Method and device for recommending post to post changing personnel and electronic equipment

Abstract

The invention relates to a method for recommending a post to a post changing person, and the method comprises the steps: obtaining recruitment associated data of the post changing person from a human resource system, and carrying out the structural processing of the recruitment associated data, and obtaining employee portrait feature data and post resume data; converting the post resume data into a working experience text; generating an evaluation text based on post information and salary data in the post resume data; inputting the post resume data into the trained skill prediction model, outputting skill keywords related to the post resume data, and further generating a skill label text; according to the employee portrait feature data, the work experience text, the evaluation text and the skill label text, generating a resume of the post changing personnel; and performing multi-dimensional feature vector matching on the resume and the recruitment post demand, determining a to-be-recommended post, and recommending the to-be-recommended post to the post changing personnel. According to the method, a large amount of post change personnel data can be vitalized, and high-quality candidate resources are provided for recruitment services.

Description

Technical Field

[0001] This invention relates to the field of recommendation system technology, and in particular to a method, apparatus, electronic device, and computer program product for recommending job positions to personnel undergoing job changes. Background Technology

[0002] Currently, enterprise human resource management systems typically accumulate a large amount of historical data on personnel changes, including departing employees and current employees slated for job transfers. This data usually encompasses recruitment-related information for these personnel and has significant talent value.

[0003] However, under the current management system, once employees leave their original positions or employers, their historical skills and experience data often lack systematic utilization, resulting in the idleness or direct loss of potential human resource value, thus causing significant losses to enterprises or human resource service agencies.

[0004] Meanwhile, most existing job matching services rely on resumes filled out by job seekers or make recommendations based on single job matching rules, lacking in-depth analysis and processing of existing historical data. This approach not only limits the accuracy of job recommendations but also reduces the efficiency for recruiters to obtain high-quality candidate resumes. Summary of the Invention

[0005] In view of this, embodiments of this application provide a method, apparatus, electronic device, and computer program product for recommending job positions to personnel undergoing job changes, in order to solve at least one technical problem.

[0006] This application provides a method for recommending jobs to personnel undergoing job changes, comprising: performing structured processing on recruitment-related data of personnel undergoing job changes obtained from a human resources system to obtain employee profile feature data and job resume data; converting the job resume data into work experience text described in natural language; generating evaluation text reflecting career development based on job information and salary data in the job resume data; inputting the job resume data into a trained skill prediction model to output skill keywords related to the job resume data, thereby generating skill tag text; generating a job application resume for the personnel undergoing job changes based on the employee profile feature data, work experience text, evaluation text, and skill tag text; and performing multi-dimensional feature vector matching between the job application resume and the job requirements to determine the recommended job and recommend it to the personnel undergoing job changes.

[0007] The job recommendation method described above generates a salary-job time series based on job information and salary data in job resume data, and then generates an evaluation text reflecting career development. The salary-job time series is a serialized data set that is arranged in chronological order and contains both job information and salary data, used to show the changing trends of salary and job.

[0008] The job recommendation method described above further includes: extracting salary nodes from the salary job time series and calculating salary change values, and comparing the salary change values ​​with industry standard salaries to determine salary increases; matching the salary job time series with a career growth trajectory template to determine job promotion speed; and generating self-evaluation text and / or peer evaluation text reflecting the job stability and learning ability of personnel changing jobs based on the salary increase and promotion speed.

[0009] The job recommendation method described above, which transforms the job resume data into work experience text described in natural language, includes: converting target fields in the job resume data into semantic tags through field mapping, and generating an input template by combining them in a preset order; the target fields include one or more of the following: job title, service object, and function; generating a natural language description text that conforms to the recruitment context based on the semantic relationship of the input template, in which the description text highlights job responsibilities and achievements; and obtaining the work experience text by performing output optimization processing on the description text, wherein the optimization processing includes one or more of the following: standardizing sentence structure and terminology, retaining quantitative indicators, and enhancing the readability of grammatical structure.

