Job information recommendation methods, devices, electronic equipment and storage media

Abstract

This application provides a job information recommendation method, relating to the field of information recommendation technology. The method includes: in response to a search operation targeting a first search term, if it is determined that a user has a job-seeking intention based on a second search term and user characteristics, then displaying recommended job information determined based on the second search term and / or user characteristics; wherein the first search term does not include job-related words, and the search type of the search operation is not a job search type; the second search term is the first search term or a search term corresponding to a historical search operation including the first search term. This application embodiment can determine whether a user currently using the search function has a job-seeking intention, and if the current user has a job-seeking intention, then displaying recommended job information that matches their job-seeking intention.

Description

Technical Field

[0001] This application relates to the field of information recommendation technology, and more specifically, to a method, apparatus, electronic device, and storage medium for recommending job information. Background Technology

[0002] One important function of recruitment apps is to meet users' job search needs.

[0003] Users with a clear job-seeking intention typically use the job search function of recruitment apps to search for positions of interest. However, there are two problems with using job searches for job hunting:

[0004] Connecting users with valuable job positions is inefficient. Users often cannot quickly provide more valuable search terms, and to connect with more valuable job positions, they have to actively make multiple searches, which is very inefficient in connecting users with valuable job positions.

[0005] Valuable job postings fail to reach some potential job seekers. Besides users who actively use the job search function, there are also a large number of users with only potential job-seeking intentions. These users have already developed potential job-seeking intentions through industry exchanges and salary discussions within recruitment apps, but they haven't actively used the job search function, preventing them from quickly connecting with valuable job postings. Summary of the Invention

[0006] This invention provides a method, apparatus, electronic device, and storage medium for recommending job information to overcome or at least partially solve the above-mentioned problems.

[0007] Firstly, a job information recommendation method is provided, which includes:

[0008] In response to a search operation targeting the first search term, if it is determined that the user has a job-seeking intention based on the second search term and user characteristics, then recommended job information determined based on the second search term and / or user characteristics is displayed.

[0009] Wherein, the first search term does not include words related to job positions, and the search type of the search operation does not belong to the job search type;

[0010] The second search term is the first search term or a search term corresponding to a historical search operation, including the first search term.

[0011] In one possible implementation, determining that a user has job-seeking intent based on a second search term and user characteristics includes:

[0012] The second search term and user features are input into a pre-trained intent prediction model to obtain the intent prediction result output by the intent prediction model.

[0013] The intent prediction model is trained using the second search terms and user characteristics of the sample users as training samples, and the presence or absence of job-seeking intent of the sample users as training labels.

[0014] The intent prediction result is used to characterize whether the current user has a job-seeking intent.

[0015] In one possible implementation, the job information recommendation method further includes:

[0016] If the search type of the search operation is determined to be a job search, then recommended job information determined based on the second search term and / or user characteristics is displayed; and / or

[0017] If it is determined that the first search term includes job-related words, then recommended job information determined based on the second search term and / or user characteristics is displayed.

[0018] In one possible implementation, displaying the recommended job information determined based on the second search term and / or user characteristics includes:

[0019] Determine the job keyword information in the second search term and / or user characteristics, and use the job keyword information as the target key;

[0020] Search for the target key in a pre-built set of key-value pairs to obtain the job information corresponding to the target key, and use the job information as candidate job information;

[0021] Based on the candidate job information, the recommended job information is determined;

[0022] In this key-value pair set, the key of each key-value pair is job keyword information, and the value is job information including the job keyword information.

[0023] In one possible implementation, determining the job keyword information in the second search term and / or user characteristics, and using the job keyword information as the target key, includes:

[0024] Use the job keywords in the second search term as the first candidate key;

[0025] Use the job keyword information corresponding to the historical job information in the user characteristics as the second candidate key;

[0026] Obtain the co-occurrence frequency of each preset job information with the first candidate key and / or the second candidate key;

[0027] The job keyword information in each preset job information with a co-occurrence frequency greater than a preset threshold is used as the third candidate key;

[0028] The first candidate key, the second candidate key, and the third candidate key are used as the target key.

[0029] In one possible implementation, determining the recommended job information based on the candidate job information includes:

[0030] The candidate job information, the second search term, and user features are input into a pre-trained filtering model to obtain the judgment result output by the filtering model;

[0031] The filtering model is trained using the user characteristics of the sample users, the second search term, and the preset job information as training samples, and the relevance between the user characteristics of the sample users and the second search term and the preset job information is used as the training label.

