User portrait label generation system for recruitment

Abstract

The invention discloses a user portrait label generation system for recruitment, and the system comprises a hierarchical data warehouse which comprises a data detail layer, a data aggregation layer and a data application layer, and is used for storing preprocessed user detail data, user behavior data and a user portrait table; the label intelligent mining module is used for calling the AI large model to execute the following operations: mining a user portrait label through time dimension extension, user behavior clustering and industry pain point analysis based on a portrait label mining instruction and existing user portrait label information, and generating a new user portrait label and label calculation logic; the label ETL generation module is used for constructing ETL processing logic depending on the user behavior data in the data summarization layer and / or the user detail data in the data detail layer based on the new user portrait label and the label calculation logic; and writing the user portrait label data generated by calculation into a user portrait table. The label output efficiency and quality can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data processing, and particularly relates to a user portrait label generation system for recruitment. BACKGROUND

[0002] In the current recruitment industry, accurate matching of positions and candidates is the core of improving recruitment efficiency and job-seeking experience. User portrait, as a key technology to achieve accurate matching, can provide a data basis for intelligent recommendation and accurate search of positions / resumes by labeling users on the job-seeking side (C side) and users on the recruitment side (B side) with various labels, such as "senior JAVA engineer", "active job seeker", and "high-intention candidate".

[0003] However, the existing user portrait label generation scheme has the following defects: on the one hand, in related technologies, a user portrait table is often directly calculated based on an original data layer, and lacks a scientific and reasonable data warehouse layering design. This results in a large amount of complex calculation logic directly depending on the underlying massive data, and there is a large amount of repeated calculation and redundant processing, which not only causes a huge waste of computing resources, but also leads to low efficiency of task execution for label output, and it is difficult to cope with the increasingly high real-time requirements of the recruitment scene.

[0004] On the other hand, the creation of labels currently mainly relies on business personnel to propose requirements, and data warehouse developers manually develop according to the requirements. This way highly depends on human experience, and not only is slow in response, but also is difficult to deeply mine potential value labels hidden in data. For example, for industry pain points in the recruitment field, such as poor resume quality of B side, high communication cost, or low job-seeking efficiency and low feedback rate of C side, it is still difficult to effectively and quickly convert them into quantifiable data label products. SUMMARY

[0005] Therefore, the embodiments of the present application provide a user portrait label generation system for recruitment to solve at least one of the above technical problems.

[0006] In a first aspect, the embodiments of the present application provide a user portrait label generation system for recruitment, comprising:

[0007] a layered data warehouse, comprising a data detail layer, a data aggregation layer, and a data application layer, the data detail layer is configured to store user detail data after preprocessing of original user data, the data aggregation layer is configured to store user behavior data after preprocessing of the original user data, and the data application layer is configured to store a user portrait table;

[0008] The label intelligent mining module is in communication connection with the hierarchical data warehouse, and is configured to receive the portrait label mining instruction and the existing user portrait label information in the user portrait table, and call the AI large model to perform the following operations: based on the portrait label mining instruction and the existing user portrait label information, the user portrait label is mined through time dimension expansion, user behavior clustering, analysis of industry pain points of the recruitment end and / or the job-seeking end, to generate new user portrait labels and corresponding label calculation logics;

[0009] The label ETL generation module is in communication connection with the hierarchical data warehouse and the label intelligent mining module, and is configured to:

[0010] After the new user portrait label and the label calculation logic are audited, based on the new user portrait label and the label calculation logic, ETL processing logic dependent on user behavior data in the data summary layer and / or user detailed data in the data detailed layer is constructed; and the calculated user portrait label data is written into the user portrait table in the data application layer.

[0011] According to some embodiments of the present application, optionally, the user portrait label generation system further comprises: an intermediate table management module in communication connection with the hierarchical data warehouse, configured to create and maintain a data intermediate table for a business party, and the data structure of the data intermediate table is derived from the user portrait table; the intermediate table management module is configured to receive a data structure change instruction from the business party; and according to the data structure change instruction, perform field addition, disablement or deletion operation on the data intermediate table, and the operation is independent of the structure change of the user portrait table itself.

[0012] According to some embodiments of the present application, optionally, the AI large model is specifically configured to perform the following operations: identify a statistical label generated based on a first statistical period in the existing user portrait label information, and based on the business meaning of the statistical label, evaluate and generate a same-type statistical label based on a second statistical period and its label calculation logic; identify a plurality of existing user portrait labels describing the same-type or same graph behavior of the user, combine or cluster analyze the plurality of existing user portrait labels, and generate a new user portrait label representing the comprehensive characteristics of the user and its label calculation logic; based on an input industry pain point instruction of the recruitment end or the job-seeking end, locate a plurality of labels related to solving the pain point in the existing user portrait labels as feature labels, and combine the plurality of feature labels to generate a new user portrait label quantifying the industry pain point and its label calculation logic.

[0013] According to some embodiments of the present application, optionally, the new user portrait label representing the comprehensive characteristics of the user includes a user active degree, a user job-seeking intention clarity, and / or a user active degree; and the new user portrait label quantifying the industry pain point includes a candidate quality score, a candidate reply willingness score, a capability matching degree, and / or an intention to hire degree.

