition layer and an output layer.

[0112] It can be understood that the label generation device of the embodiment of the present application can correspond to the execution entity of the label generation method provided in the embodiment of the present application. The specific details of the operation and / or function of each module / unit of the model generation device can refer to the description of the corresponding parts of the label generation method provided in the above-mentioned embodiment of the present application. For the sake of brevity, they will not be repeated here.

[0113] 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 6 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 601 and a memory 602 storing computer program instructions. When the processor 601 executes the computer program instructions, the process or function of any of the above-mentioned embodiments of the method is implemented.

[0114] Specifically, the processor 601 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 602 may include a large-capacity memory for data or instructions. For example, the memory 602 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 device. For another example, the memory 602 may include a removable or non-removable (or fixed) medium. For another example, the memory 602 may be inside or outside the integrated gateway disaster recovery device. The memory 602 may be a non-volatile solid-state memory. In other words, the memory 602 usually 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 executed by one or more processors), the operation described in the method of the embodiment of the present application can be performed. The processor 601 implements the process or function of any method in the above embodiments by reading and executing the computer program instructions stored in the memory 602.

[0115] In one example, Figure 6 The electronic device shown may also include a communication interface 603 and a bus 610. Among them, the processor 601, the memory 602, and the communication interface 603 are connected through the bus 610 and complete the communication between each other. The communication interface 603 is mainly used to realize the communication between each module, device, unit and / or device in the embodiment of the present application. The bus 610 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 610 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.

[0116] 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.

[0117] 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.

[0118] 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.

[0119] 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.

[0120] 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 label generation method, characterized in that: The method comprises: Acquire user information of a job seeker and position information of positions related to the job seeker, wherein the positions related to the job seeker include positions for which the job seeker has viewed but not submitted resumes and positions for which the job seeker has viewed but submitted resumes; Inputting the user information of the job seeker and the position information of the position related to the job seeker into a pre-trained target label generation model, and outputting the job seeker's requirement label through the target label generation model; Wherein, the target label generation model includes an embedding layer, a feature recognition layer and an output layer; The outputting of the job seeker's demand label through the target label generation model includes: converting the user information into a user feature vector and converting the position information into a position feature vector through the embedding layer, determining the target features that the job seeker is concerned about and the demand intentions corresponding to the target features based on the user feature vector and the position feature vector through the feature recognition layer, converting the target features into demand labels through the output layer, outputting the demand labels, and outputting the demand intentions corresponding to the demand labels.

2. The method according to claim 1, characterized in that The user feature vector includes a first feature vector and a second feature vector; The converting the user information into a user feature vector comprises: Identifying unstructured feature information in the user information and structured feature information corresponding to the first preset attribute, wherein the user information includes at least any one of the following information: resume information, job communication information, job viewing behavior information, and job application behavior information, and the unstructured information is data without a predetermined format; Determine the attribute value of the structured feature information corresponding to each of the first preset attributes to obtain the first feature vector; Semantic understanding is performed on the unstructured feature information, text features are extracted, and the text features are converted into the second feature vector.

3. The method according to claim 2, characterized in that The converting the position information into a position feature vector comprises: Identifying structured feature information corresponding to a second preset attribute in the position information; Determine the attribute value of the structured feature information corresponding to each of the second preset attributes to obtain the position feature vector.

4. The method according to claim 1, characterized in that The feature recognition layer includes: an attention layer and a fully connected layer, and the position feature vector includes a first position feature vector for viewing positions corresponding to unsubmitted resumes and a second position feature vector for viewing positions corresponding to submitted resumes; The determining, by the feature recognition layer according to the user feature vector and the position feature vector, the target feature that the job seeker is concerned about and the demand intention corresponding to the target feature includes: Determining difference information according to the user feature vector, the first position feature vector, and the second position feature vector through the attention layer, and generating a feature vector of the target feature according to the difference information; The feature vector of the target feature is identified through the fully connected layer to determine the demand intention corresponding to the target feature.

5. The method according to claim 1, characterized in that The method further comprises: In response to the job seeker visiting the recruitment platform, acquiring the job seeker's demand tag information, wherein the demand tag information includes the job seeker's demand tag and demand intention; Adjusting the recall weight corresponding to the demand label information in the job recall model, and determining the first matching job for the demand label information according to the adjusted job recall model; The first matching position is displayed to the job seeker.

6. The method according to claim 5, characterized in that The tag information also includes the job seeker's customized tag information; Determining a second matching position for the custom tag information according to the adjusted position recall model; The displaying the first matching position to the job seeker includes: Inputting the first matching position, the demand tag information, the second matching position and the custom tag information into a sorting model, and outputting the display order corresponding to the first matching position and the second matching position respectively through the sorting model, so that the priority display probability of the first matching position is greater than the priority display probability of the second matching position; Based on the display order, the first matching position is displayed to the job seeker, and based on the display order, the second matching position is displayed to the job seeker.

7. The method according to claim 1, characterized in that Before obtaining the user information of the job seeker and the position information of the position associated with the job seeker, the method includes: Acquire historical job-seeking behavior information within a preset time period, wherein the historical job-seeking behavior information includes a plurality of historical job-seeking users and associated positions of each of the historical job-seeking users; Acquire the historical user information of each of the historical job-seeking users, the historical position information of the positions associated with the historical job-seeking users, and the preset user tag of each of the historical job-seeking users; Generate a training sample set according to the historical user information of each of the historical job-seeking users, the historical position information of the positions associated with the historical job-seeking users, and the preset user label of each of the historical job-seeking users; Based on the training sample set, the pre-constructed label generation model is trained until the preset training conditions are met to obtain a target label generation model, wherein the label generation model includes an embedding layer, a feature recognition layer and an output layer.

8. A label generating device, characterized in that: The device comprises: An acquisition module, used to acquire user information of a job seeker and position information of positions related to the job seeker, wherein the positions related to the job seeker include positions for which the job seeker has viewed but not submitted resumes and positions for which the job seeker has viewed but submitted resumes; A processing module, configured to input the user information of the job seeker and the position information of the position related to the job seeker into a pre-trained target label generation model, and output the demand label of the job seeker through the target label generation model; Wherein, the target label generation model includes an embedding layer, a feature recognition layer and an output layer; Among them, the processing module outputs the job seeker's demand label through the target label generation model, including: converting the user information into a user feature vector through the embedding layer, and converting the position information into a position feature vector, determining the target features that the job seeker is concerned about and the demand intentions corresponding to the target features according to the user feature vector and the position feature vector through the feature recognition layer, converting the target features into demand labels through the output layer, outputting the demand labels, and outputting the demand intentions corresponding to the demand labels.

9. 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 7 is implemented.

10. 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 7 is implemented.

11. 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 7.

Images