[0032] The judgment result is used to characterize whether the currently input candidate job information is related to the user characteristics and the second search term;

[0033] If the currently input candidate job information is related to the user characteristics and the second search term, the currently input candidate job information will be used as the filtered candidate job information.

[0034] The filtered candidate job information, user characteristics, and second search term are input into a preset click-through rate model to obtain a first probability value output by the preset click-through rate model, which is used to characterize the user's click on the recommended job information.

[0035] The filtered candidate job information is sorted in descending order of the first probability value. The first preset number of job information in the sorted filtered candidate job information is selected as the recommended job information.

[0036] In one possible implementation, obtaining the first probability value output by the preset click-through rate model, which characterizes the user's click on the recommended job information, further includes:

[0037] The filtered candidate job information, user characteristics, and second search term are input into a preset submission rate model to obtain a second probability value output by the preset submission rate model, which is used to characterize the user's submission of a resume to the recommended job information.

[0038] The product of the first probability value and the second probability value is used as the sorting value;

[0039] The filtered candidate job information is sorted in descending order of the sorting value. The first preset number of job information from the sorted filtered candidate job information is selected as the recommended job information.

[0040] Secondly, a job information recommendation device is provided, comprising:

[0041] The display unit is used to respond to a search operation for a first search term. If it is determined that the user has a job-seeking intention based on the second search term and user characteristics, then it displays recommended job information determined based on the second search term and / or user characteristics.

[0042] Wherein, the first search term does not include words related to job positions, and the search type of the search operation does not belong to the job search type;

[0043] The second search term is the first search term or a search term corresponding to a historical search operation, including the first search term.

[0044] Thirdly, embodiments of the present invention provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method provided in the first aspect.

[0045] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method provided in the first aspect.

[0046] Fifthly, embodiments of the present invention provide a computer program including computer instructions stored in a computer-readable storage medium. When a processor of a computer device reads the computer instructions from the computer-readable storage medium, the processor executes the computer instructions, causing the computer device to perform steps implementing the method provided in the first aspect.

[0047] The job information recommendation method, apparatus, electronic device, and storage medium provided in this invention can determine whether a user currently using the search function has a job-seeking intention. If the user has a job-seeking intention, recommended job information matching their job-seeking intention is displayed to the user. This includes job recommendations from search results of both job search and non-job search behaviors, enabling valuable positions to be matched more efficiently with more users with job-seeking intentions, and better meeting the user's job-seeking needs. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below.

[0049] Figure 1 A flowchart illustrating a job information recommendation method provided in an embodiment of this application;

[0050] Figure 2 A schematic diagram illustrating the construction logic of a job candidate set provided in an embodiment of this application;

[0051] Figure 3 This application provides a flowchart illustrating a job information recommendation system.

[0052] Figure 4 This is a schematic diagram of the structure of a job information recommendation device provided in an embodiment of this application;

[0053] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0054] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting the invention.

[0055] Those skilled in the art will understand that, unless explicitly stated otherwise, the singular forms “a,” “an,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in the specification of this application means the presence of features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0056] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0057] In existing recruitment apps, users with a clear job-seeking intention typically use the app's job search function to find positions of interest. However, using job searches for employment presents two problems:

[0058] 1. The connection between users and valuable job positions is inefficient. Users often cannot quickly provide more valuable search terms. For example, a user whose current job title is "backend engineer" will likely use related terms like "backend development" or "Java backend" in their first few job searches. Since the search function primarily sorts results based on the relevance of search terms to job titles, other potentially valuable job titles such as "senior middle platform engineer," "backend system architect," and "infrastructure developer" are not displayed prominently due to their poor relevance to the search terms. Users must actively perform multiple searches to connect with these more valuable positions, which is highly inefficient in linking users with valuable job opportunities.

[0059] 2. Valuable job postings fail to reach some potential job seekers. Besides users actively using the job search function, there are many users with only potential job-seeking intentions. These users have already developed these intentions through industry exchanges and salary discussions within the platform, but haven't actively used the job search function, making it difficult to quickly connect them with valuable job postings. For example, if a user searches for "Company A benefits," and this indicates a potential job-seeking intention, a straightforward and valuable solution would be to recommend relevant positions from Company A to the user at this point.

[0060] The job information recommendation method, apparatus, electronic device, and computer-readable storage medium provided in this application are intended to solve the technical problems existing in the prior art.