[0014] According to some embodiments of this application, optionally, the user profile tag generation system further includes: a tag quality monitoring module, which is connected to a hierarchical data warehouse and is used to perform overall quality monitoring and key field monitoring on the user profile table at preset intervals; wherein, overall quality monitoring includes monitoring the total amount of abnormal data and the number of fields with all null values; key field monitoring includes monitoring whether the distribution of enumeration values ​​in enumeration-type tags is abnormal and monitoring the proportion of numerical tags whose values ​​exceed preset thresholds.

[0015] According to some embodiments of this application, optionally, the tag quality monitoring module is specifically used to perform the following operations: calculate the total amount of data in the user profile table for the day, and determine whether its year-on-year change rate with the total amount of data yesterday and / or its month-on-month change rate with the total amount of data in the same period last week exceeds a preset first threshold range; if it exceeds, it is determined that the total amount of data is abnormal; traverse each field of the user profile table, calculate its null value rate, and determine whether the number of fields with a null value rate of 100% is greater than zero; if it is greater than zero, trigger an alarm; for a specified enumeration type tag, count the proportion of each enumeration value in the data for the day, and determine whether the fluctuation range of the proportion of any enumeration value and the proportion in the same period of history exceeds a preset second threshold range; if it exceeds, it is determined that the distribution is abnormal; for a specified numerical type tag, count the number of records whose values ​​exceed a preset reasonable value range, and calculate the percentage of this number to the total amount of data; if this percentage exceeds a preset third threshold, it is determined to be abnormal.

[0016] According to some embodiments of this application, optionally, the user profile tag generation system further includes: a tag usage frequency monitoring module, which is used to collect task configuration logs of the downstream user identification platform, interface call logs that provide profile data, and query logs of the downstream business database, aggregate and analyze the usage frequency of each user profile tag, and issue an alarm for user profile tags whose usage frequency is lower than a preset threshold.

[0017] According to some embodiments of this application, optionally, the user profile tag generation system further includes: a common tag management module, used to disable or delete the corresponding field in the data intermediate table when it is necessary to take the user profile tag offline in the user profile table, so that the business party cannot perceive it; and to comment out the generation logic of the user profile tag in the user profile table and set the field value to NULL while retaining the field structure; when it is necessary to re-enable, to undo the disabling or deletion operation in the data intermediate table and release the commented generation logic in the user profile table.

[0018] According to some embodiments of this application, optionally, the user profile tag generation system further includes: a temporary tag management module, used to classify user profile tags related to activities or individual businesses as temporary tags in response to the temporary tag marking instructions of developers; wherein, the temporary tags are not stored in the user profile table, and have task codes and output tables independent of the main generation task of the user profile table, and the data of the temporary tags are provided to the business party after being merged with the data of the user profile table through data synchronization.

[0019] According to some embodiments of this application, optionally, the temporary tag management module is also used to pause the independent task of the temporary tag when it needs to be taken offline, and retain the task code and output table structure of the independent task; when it needs to be reactivated, the independent task is reactivated.

[0020] The user profile tag generation system for recruitment provided by the embodiments of this application, on the one hand, by constructing a layered data warehouse including a Data Detail Layer (DWD), a Data Aggregation Layer (DWS), and a Data Application Layer (DM), can build complex tag calculation logic on pre-processed and lightly summarized data, avoiding repetitive calculations on massive amounts of raw data, simplifying ETL processing logic, reducing processing time and required CPU, memory, and other computing resources, and improving tag output efficiency. On the other hand, by calling AI large-scale models through the tag intelligent mining module and integrating them with the needs of the recruitment field, it can automatically and intelligently generate new user profile tags and their tag calculation logic based on time dimension expansion, user behavior clustering, and analysis of industry pain points. This changes the past single and passive mode that relied solely on manual demand submission, shortening the cycle from tag discovery to online deployment. The system can proactively mine deep-level, high-value tags such as "user willingness to actively seek employment" and "candidate quality score," solving industry pain points such as "difficulty in recruitment" and "low job search efficiency," and helping to enhance the support and value of user profiles for recruitment business. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings of the embodiments of this application will be briefly described below.

[0022] Figure 1 This is a structural block diagram of a user profile tag generation system for recruitment provided in an embodiment of this application.

[0023] Figure 2 This is a schematic diagram of the execution flow of an AI large model in a user profile tag generation system for recruitment provided in an embodiment of this application.

[0024] Figure 3Another structural block diagram of the user profile tag generation system for recruitment provided in the embodiments of this application.

[0025] Figure 4 This is a schematic diagram of the execution flow of the tag quality monitoring module in the user profile tag generation system for recruitment provided in this application embodiment.

[0026] Figure 5 This is another structural block diagram of the user profile tag generation system for recruitment provided in the embodiments of this application. Detailed Implementation

[0027] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0028] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0029] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0030] Various modifications and variations can be made to this application without departing from its spirit or scope, which will be apparent to those skilled in the art. Therefore, this application is intended to cover modifications and variations falling within the scope of the corresponding claims (the claimed technical solutions) and their equivalents. It should be noted that the embodiments provided in this application can be combined with each other without contradiction.

[0031] Before describing the technical solutions provided in the embodiments of this application, in order to facilitate understanding of the embodiments of this application, this application first specifically explains the problems existing in the related technologies:

[0032] In today's recruitment industry, accurate job matching is key to improving recruitment efficiency and the job seeker experience. User profiling, as a crucial technological support for achieving accurate matching, provides a data foundation for intelligent job / resume recommendations and precise searches by tagging job seekers (C-end) and recruiting companies (B-end) with various labels such as "Senior Java Engineer," "Active Job Seeker," and "Highly Intended Candidate."