[0061] The technical solutions of this application and how they solve the technical problems existing in the prior art are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0062] Figure 1 This is a flowchart of a job information recommendation method provided in an embodiment of this application, such as... Figure 1 As shown, the method may include:

[0063] S101. A search box is displayed on the display page of the smart terminal, and the search box is used to input search terms;

[0064] S102. In response to a search operation for the first search term, if it is determined that the user has a job-seeking intention based on the second search term and user characteristics, then the recommended job information determined based on the second search term and / or user characteristics is displayed.

[0065] Wherein, the first search term does not include words related to job positions, and the search type of the search operation does not belong to the job search type;

[0066] The second search term is the first search term or a search term corresponding to a historical search operation, including the first search term.

[0067] In the embodiments of this application, the current smart terminal displays a search box. After the user enters a first search term in the search box and triggers a search operation, the corresponding page after the search based on the first search term is displayed. If it is determined that the user has a job-seeking intention based on the second search term and user characteristics, then recommended job information determined based on the second search term and / or user characteristics is displayed.

[0068] Among these, "user's job-seeking intention" can mean that the user hopes to find a suitable job. The second search term can be the search term currently entered by the user, the user's historical search terms stored on the smart terminal or server, or a sequence of search terms generated by the user's continuous searches within a certain period of time. For example, if a user enters the search terms "Alibaba", "Tencent", "ByteDance", and "sales" in the search box within 10 minutes, these search terms form a search sequence ["Alibaba", "Tencent", "ByteDance", "sales"], and this search sequence is used as the second search term.

[0069] User characteristics can be attribute information, such as user ID, industry, company, age, current job information, and historical job information; they can also be behavioral information, such as keywords contained in the user's recent active pages and clicked activity; or they can be statistical information, such as the number of times a user logs into a certain application within a certain period of time and the duration of their activity.

[0070] It can display recommended job information determined by both the second search term and user characteristics, or it can display recommended job information determined by either the second search term or a certain type of information from user characteristics.

[0071] The display location of the recommended job information is not specifically limited in this application, as long as it allows users to see the recommended job information without affecting the user's overall operating experience.

[0072] Specifically, determining that a user has job-seeking intentions based on the second search term and user characteristics may include:

[0073] The second search term and user features are input into a pre-trained intent prediction model to obtain the intent prediction result output by the intent prediction model.

[0074] In the embodiments of this application, the intent prediction model can be a binary classification model, which can be any one of logistic regression classification model, support vector machine classification model, or deep neural network classification model, etc. Furthermore, the implementation of this invention is not limited to the binary classification model form corresponding to the above-mentioned intent prediction model, and the use of other binary classification prediction models does not affect the achievement of the invention's objective.

[0075] The user's second search term, i.e., historical search term or search term sequence generated by the user's continuous searches within a certain period of time, as well as user feature information including user attribute features, user behavior features, and user statistical features, are input into the pre-trained intent prediction model.

[0076] The intent prediction model is trained using the second search terms and user characteristics of the sample users as training samples, and the presence or absence of job-seeking intent of the sample users as training labels.

[0077] The intent prediction result is used to characterize whether the current user has a job-seeking intent.

[0078] In the embodiments of this application, the second search term information of multiple sample users and the user feature information corresponding to the users are used as training samples to train the intent prediction model.

[0079] After obtaining information on the second search terms and user characteristics of multiple sample users, it is also necessary to obtain historical operational behavior data of these users within a certain time period. For example, at least 180 days of historical operational behavior data can be obtained for multiple sample users. If a user's historical operational behavior data includes behaviors such as: the input search terms pointing to specific positions, the search type being a job search, and actively sending messages to headhunters, and such historical operational behaviors of the sample user occur multiple times, then this sample user is defined as a positive sample, indicating that the user has job-seeking intentions, i.e., the user hopes to find a suitable job. Conversely, if these behaviors are not present, the sample user is defined as a negative sample, indicating that the user does not have job-seeking intentions, i.e., the user has a low job-seeking desire.

[0080] Furthermore, the job information recommendation method of this application may also include:

[0081] If the search type of the search operation is determined to be a job search type, then recommended job information determined based on the second search term and / or user characteristics is displayed;

[0082] In embodiments of this application, the user's search operation can be of various types, including general search types and job search types. If the current user's search operation type is a job search type, it indicates that the current user expects to find a suitable job and has a job-seeking intention. If it is determined that the user has a job-seeking intention, then recommended job information determined based on the second search term and / or user characteristics is displayed.

[0083] Based on the above embodiments, as an optional embodiment, when the user has not yet performed a search operation on the search term, i.e., when the user uses the search box for the first time, the search interface displays the search box. At this time, when the user uses the search box to search, it will default to a general search type and display a general search page related to the current user's search term. The user can choose to convert the current general search page, for example, the current general search page can be converted to a job search page. If the current terminal's display page is a job search page, and the user uses the search box on the display page to perform a search operation, then the search operation type is a job search type, that is, it is assumed that the current user has a job-seeking intention, and then recommended job information determined based on the second search term and / or user characteristics is displayed.

[0084] And / or if it is determined that the first search term includes job-related words, then recommended job information determined based on the second search term and / or user characteristics is displayed.

[0085] In the embodiments of this application, if the first search term entered by the user contains job-related words, and a search is performed based on the first search term, it indicates that the user expects to find a suitable job and has a job-seeking intention. For example, if the user currently enters "salary for backend development," which contains the job-related word "backend development," and a search operation is triggered based on this search term, it indicates that the user expects to find information related to this search term, indirectly indicating that the user expects to find a job related to backend development, and indicating that the user has a job-seeking intention. If it is determined that the user has a job-seeking intention, recommended job information determined based on the second search term and / or user characteristics is displayed.

[0086] Specifically, displaying recommended job information determined based on the second search term and / or user characteristics may include:

[0087] Determine the job keyword information in the second search term and / or user characteristics, and use the job keyword information as the target key;

[0088] Search for the target key in a pre-built set of key-value pairs to obtain the job information corresponding to the target key, and use the job information as candidate job information;

[0089] In this key-value pair set, the key of each key-value pair is job keyword information, and the value is job information including the job keyword information.

[0090] In the embodiments of this application, such as Figure 2 As shown, the server needs to pre-build an offline candidate set, which can be a collection of multiple key-value pairs. The key of each key-value pair is job keyword information, such as: data development, product manager, big data architecture, etc. The value is the job information containing the job keyword information; for example, with product manager as the key, the corresponding values ​​are: senior product manager, navigation algorithm product manager, or overseas product management.

[0091] Different keys correspond to different values, and the same key can correspond to multiple job postings. Furthermore, the multiple job postings corresponding to each key can be pre-sorted based on factors such as the posting time, the number of postings, or the resume response rate. The key-value pair set corresponding to the aforementioned offline candidate set is then stored in a suitable storage system, such as Pika.

[0092] After constructing the offline candidate set and storing it in the corresponding location, it is necessary to determine the target key based on the second search term and / or user characteristics, and search for job information corresponding to the target key in the key-value pair set corresponding to the offline candidate set, and use the searched job information as candidate job information.

[0093] Based on the candidate job information, the recommended job information is determined;

[0094] In the embodiments of this application, since the amount of candidate job information obtained is too large to be directly displayed as recommended job information, it needs to be processed in order to obtain and display recommended job information.

[0095] Based on the above embodiments, as an optional embodiment, this application pre-constructs a dual-tower model. In the offline stage, the dual-tower model is used to process the massive job information in the underlying database to obtain the job information embedding vectors. The embedding vectors of the massive job information are stored in an ANN retrieval system, such as Faiss, and a vector proximity index service is constructed. In the online stage, the dual-tower model is used to obtain the embedding vectors corresponding to the latest user features. Based on the embedding vectors of user features and the embedding vectors of massive job information, the constructed vector proximity index service is combined to obtain a certain number of job information and use the certain number of job information as candidate job information.

[0096] Based on the above embodiments, as an optional embodiment, this application can set an adjustable threshold M when constructing the vector proximity index service. For example, if the threshold M is 100, it means that the 100 nearest job information will be returned in the vector proximity index service, that is, the candidate job information obtained based on this scheme includes 100 job information.

[0097] Specifically, determining the job keyword information in the second search term and / or user characteristics, and using the job keyword information as the target key, may include:

[0098] Use the job keywords in the second search term as the first candidate key;

[0099] In embodiments of this application, job keyword information in the second search term is used as the first candidate key, and the second search term may include: the first search term, search terms corresponding to the user's historical search operations, and a search sequence composed of search terms generated by the user's continuous searches within a certain time period. For example, if the search sequence composed of search terms continuously entered by the user within 10 minutes is ["sales", "sales manager", "key account manager", "online sales"], the job keyword information of ["sales manager", "key account manager", "online sales"] is used as the first candidate key.

[0100] Use the job keyword information corresponding to the historical job information in the user characteristics as the second candidate key;

[0101] In embodiments of this application, the user's historical job information is obtained from the user's characteristics. The user's historical job information may include the user's current job information. Job keywords in the user's historical job information are used as second candidate keys. For example, if the user's current job information is "Technical Lead", and the job sequence formed by the job information in their career experience is ["Senior Algorithm Engineer", "Recommendation Algorithm Engineer"], the job keywords in the corresponding job sequence ["Senior Algorithm Engineer", "Recommendation Algorithm Engineer", "Technical Lead"] are used as second candidate keys.