[0033] However, existing user profile tag generation solutions have the following drawbacks: On the one hand, in related technologies, user profile tables are often directly based on statistical calculations from the raw data layer, lacking a scientifically sound and reasonable data warehouse layered design. This results in a large amount of complex computational logic directly relying on the underlying massive amounts of data, leading to a large amount of repetitive calculations and redundant processing. This not only causes a huge waste of computing resources but also results in low task execution efficiency for tag generation, making it difficult to cope with increasingly demanding real-time recruitment scenarios.

[0034] On the other hand, currently, tag creation mainly relies on business personnel submitting requirements, which are then manually developed by data warehouse developers. This approach is highly dependent on human experience, resulting in slow response times and difficulty in deeply mining the potential value tags hidden within the data. For example, industry pain points in the recruitment field, such as poor resume quality and high communication costs for B2B clients, or low job search efficiency and low response rates for B2C clients, are still difficult to effectively and quickly translate into quantifiable data tag products. Furthermore, while AI technology has demonstrated its value in general fields, in the recruitment field, there is a lack of mature and systematic solutions for enabling AI to understand complex industry pain points and proactively discover and generate valuable new tags.

[0035] In view of this, the present application provides a user profile tag generation system for recruitment, which can solve at least one technical problem in the current recruitment field: low efficiency of user profile tag production, serious waste of resources, and lack of intelligent mining capabilities.

[0036] The following describes the user profile tag generation system for recruitment provided in the embodiments of this application.

[0037] Figure 1 This is a structural block diagram of a user profile tag generation system for recruitment provided in an embodiment of this application. Figure 1 As shown, the user profile tag generation system 10 for recruitment provided in this application embodiment may include a hierarchical data warehouse 110, a tag intelligent mining module 120, and a tag ETL generation module 130. The modules work together to achieve efficient and intelligent generation of user profile tags.

[0038] The layered data warehouse 110 adopts a layered design concept to simplify the data processing flow, avoid redundant calculations, and improve tag output efficiency. Specifically, the layered data warehouse 110 can include at least a data detail layer (DWD layer) 111, a data aggregation layer (DWS layer) 112, and a data application layer (DM layer) 113. The data detail layer 111 can be used to store user detail data after preprocessing the original user data, such as user detail data obtained after cleaning and integrating the original user data from the data storage layer (ODS layer). User detail data can be stored in the user detail table (or user basic information table) in the data detail layer 111. User detail data can include multiple basic information fields for users, such as user ID, name, age, education, major, work experience, and hobbies.

[0039] The data aggregation layer 112 can be used to store user behavior data after preprocessing the original user data. For example, the data aggregation layer 112 can divide data domains according to business scenarios, such as chat domains, order transaction domains, etc., and summarize the user behavior data in the original user data according to different data domains to form wide tables of user behavior data corresponding to each data domain. Each wide table of user behavior data can store the user behavior data corresponding to that data domain, such as aggregated metrics such as the number of times a user browses, submits, and chats on that day.

[0040] The data application layer 113 can be used to store user profile tables. These tables can store multiple user profile tags (or fields), such as user ID, age, education level, current city, number of app views in the past 7 days, number of resumes submitted in the past 30 days, and active time periods. The multiple user profile tags and their data in the user profile table depend on the user detail data in the data detail layer 111 and / or the user behavior data in the data aggregation layer 112. Taking the "number of views in the past 7 days" tag in the user profile table as an example, it only requires reading and summing the daily views from the most recent 7 days in the data aggregation layer 112.

[0041] In this way, the heavy and repetitive low-level calculations are completed in advance at the DWD / DWS layer, forming a common data middleware layer. The user profile table only needs to be obtained by performing lightweight calculations or aggregations based on these pre-processed, relatively small-volume middleware data, which avoids starting calculations from massive amounts of raw data every time and improves data processing efficiency.

[0042] The tag intelligent mining module 120 is connected to the hierarchical data warehouse 110. It can be used to receive profile tag mining instructions and existing user profile tag information in the user profile table, and call the AI ​​big model to perform the following operations: based on the profile tag mining instructions and existing user profile tag information, to mine user profile tags by expanding the time dimension, clustering user behavior, and analyzing industry pain points of the recruitment end and / or job seeker end, and generate new user profile tags and their corresponding tag calculation logic.

[0043] Specifically, the tag intelligent mining module 120 can receive profile tag mining instructions from business parties or developers, such as profile tag mining logic or profile tag mining requirements, and access the user profile table in the data application layer 113 to obtain existing user profile tag information as the basis for analysis. Then, the tag intelligent mining module 120 performs the following operations by calling a pre-trained or fine-tuned AI model: based on the profile tag mining instructions and existing user profile tag information, it mines user profile tags by expanding the time dimension, clustering user behavior, and analyzing industry pain points on the recruitment and / or job-seeking sides, generating new user profile tags and their corresponding tag calculation logic.