[0102] Obtain the co-occurrence frequency of each preset job information with the first candidate key and / or the second candidate key;

[0103] In the embodiments of this application, the massive job information in the underlying database of the server is used as each preset job information, and the co-occurrence frequency of each preset job information with the first candidate key and / or the second candidate key is obtained. Based on the level of co-occurrence frequency, the level of correlation between each preset job information and the first candidate key and / or the second candidate key can be determined.

[0104] The job keyword information in each preset job information with a co-occurrence frequency greater than a preset threshold is used as the third candidate key;

[0105] In embodiments of this application, if the co-occurrence frequency of a preset job posting with the first candidate key and / or the second candidate key is greater than a preset threshold, it indicates that the preset job posting has a high correlation with the first candidate key and / or the second candidate key, and the job keyword information corresponding to the preset job posting is used as the third candidate key. The co-occurrence frequency can be the frequency with which the second job posting appears on the search results page after a user searches using the first job posting as the search term. If this frequency exceeds the preset threshold, it indicates that the second job posting has a high correlation with the first job posting. This application does not specifically limit the value of the threshold, as long as it can characterize that the preset job posting has a high correlation with the first candidate key and / or the second candidate key after the co-occurrence frequency exceeds the threshold.

[0106] Based on the above embodiments, as an optional embodiment, this application uses job keyword information from job information that is highly correlated with the first candidate key and / or the second candidate key as the third candidate key. This application does not specifically limit how to determine whether a certain job information is highly correlated with the first candidate key and / or the second candidate key, as long as it can characterize that the job information is highly correlated with the first candidate key and / or the second candidate key. For example, historical job sequences corresponding to the historical job information of multiple users can be obtained. If two or more job information appear simultaneously in the historical job sequence, it can be determined that these two or more job information are highly correlated and can be used as each other as the third candidate key.

[0107] The first candidate key, the second candidate key, and the third candidate key are used as the target key.

[0108] In the embodiments of this application, the first candidate key, the second candidate key, and the third candidate key can be used as the target key, and the job information corresponding to the target key can be obtained from the key-value pair set in the offline candidate set, and the job information can be used as the candidate job information.

[0109] Specifically, determining the recommended job information based on the candidate job information includes:

[0110] The candidate job information, the second search term, and user features are input into a pre-trained filtering model to obtain the judgment result output by the filtering model;

[0111] In the embodiments of this application, due to the large number of candidate job postings, and the inevitable presence of some job postings that negatively impact user experience, filtering of the candidate job postings is necessary. The preset filtering model in this application can be a binary classification model. The input data can be a combination of <<user characteristics, second search keyword>, and candidate job posting characteristics>. The candidate job posting characteristics can include: job title, job description, work experience requirements, education requirements, job highlights, salary, work location, posting time, information completeness, and poster information. User characteristics can include: user ID, industry, company, current job, historical job list, skill tags, age, region, and historical company list. The second search keyword can include: the text of the second search keyword, keywords within the second search keyword, and job or company information contained within the second search keyword.

[0112] The filtering model is trained using user characteristics of sample users, a second search term, and preset job information as training samples, and using the relevance between the user characteristics of the sample users and the second search term and the preset job information as training labels. In embodiments of this application, the training samples of the filtering model can be user characteristics of sample users, a second search term, and preset job information, with positive samples indicating relevance between the user characteristics of the sample users and the second search term and the preset job information, and negative samples indicating no relevance. The filtering model is obtained through iterative training using the above training samples.

[0113] The judgment result is used to characterize whether the currently input candidate job information is related to the user characteristics and the second search term;

[0114] If the currently input candidate job information is related to the user characteristics and the second search term, the currently input candidate job information will be used as the filtered candidate job information.

[0115] In the embodiments of this application, the judgment result of each candidate job information is obtained through a filtering model. If the judgment result is that the currently input candidate job information is related to the user feature and the second search term, the currently input candidate job information is used as the filtered candidate job information.

[0116] Based on the above implementations, as an optional embodiment, this application can filter candidate job information based on multiple rules, such as filtering candidate job information whose posting company is the same as the user's current company, or filtering candidate job information that is suspected of being false (such as being reported for inaccurate salary information). This application does not specifically limit the specific filtering rules, as long as they can filter some candidate job information that is not good for the user or job posting.

[0117] The filtered candidate job information, user characteristics, and second search term are input into a preset click-through rate model to obtain a first probability value output by the preset click-through rate model, which is used to characterize the user's click on the recommended job information.

[0118] In the embodiments of this application, after obtaining the filtered job information, it is necessary to sort it to obtain job information with a higher sequence, and then display it as recommended job information.

[0119] The click-through rate (CTR) model presupposed in this application can be a binary classification model, whose input is a data combination of <<user characteristics, second search keyword>, and filtered candidate job information>. Here, user click behavior can be defined as a user clicking to view the details of recommended job information. Records where a user browses but does not click are defined as negative samples, and records where a user browses and clicks to view job details are defined as positive samples. The CTR model can be obtained through iterative training using the above training samples. Simultaneously, during the online phase, the latest user characteristic information is acquired in real time and input into the CTR model using the data combination of <<user characteristics, second search keyword>, and filtered candidate job information>.

[0120] In the embodiments of this application, based on a click-through rate (CTR) model, this application can obtain a probability value representing the probability that a user browses and clicks on recommended job information. A corresponding probability value can be obtained for each filtered candidate job posting; the higher the probability value, the more likely the user is to browse and click on the recommended job information.

[0121] The filtered candidate job information is sorted in descending order of the first probability value, and the first preset number of job information in the sorted filtered candidate job information is selected as the recommended job information.

[0122] In the embodiments of this application, the magnitude of the obtained first probability value can be used as the sorting criterion to sort the filtered recommended job information to obtain a sorted list of recommended job information, and the top n job information in the sorted list can be selected as recommended job information according to the scenario requirements and displayed in an appropriate position on the corresponding display page.

[0123] Furthermore, the step of obtaining the first probability value output by the preset click-through rate model, used to characterize the user's click on the recommended job information, may further include:

[0124] The filtered candidate job information, user characteristics, and second search term are input into a preset submission rate model to obtain a second probability value output by the preset submission rate model, which is used to characterize the user's submission of a resume to the recommended job information.

[0125] In the embodiments of this application, after obtaining the probability value of the click-through rate model corresponding to each filtered candidate job information, the data combination of <<user features, second search keyword>, and filtered candidate job information> can be input into a preset submission rate model to obtain the probability value of the submission rate model. The preset submission rate model can be a binary classification model, where the submission behavior can be a user submitting a resume to the recommended job information. The training sample data can be user features, the second search keyword, and the filtered candidate job information. A positive sample is defined as a user who clicked on the recommended job information and submitted a resume, and a negative sample is defined as a user who clicked on the job recommendation but did not submit a resume. Through iterative training of the above training samples, a trained submission rate model can be obtained. By inputting the data combination of <<user features, second search keyword>, and filtered candidate job information> into the submission rate model, a probability value representing a user clicking on the recommended job information and submitting a resume can be obtained. A corresponding probability value can be obtained for each filtered candidate job information; the larger the probability value, the greater the probability that the user clicked on the recommended job information and submitted a resume.

[0126] The product of the first probability value and the second probability value is used as the sorting value;

[0127] The filtered candidate job information is sorted in descending order of the sorting value. The first preset number of job information from the sorted filtered candidate job information is selected as the recommended job information.

[0128] In the embodiments of this application, the probability value of the click-through rate model, the probability value of the delivery rate model, or the product of the probability values ​​of the two models can be used as the sorting criterion for the filtered candidate job information. This application does not specifically limit which probability value is used as the sorting criterion for the filtered candidate job information; any value that meets the needs of the current scenario is acceptable.

[0129] In the embodiments of this application, the probability value of the click-through rate model, the probability value of the delivery rate model, or the product of the probability values ​​of the two models can be used as the sorting criterion for the filtered candidate job information. This application does not specifically limit which probability value is used as the sorting criterion for the filtered candidate job information; any value that meets the needs of the current scenario is acceptable.

[0130] Based on the above embodiments, as an optional embodiment, such as... Figure 3 As shown, this application also designs a job information recommendation system, which is a gRPC service implemented in Golang language, deployed on multiple Linux servers, and uses Nginx for service management and load balancing. The job information recommendation system may include four sub-modules: intent prediction 201, candidate set recall 202, candidate set filtering 203, and model ranking 204. The intent prediction 201 includes: predicting whether a user has a job-seeking intent through an intent prediction model or determining whether a user has a job-seeking intent through certain rules; the candidate set recall 202 includes: constructing an offline candidate set and recalling candidate job information, wherein the recall can take the form of constructing a first and second target key and searching for job information corresponding to the first and second target keys, or constructing a third target key and searching for job information corresponding to the third target key; the candidate set filtering 203 includes: obtaining filtered candidate job information through a filtering model, or filtering candidate job information through other methods; the model ranking 204 includes: obtaining the probability values ​​corresponding to the filtered candidate job information through a click-through rate model and / or delivery rate model, arranging the filtered candidate job information according to the order of the probability values, and selecting the top n job information as recommended job information for display.