[0044] For example, time dimension expansion can include identifying existing short-term statistical tags, such as the number of times in the past day, analyzing and suggesting the generation of similar tags for medium- and long-term dimensions (such as the past 7 days and the past 30 days). User behavior clustering can include comprehensively analyzing multiple tags describing similar user behaviors (such as submitting applications, chatting, and refreshing resumes) to mine comprehensive feature tags that reflect users' deeper intentions, such as a user's willingness to actively reach out. Analyzing industry pain points on the recruitment and / or job-seeking sides can include identifying relevant tags as feature tags based on industry pain points on the recruitment and / or job-seeking sides, and constructing new tags that can quantify these pain points, such as a candidate quality score. Tag calculation logic can be used to define the data sources and calculation rules for new user profile tags, that is, which fields in the data detail layer 111 and / or data aggregation layer 112 are needed to calculate the tag value of a new user profile tag, and how these fields are used to calculate the tag value of the new user profile tag.

[0045] After the tag intelligent mining module 120 generates new user profile tags and their corresponding tag calculation logic, it can be manually reviewed to evaluate the tag value.

[0046] The tag ETL generation module 130 can communicate with the hierarchical data warehouse 110 and the tag intelligent mining module 120. The tag ETL generation module 130 can be used to construct ETL processing logic that depends on user behavior data in the data summary layer and / or user detail data in the data detail layer after the new user profile tags and tag calculation logic have been approved; and write the calculated user profile tag data into the user profile table in the data application layer.

[0047] Specifically, after the new user profile tags and their calculation logic have passed manual review, the tag ETL generation module 130 can construct corresponding ETL (Extract-Transform-Load) processing logic based on the new user profile tags and tag calculation logic. This logic relies on user behavior data from the data aggregation layer and / or user detail data from the data detail layer, utilizing intermediate results from the hierarchical data warehouse to avoid calculations from raw data, thus ensuring processing efficiency. Then, the tag ETL generation module 130 executes this ETL logic, writing the calculated user profile tag data into the user profile table in the data application layer, completing the new tag deployment process. Extraction involves reading data from the data detail layer and / or data aggregation layer tables in the data warehouse; Transformation involves calculating and transforming the read data according to the tag calculation logic to generate tag values ​​for each user; and Loading involves writing the calculated tag results into the user profile table in the data application layer.

[0048] The user profile tag generation system for recruitment provided in this application, on the one hand, constructs a layered data warehouse including a Data Detail Layer (DWD), a Data Aggregation Layer (DWS), and a Data Application Layer (DM). This allows complex tag calculation logic to be built on pre-processed and lightly summarized data, avoiding repetitive calculations on massive amounts of raw data. This simplifies ETL processing logic, reduces processing time and required CPU, memory, and other computing resources, and improves tag output efficiency. On the other hand, by calling an AI large-scale model through a tag intelligent mining module and integrating it with the needs of the recruitment field, the system can automatically and intelligently generate new user profile tags and their tag calculation logic based on time dimension expansion, user behavior clustering, and analysis of industry pain points. This changes the past single, passive mode that relied solely on manual demand submission, shortening the cycle from tag discovery to deployment. The system can proactively mine deep-level, high-value tags such as "user willingness to actively seek employment" and "candidate quality score," addressing industry pain points such as "difficulty in recruitment" and "low job search efficiency," and helping to enhance the support and value of user profiles for recruitment business.

[0049] To make it easier to understand, the following example illustrates the process of AI large-scale model intelligently mining user profile tags.

[0050] Figure 2 This is a schematic diagram illustrating the execution flow of an AI large-scale model in a user profile tag generation system for recruitment provided in an embodiment of this application. Figure 2 As shown, according to some embodiments of this application, optionally, the AI ​​large model can be specifically configured to perform the following steps S201 to S203.

[0051] S201: Identify statistical tags generated based on the first statistical period from existing user profile tag information, and based on the business meaning of the statistical tags, evaluate and generate the same type of statistical tags based on the second statistical period and their tag calculation logic.

[0052] In S201, the AI ​​big model can first scan and identify existing user profile tags in the user profile table of the data application layer, and locate statistical tags generated based on the first statistical period (such as the last day) from these tags, such as the number of applications submitted on the day, the number of job views on the day, and the number of logins on the day.

[0053] Next, the AI ​​model can analyze the business implications of various statistical tags. For example, AI model analysis might show that the number of job applications is a core indicator of a user's job-seeking intention. However, daily data fluctuates greatly, requiring a longer period to smooth out these fluctuations and reflect true intentions. Therefore, it's recommended to expand the analysis to include the number of applications submitted in the past 7 days and / or the past 30 days. As another example, AI model analysis might show that logging in is a prerequisite for using the app, but it contains little information. More attention should be paid to post-login behavior, such as browsing and applying for jobs. Therefore, monitoring daily login counts is sufficient to determine user activity, and aggregation value is limited; thus, there's no need to expand the analysis to include the number of logins in the past 7 days and / or the past 30 days.

[0054] Next, the AI ​​big model can automatically generate similar statistical tags based on the second statistical period (such as the last 7 days or the last 30 days) and their complete tag calculation logic, such as new tags "delivery count in the last 7 days" and / or "delivery count in the last 30 days" and their corresponding tag calculation logic.

[0055] S202: Identify multiple existing user profile tags that describe users with similar or identical behaviors, perform weighted combination or cluster analysis on multiple existing user profile tags, and generate new user profile tags that represent the comprehensive characteristics of users and their tag calculation logic.

[0056] In S202, the AI ​​big data model identifies multiple existing user profile tags describing similar or consensus-based user behaviors. For example, the AI ​​big data model finds that tags such as the number of applications submitted (apply_cnt), the number of chats with HR (chat_hr_cnt), and the number of resumes sent (send_resume_cnt) all represent users' proactive outreach to recruiting companies. Next, the AI ​​big data model can perform cluster analysis or weighted combination based on business rules on these multiple existing user profile tags representing similar or consensus-based behaviors to generate a new user profile tag that can characterize the user's comprehensive characteristics, along with its tag calculation logic. For example, in some examples, the new user profile tag generated by the AI ​​big data model that characterizes the user's comprehensive characteristics includes, but is not limited to, the user's proactive outreach intention score, the clarity of the user's job search intention, and / or the user's activity level. Taking the user's proactive outreach intention score as an example, its tag calculation logic can be: User's proactive outreach intention score = (Number of applications submitted in the last 7 days * 0.5 + Number of chats with HR in the last 7 days * 0.3 + Number of resumes sent in the last 7 days * 0.2) * 10.

[0057] In this way, deep mining through large AI models can overcome the limitations of human thinking, extract deeper user characteristics from scattered data, and generate more valuable derivative tags.

[0058] S203: Based on the industry pain point instructions input from the recruitment or job search end, locate multiple tags in the existing user profile tags that are related to solving the pain point as feature tags, and perform weighted combination of multiple feature tags to generate new user profile tags that quantify industry pain points and their tag calculation logic.

[0059] In S203, the AI ​​large model receives industry pain point instructions from business users or developers, addressing issues on the recruitment or job-seeking sides. For example, it can resolve the issue of inconsistent resume quality received by B-end companies or improve the job-seeking efficiency for C-end users.

[0060] Based on semantic understanding of pain point commands, the AI ​​big data model can locate multiple tags related to solving the pain point from existing user profile tags as feature tags. For example, for the industry pain point of resume quality, the AI ​​big data model locates tags such as years of work experience, education level, skill matching degree, and historical interview pass rate. Then, the AI ​​big data model can weight and combine these feature tags to construct a new user profile tag and its tag calculation logic that can directly quantify the industry pain point. For example, in some examples, the new user profile tag generated by the AI ​​big data model to quantify the industry pain point includes candidate quality score, candidate response willingness score, ability matching degree, and / or job application willingness. Taking the candidate quality score as an example, its tag calculation logic can be: Candidate Quality Score = (Years of Work Experience Score * 0.2 + Education Level Score * 0.2 + Skill Matching Degree * 0.3 + Interview Pass Rate * 0.3) * 100.

[0061] In this way, the AI ​​big data model can perform tag mining from three levels: time dimension expansion, user behavior clustering, and analysis of industry pain points on the recruitment and / or job-seeking sides. It can proactively discover deep and high-value tags such as "user willingness to actively seek help" and "candidate quality score", which can solve industry pain points such as "difficulty in recruitment" and "low efficiency in job seeking". This will help improve the support and value of user profiles for recruitment business.

[0062] Figure 3 Another structural block diagram of the user profile tag generation system for recruitment provided in this application embodiment. For example... Figure 3 As shown, according to some embodiments of this application, optionally, the user profile tag generation system 10 for recruitment may further include an intermediate table management module 140. The intermediate table management module 140 can communicate with the hierarchical data warehouse 110 and is used to create and maintain a data intermediate table for business users. The data structure of the data intermediate table originates from the user profile table in the data application layer. The data intermediate table can serve as a copy of the user profile table. The data intermediate table can serve as a data source directly accessed and used by downstream business users (such as precision marketing platforms, intelligent recommendation systems, and BI analysis tools).

[0063] The intermediate table management module 140 can be configured to: receive data structure change instructions from the business side; and, according to the data structure change instructions, add, disable, or delete fields in the intermediate data table, with the operation being independent of the structural changes in the user profile table itself.

[0064] Specifically, the intermediate table management module 140 can receive data structure change instructions from business stakeholders. These instructions can stem from changes in business requirements, such as needing to remove an unused tag field or enable a newly added tag field. Based on the received instructions, the intermediate table management module 140 can independently add, disable, or delete fields in the intermediate data table. This operation can be independent of structural changes to the user profile table within the data warehouse. That is, when a business needs to remove a tag, it only needs to modify the intermediate data table through the intermediate table management module 140, and the business application will immediately lose access to that field, achieving "second-level" removal. The user profile table within the data warehouse can retain the field for a period to ensure the needs of internal data analysis. Finally, based on internal policies, a decision is made on whether to delete the field from the user profile table, thus achieving the orderly removal of tags. In this way, the business side's change operations are restricted to the intermediate data table, which decouples the user profile tag data in the data warehouse from the business side's user profile tag data. This effectively avoids risks such as ETL task failure, data backtracking difficulties, and access control issues that may arise from directly modifying the core user profile table structure.

[0065] like Figure 3 As shown, according to some embodiments of this application, optionally, in order to ensure the reliability of the final output data, the user profile tag generation system 10 for recruitment may further include a tag quality monitoring module 150. The tag quality monitoring module 150 is communicatively connected to the hierarchical data warehouse 110 and can be used to perform overall quality monitoring and key field monitoring of the user profile table at preset intervals.

[0066] Overall quality monitoring can include monitoring the total amount of abnormal data and the number of fields with completely empty values. Monitoring key fields can include monitoring whether the distribution of enumeration values ​​in enumeration-type tags is abnormal, and monitoring the percentage of numerical tags whose values ​​exceed a preset threshold. The preset duration can be flexibly adjusted according to actual needs; this application does not impose any limitations on it.