[0131] Based on the above embodiments, as an optional implementation, the intent prediction model, filtering model, click-through rate model, and delivery rate model in this application system all utilize deep learning technology. Offline training uses open-source deep learning frameworks (such as Tensorflow), while online training provides online inference services for deep models to perform the corresponding model prediction functions. Also as an optional implementation, user characteristics and job information can be stored in an independent external feature query service (feature server), and the online system retrieves user characteristics and job information by calling the feature server.

[0132] This application provides a job information recommendation device, such as... Figure 4As shown, the device may include: a display unit 301, specifically: the display unit 301 is used to respond to a search operation for a first search term, and if it is determined that the user has a job-seeking intention based on the second search term and user characteristics, then display recommended job information determined based on the second search term and / or user characteristics;

[0133] Wherein, the first search term does not include words related to job positions, and the search type of the search operation does not belong to the job search type;

[0134] The second search term is the first search term or a search term corresponding to a historical search operation, including the first search term.

[0135] The job information recommendation device provided in this embodiment of the invention specifically executes the process described in the above method embodiment. For details, please refer to the content of the above job information recommendation method embodiment; it will not be repeated here. The job information recommendation device provided in this embodiment of the invention recommends jobs from search results, including both job searches and non-job searches, enabling valuable positions to be matched more efficiently with more users with job-seeking intentions, thus better meeting users' job-seeking needs.

[0136] This application provides an electronic device, which includes a memory and a processor; at least one program stored in the memory, which, when executed by the processor, can, compared with the prior art, perform job recommendations in search results including job searches and non-job searches, so that valuable positions can be matched more efficiently with more users with job-seeking intentions, and better meet the job-seeking needs of users.

[0137] In one alternative embodiment, an electronic device is provided, such as Figure 5 As shown, Figure 5 The illustrated electronic device 4000 includes a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may also include a transceiver 4004. It should be noted that in practical applications, the transceiver 4004 is not limited to one type, and the structure of this electronic device 4000 does not constitute a limitation on the embodiments of this application.

[0138] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0139] Bus 4002 may include a pathway for transmitting information between the aforementioned components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 4002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0140] The memory 4003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0141] The memory 4003 stores application code that executes the scheme of this application, and its execution is controlled by the processor 4001. The processor 4001 executes the application code stored in the memory 4003 to implement the content shown in the foregoing method embodiments.

[0142] This application provides a computer-readable storage medium storing a computer program that, when run on a computer, enables the computer to execute the corresponding content described in the aforementioned method embodiments. Compared with the prior art, by recommending jobs in search results including both job searches and non-job searches, valuable positions are more efficiently matched with more users with job-seeking intentions, thus better meeting users' job-seeking needs.

[0143] This application provides a computer program including computer instructions stored in a computer-readable storage medium. When a processor of a computer device reads the computer instructions from the computer-readable storage medium, the processor executes the computer instructions, causing the computer device to perform the content shown in the foregoing method embodiments. Compared with the prior art, by recommending jobs in search results including both job searches and non-job searches, valuable positions are matched more efficiently with more users with job-seeking intentions, better meeting users' job-seeking needs.

[0144] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0145] The above are only some embodiments of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A job information recommendation method, characterized in that, include: In response to a search operation targeting the first search term, if it is determined that the user has a job-seeking intention based on the second search term and user characteristics, then recommended job information determined based on the second search term and / or user characteristics is displayed. Wherein, the first search term does not include words related to job positions, and the search type of the search operation does not belong to the job search type; The second search term is the first search term or a search term corresponding to a historical search operation, including the first search term. The display of recommended job information determined based on the second search term and / or user characteristics includes: Determine the job keyword information in the second search term and / or user characteristics, and use the job keyword information as the target key; Search for the target key in a pre-built set of key-value pairs to obtain the job information corresponding to the target key, and use the job information as candidate job information; Based on the candidate job information, the recommended job information is determined; In the key-value pair set, the key of each key-value pair is job keyword information, and the value is job information including the job keyword information; The method for obtaining candidate job information further includes: The job information in the underlying database is processed by a pre-built dual-tower model to obtain the embedding vector of the job information; The embedding vector corresponding to the latest user features is obtained through the dual-tower model. Based on the embedding vector of the job information, the embedding vector of the user information, and the constructed vector proximity index service, a preset number of job information is obtained, and the preset number of job information is used as candidate job information.