[0067] Figure 4 This is a schematic diagram illustrating the execution flow of a tag quality monitoring module in a user profile tag generation system for recruitment provided in this application embodiment. Figure 4 As shown, in some specific embodiments, optionally, the label quality monitoring module 150 is specifically used to perform the following steps S401 to S404.

[0068] S401: Calculate the total amount of data in the user profile table for the day, and determine whether its year-on-year change rate with the total amount of data yesterday and / or month-on-month change rate with the total amount of data in the same period last week exceed the preset first threshold range; if it exceeds the threshold, it is determined that the total amount of data is abnormal.

[0069] Taking a preset duration of one day as an example, in step S401, the total data volume of the user profile table for that day, i.e., the total number of records, is calculated. This is then compared to the total data volume of the previous day using a year-on-year change rate calculation, and / or compared to the total data volume of the same period last week using a month-on-month change rate calculation. Next, it is determined whether the aforementioned change rates exceed a preset first threshold range. If they do, the total data volume is deemed abnormal. The size of the first threshold range can be flexibly adjusted according to actual circumstances; this application does not limit this. The detection in step S401 can effectively detect whether data has sharply decreased due to ETL task failure or surged due to data duplication.

[0070] S402: Traverse each field of the user profile table, calculate its null value rate, and determine whether the number of fields with a null value rate of 100% is greater than zero; if it is greater than zero, trigger an alarm.

[0071] In S402, each field in the user profile table can be iterated through to calculate its null value rate, which is the number of NULL values ​​in that field divided by the total amount of data. Then, it is determined whether the number of fields with a null value rate of 100% is greater than zero. If it is greater than zero, an alarm is triggered immediately. S402 detection can identify cases where the entire field data is lost due to data source anomalies or calculation logic errors.

[0072] S403: For a specified enumeration label, calculate the proportion of each enumeration value in the daily data, and determine whether the fluctuation range of the proportion of any enumeration value compared with the proportion in the same period of the previous year exceeds the preset second threshold range; if it exceeds the threshold, it is determined to be an abnormal distribution.

[0073] In S403, for a specified enumerated tag, such as user activity level, the proportion of each enumerated value in the daily data can be calculated. It is then determined whether the fluctuation range of the proportion of any enumerated value compared to the proportion in the same historical period exceeds a preset second threshold range. If it does, it is judged as an abnormal distribution. The size of the second threshold range can be flexibly adjusted according to actual conditions, and this application does not limit it. The detection through S403 can effectively identify sudden changes in data distribution caused by changes in business rules or errors in calculation logic.

[0074] S404: For a specified numeric label, count the number of records whose values ​​exceed the preset reasonable value range, and calculate the percentage of this number to the total data volume; if the percentage exceeds the preset third threshold, it is determined to be abnormal.

[0075] In S404, for a specified numerical tag, such as the number of deliveries in the last 7 days, the number of records whose values ​​exceed a preset reasonable range can be counted. Then, the percentage of this abnormal number to the total data volume is calculated. If this percentage exceeds a preset third threshold, it is determined to be abnormal. The size of the third threshold can be flexibly adjusted according to the actual situation, and this application does not limit it. The detection through S404 can effectively avoid false alarms caused by individual dirty data.

[0076] Thus, by conducting overall quality monitoring and key field monitoring of the user profile table, we can shift from passive discovery to proactive early warning, which helps ensure the accuracy and reliability of the profile data used by downstream businesses. Furthermore, by monitoring various detection methods, such as anomalies in total data volume, the number of completely null fields, whether the distribution of enumeration values ​​in enumeration-type tags is abnormal, and the percentage of numerical tags exceeding preset thresholds, we can quickly pinpoint whether the problem is an overall task failure, a missing data source, or a calculation logic error in a specific tag, thereby shortening troubleshooting and recovery time.

[0077] like Figure 3 As shown, according to some embodiments of this application, optionally, the user profile tag generation system 10 for recruitment may also include a tag usage frequency monitoring module 160, which is used to collect task configuration logs of the downstream user acquisition platform, interface call logs that provide profile data, and query logs of the downstream business database, aggregate and analyze the usage frequency of each user profile tag, and issue an alarm for user profile tags whose usage frequency is lower than a preset threshold.

[0078] The tag usage frequency monitoring module 160 can collect usage logs from downstream systems through multiple channels to understand the actual usage of each user profile tag in downstream businesses and assist in tag operation and maintenance decisions based on usage frequency data. Specifically, when operations or analysts create tasks on the user segmentation platform, they configure the tag conditions that depend on them. By collecting task configuration logs from downstream user segmentation platforms, the tag usage frequency monitoring module 160 can accurately determine which tags are used to generate business audience packages. When recommendation systems, search systems, etc., call user profile data through API interfaces, the tag usage frequency monitoring module 160 can record the request parameters of each call, thereby analyzing which tags are specifically included in the interface return value. For applications that directly query business databases (such as Elasticsearch, HBase), the tag usage frequency monitoring module 160 can extract the tag fields used in the query conditions or sorting conditions by parsing SQL or DSL query statements, thus obtaining the query logs of the downstream business database.

[0079] Next, the tag usage frequency monitoring module 160 can clean, parse, and aggregate logs collected from multiple channels, group them according to the names of various user profile tags, and count the total number of times each tag is used within a preset period, thus obtaining its usage frequency. Then, the tag usage frequency monitoring module 160 can compare the usage frequency of each tag with a preset threshold. For user profile tags whose usage frequency within the preset period is lower than the preset threshold, the tag usage frequency monitoring module 160 can automatically trigger an alarm. The preset period can be flexibly adjusted according to actual conditions, such as 7 days, 10 days, 30 days, or 60 days, etc., and this application does not limit it. The preset threshold can also be flexibly adjusted according to actual conditions, such as 0, 1, or 3, etc., and this application does not limit it.

[0080] Thus, by introducing the tag usage frequency monitoring module 160, the business value of tags can be quantified by the objective indicator of usage frequency, providing an objective basis for tag retirement decisions, thereby timely cleaning up invalid tags, saving data storage and computing costs, and avoiding resource waste.

[0081] Figure 5 This is another structural block diagram of the user profile tag generation system for recruitment provided in the embodiments of this application. For example... Figure 5 As shown, according to some embodiments of this application, the user profile tag generation system 10 for recruitment may optionally include a frequently used tag management module 170. The frequently used tag management module 170 can be used to manage the process of deactivating and reactivating user profile tags (or frequently used tags).

[0082] The frequently used tag management module 170 can disable or delete corresponding fields in the intermediate data table when user profile tags in the user profile table need to be taken offline, so that business users are unaware of it. That is, when a frequently used tag needs to be taken offline, the frequently used tag management module 170 can first disable or delete the corresponding field in the intermediate data table. This immediately prevents downstream business users from seeing or using the field, achieving instantaneous offline functionality without any awareness from the business users. Simultaneously, in the user profile table, the frequently used tag management module 170 can comment out the user profile tag generation logic and set the field values ​​to NULL, while retaining the complete field structure. This avoids high-risk operations such as directly executing ALTER TABLE DROP COLUMN, ensuring the stability of the core table structure and preparing for possible future re-enabling of tags.

[0083] When it needs to be reactivated, the common tag management module 170 can remove the disabled or deleted operations in the data intermediate table, making it visible to the business again, and release the generation logic of the annotated user profile table.

[0084] Thus, by introducing the commonly used tag management module 170, the heavyweight database schema change operation of tag deactivation / activation can be transformed into a lightweight, reversible state switching operation, reducing operational risks and helping to ensure the consistency of data backtracking.

[0085] like Figure 5 As shown, according to some embodiments of this application, optionally, the user profile tag generation system 10 for recruitment may further include a temporary tag management module 180, used to classify user profile tags related to activities or individual business as temporary tags in response to temporary tag marking instructions from developers. Specifically, for some user profile tags, after review, developers may consider some user profile tags to be time-sensitive, such as those associated with activities, like spring and autumn recruitment events in the recruitment industry, where the corresponding tags need to be removed after the event ends, or those with poor universality, only meeting the needs of a single business (such as operations, marketing, or product). Therefore, the temporary tag management module 180 can classify user profile tags related to activities or individual business as temporary tags in response to temporary tag marking instructions from developers. For temporary tags, the temporary tag management module 180 can create task code and a dedicated output table independent of the main user profile table generation task. Their data is not stored in the user profile table, but is provided to the business side after being merged with the user profile table data through data synchronization.

[0086] According to some embodiments of this application, optionally, the temporary tag management module 180 is also used to pause the independent task of the temporary tag when it needs to be taken offline, and retain the task code and output table structure of the independent task; when it needs to be reactivated, the independent task is reactivated.

[0087] Specifically, when a temporary tag needs to be taken offline, the temporary tag management module 180 only needs to pause the temporary tag's independent task. After the task is paused, data production stops, and the data in its dedicated output table also stops. This offline operation does not affect the user profile table generation task or other tags, and the task code and output table structure of the temporary tag are preserved. When it needs to be reactivated, the temporary tag management module 180 can simply reactivate the independent task, which can then resume execution from the pause point and restart data production. The entire process is quick and simple.

[0088] Thus, by introducing the temporary tag management module 180, the impact of temporary tags can be limited to themselves, achieving physical isolation from the core data in the user profile table. The deactivation and activation of temporary tags become state management of independent tasks, with lower risk and higher flexibility, adapting to short-term and ever-changing business needs.

[0089] The block diagrams of the systems exemplified in embodiments of this application have been described above, along with related aspects. It should be understood that each block or combination thereof in the block diagrams may be implemented by computer program instructions, by dedicated hardware performing a specified function or action, or by a combination of dedicated hardware and computer instructions. For example, these computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to form a machine such that these instructions, executed via such processor, enable the implementation of the function / action specified in each block or combination thereof in the flowcharts and / or block diagrams. Such a processor may be a general-purpose processor, a dedicated processor, a special-purpose application processor, or a field-programmable logic circuit.

[0090] The functional blocks shown in the structural block diagrams of this application can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc.; when implemented in software, they are programs or code segments used to perform the required tasks. Programs or code segments can be stored in memory or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. Code segments can be downloaded via computer networks such as the Internet or intranets.

[0091] It should be noted that this application is not limited to the specific configurations and processes described above or shown in the figures. The above descriptions are merely specific embodiments of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the described systems, devices, modules, or units can be referred to the corresponding processes in the method embodiments, and need not be repeated here. It should be understood that the scope of protection of this application is not limited thereto. Any person skilled in the art can conceive of various equivalent modifications or substitutions within the scope of the technology disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application.