2. The job information recommendation method according to claim 1, characterized in that, The determination that a user has job-seeking intentions based on the second search term and user characteristics includes: The second search term and user features are input into a pre-trained intent prediction model to obtain the intent prediction result output by the intent prediction model. The intent prediction model is trained using the second search terms and user characteristics of the sample users as training samples, and the presence or absence of job-seeking intent of the sample users as training labels. The intent prediction result is used to characterize whether the current user has a job-seeking intent.

3. The job information recommendation method according to claim 1, characterized in that, The method further includes: If the search type of the search operation is determined to be a job search, then recommended job information determined based on the second search term and / or user characteristics is displayed; and / or If it is determined that the first search term includes job-related words, then recommended job information determined based on the second search term and / or user characteristics is displayed.

4. The job information recommendation method according to claim 1, characterized in that, The step of determining the job keyword information in the second search term and / or user characteristics, and using the job keyword information as the target key, includes: Use the job keywords in the second search term as the first candidate key; Use the job keyword information corresponding to the historical job information in the user characteristics as the second candidate key; Obtain the co-occurrence frequency of each preset job information with the first candidate key and / or the second candidate key; The job keyword information in each preset job information with a co-occurrence frequency greater than a preset threshold is used as the third candidate key; The first candidate key, the second candidate key, and the third candidate key are used as the target key.

5. The job information recommendation method according to claim 1, characterized in that, The step of determining the recommended job information based on the candidate job information includes: The candidate job information, the second search term, and user features are input into a pre-trained filtering model to obtain the judgment result output by the filtering model; The filtering model is trained using the user characteristics of the sample users, the second search term, and the preset job information as training samples, and the relevance between the user characteristics of the sample users and the second search term and the preset job information is used as the training label. The judgment result is used to characterize whether the currently input candidate job information is related to the user characteristics and the second search term; If the currently input candidate job information is related to the user characteristics and the second search term, the currently input candidate job information will be used as the filtered candidate job information. The filtered candidate job information, user characteristics, and second search term are input into a preset click-through rate model to obtain a first probability value output by the preset click-through rate model, which is used to characterize the user's click on the recommended job information. The filtered candidate job information is sorted in descending order of the first probability value. The first preset number of job information in the sorted filtered candidate job information is selected as the recommended job information.

6. The job information recommendation method according to claim 5, characterized in that, The process of obtaining the first probability value output by the preset click-through rate model, used to characterize a user clicking on recommended job information, further includes: The filtered candidate job information, user characteristics, and second search term are input into a preset submission rate model to obtain a second probability value output by the preset submission rate model, which is used to characterize the user's submission of a resume to the recommended job information. The product of the first probability value and the second probability value is used as the sorting value; The filtered candidate job information is sorted in descending order of the sorting value. The first preset number of job information from the sorted filtered candidate job information is selected as the recommended job information.

7. A job information recommendation device, characterized in that, include: The display unit is used to respond to a search operation for a first search term. If it is determined that the user has a job-seeking intention based on the second search term and user characteristics, then it displays recommended job information determined based on the second search term and / or user characteristics. Wherein, the first search term does not include words related to job positions, and the search type of the search operation does not belong to the job search type; The second search term is the first search term or a search term corresponding to a historical search operation, including the first search term. The display of recommended job information determined based on the second search term and / or user characteristics includes: Determine the job keyword information in the second search term and / or user characteristics, and use the job keyword information as the target key; Search for the target key in a pre-built set of key-value pairs to obtain the job information corresponding to the target key, and use the job information as candidate job information; Based on the candidate job information, the recommended job information is determined; In the key-value pair set, the key of each key-value pair is job keyword information, and the value is job information including the job keyword information; The method for obtaining candidate job information further includes: The job information in the underlying database is processed by a pre-built dual-tower model to obtain the embedding vector of the job information; The embedding vector corresponding to the latest user features is obtained through the dual-tower model. Based on the embedding vector of the job information, the embedding vector of the user information, and the constructed vector proximity index service, a preset number of job information is obtained, and the preset number of job information is used as candidate job information.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the job information recommendation method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause the computer to perform the steps of the job information recommendation method as described in any one of claims 1 to 6.

Images