Claims

1. A user profile tag generation system for recruitment, characterized in that, include: A layered data warehouse includes a data detail layer, a data aggregation layer, and a data application layer. The data detail layer stores user detail data after preprocessing the original user data, the data aggregation layer stores user behavior data after preprocessing the original user data, and the data application layer stores user profile tables. The tag intelligent mining module, which communicates with the hierarchical data warehouse, is used to receive user profile tag mining instructions and existing user profile tag information in the user profile table, and to call the AI ​​big model to perform the following operations: Based on the user profile tag mining instructions and existing user profile tag information, user profile tags are mined by expanding the time dimension, clustering user behavior, and analyzing industry pain points on the recruitment and / or job search sides, so as to generate new user profile tags and their corresponding tag calculation logic. The tag ETL generation module, which is communicatively connected to the hierarchical data warehouse and the tag intelligent mining module, is configured as follows: After the new user profile tags and tag calculation logic are approved, an ETL processing logic that depends on user behavior data in the data aggregation layer and / or user detail data in the data detail layer is constructed based on the new user profile tags and tag calculation logic. The calculated user profile tag data is then written into the user profile table of the data application layer.

2. The system according to claim 1, characterized in that, The system also includes: The intermediate table management module, which is communicatively connected to the hierarchical data warehouse, is used to create and maintain intermediate data tables for business users. The data structure of the intermediate data tables is derived from the user profile table. The intermediate table management module is configured to: receive data structure change instructions from business users; and, according to the data structure change instructions, add, disable, or delete fields in the intermediate data tables, with the operation being independent of the structural changes in the user profile table itself.

3. The system according to claim 1, characterized in that, The large AI model is specifically configured to perform the following operations: Identify statistical tags generated based on the first statistical period from existing user profile tag information, and evaluate and generate similar statistical tags and their tag calculation logic based on the business meaning of the statistical tags based on the second statistical period. Identify multiple existing user profile tags that describe users’ similar or identical behaviors, perform weighted combination or cluster analysis on the multiple existing user profile tags, and generate new user profile tags that represent the comprehensive characteristics of users and their tag calculation logic. Based on the industry pain point instructions input from the recruitment or job search end, multiple tags related to solving the pain point in the existing user profile tags are located as feature tags. The multiple feature tags are then weighted and combined to generate new user profile tags that quantify the industry pain point and their tag calculation logic.

4. The system according to claim 3, characterized in that, New user profile tags that characterize comprehensive user features include user willingness to actively seek help, clarity of user job search intentions, and / or user activity; new user profile tags that quantify industry pain points include candidate quality score, candidate willingness to respond score, ability matching degree, and / or willingness to join the company.

5. The system according to claim 1, characterized in that, The system also includes: The tag quality monitoring module is connected to the hierarchical data warehouse and is used to perform overall quality monitoring and key field monitoring on the user profile table at preset intervals. The overall quality monitoring includes monitoring the total amount of abnormal data and the number of fields with all empty values; the key field monitoring includes monitoring whether the distribution of enumeration values ​​in enumeration-type labels is abnormal and monitoring the proportion of numerical labels whose values ​​exceed a preset threshold.

6. The system according to claim 5, characterized in that, The label quality monitoring module is specifically used to perform the following operations: Calculate the total amount of data in the user profile table for the day, and determine whether its year-on-year change rate with the total amount of data yesterday and / or its month-on-month change rate with the total amount of data in the same period last week exceed the preset first threshold range; if it exceeds the threshold, it is determined that the total amount of data is abnormal. Iterate through each field in the user profile table, calculate its null value rate, and determine whether the number of fields with a null value rate of 100% is greater than zero; if it is greater than zero, trigger an alarm. For a given enumeration label, calculate the proportion of each enumeration value in the daily data, and determine whether the fluctuation range of the proportion of any enumeration value compared with the proportion in the same period of the previous year exceeds the preset second threshold range; if it exceeds the threshold, it is determined to be an abnormal distribution. For a given numeric label, count the number of records whose values ​​exceed a preset reasonable range, and calculate the percentage of that number to the total data volume; if the percentage exceeds a preset third threshold, it is considered abnormal.

7. The system according to claim 1, characterized in that, The system also includes: The tag usage frequency monitoring module is used to collect task configuration logs from downstream user segmentation platforms, interface call logs that provide profile data, and query logs from downstream business databases. It aggregates and analyzes these logs to obtain the usage frequency of each user profile tag and issues alerts for user profile tags whose usage frequency is lower than a preset threshold.

8. The system according to claim 2, characterized in that, The system also includes: The common tag management module is used to disable or delete the corresponding field in the data intermediate table when a user profile tag in the user profile table needs to be taken offline, so that the business side cannot be aware of it; and to comment out the generation logic of the user profile tag in the user profile table and set the field value to NULL while retaining the field structure; when it needs to be re-enabled, the disabling or deletion operation in the data intermediate table is lifted, and the commented generation logic in the user profile table is unblocked.

9. The system according to claim 1, characterized in that, The system also includes: The temporary tag management module is used to respond to developers' temporary tag marking instructions and classify user profile tags related to activities or individual business as temporary tags; The temporary tags are not stored in the user profile table and have task codes and output tables that are independent of the main generation task of the user profile table. The data of the temporary tags is provided to the business side after being merged with the data of the user profile table through data synchronization.

10. The system according to claim 9, characterized in that, The temporary tag management module is also used to pause the independent task of the temporary tag when it needs to be taken offline, while retaining the task code and output table structure of the independent task; and to re-enable the independent task when it needs to be reactivated.

Images