[0010] The job recommendation method described above includes the following training method for the skill prediction model: collecting job descriptions and job skill information from massive resumes and job postings as pre-training data, and preprocessing the pre-training data; inputting the pre-training data into the model to be trained, and performing pre-training through vectorization mapping until a preset training index is reached; collecting task-specific data with job-skill correspondence annotations, constructing fine-tuning training data including input samples and output samples, inputting the fine-tuning training data into the pre-training model for fine-tuning until a preset training index is reached; deploying the fine-tuned model to the recommendation system, and performing quantitative compression and / or periodic maintenance according to business needs.

[0011] The job recommendation method described above, which matches the job applicant resumes with the job requirements using multi-dimensional feature vectors, includes: generating multi-dimensional feature vectors for both the job applicant resumes and the job requirements; calculating a weighted similarity based on the multi-dimensional feature vectors to determine one or more job positions to be recommended; and recommending these job positions to the applicants whose positions have changed.

[0012] The job recommendation method described above involves structuring the recruitment-related data by: matching the original job titles in the recruitment-related data with standard job titles in a preset job dictionary to output standardized job titles; and / or mapping the original industry description information in the recruitment-related data to industry categories to output target industry tags that conform to the industry classification standards of the recruitment system; and / or using a rule engine to identify and delete sensitive information fields that are irrelevant to job recommendations.

[0013] According to another aspect of this application, a job recommendation device for personnel undergoing job changes is proposed, comprising: a structured processing module for performing structured processing on recruitment-related data of personnel undergoing job changes obtained from a human resources system to obtain employee profile feature data and job resume data; an experience text generation module for converting the job resume data into work experience text described in natural language; an evaluation text generation module for generating evaluation text reflecting career development based on job information and salary data in the job resume data; a skill text generation module for inputting the job resume data into a trained skill prediction model, outputting skill keywords related to the job resume data, and then generating skill tag text; a resume generation module for generating a job application resume for the personnel undergoing job changes based on the employee profile feature data, work experience text, evaluation text, and skill tag text; and a job recommendation module for performing multi-dimensional feature vector matching between the job application resume and the job requirements to determine the job to be recommended and recommending it to the personnel undergoing job changes.

[0014] According to another aspect of this application, an electronic device is proposed, including a processor and a memory, wherein a set of computer program instructions is stored on the memory, and the method for recommending positions to personnel undergoing job changes as described above is implemented when the processor executes the set of computer program instructions in the memory.

[0015] According to another aspect of this application, a computer program product is proposed, which includes a set of computer program instructions that, when executed by a processor, implement the method of recommending positions to personnel undergoing job changes as described above.

[0016] This method cleanses, structures, and semantically processes multi-source data, including historical job positions and salary records of individuals who have changed jobs, to generate complete and naturally expressed resumes. These resumes are then used for job recommendations, effectively utilizing a large amount of data on individuals with changing jobs and providing high-quality candidate resources for recruitment. Furthermore, multi-dimensional matching calculations improve the accuracy of job recommendations for individuals with changing jobs. Attached Figure Description

[0017] The preferred embodiments of the present invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0018] Figure 1 This is a flowchart of a method for recommending job positions to personnel undergoing job changes, according to one embodiment of this application.

[0019] Figure 2 This is a flowchart of the method for generating work experience text in step S120.

[0020] Figure 3 This is a flowchart of a method for generating evaluation text based on information job time series according to an embodiment of this application.

[0021] Figure 4 This is a flowchart illustrating a prediction model training method according to an embodiment of this application.

[0022] Figure 5 This is a flowchart of a pre-training and fine-tuning method for a prediction model according to an embodiment of this application.

[0023] Figure 6 This is a schematic diagram of a job recommendation device for personnel undergoing job changes, according to one embodiment of this application.

[0024] Figure 7 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] In the following detailed description, reference can be made to the accompanying drawings, which form part of this application and illustrate specific embodiments of the present application. In the drawings, similar reference numerals describe substantially similar components in different figures. Specific embodiments of the present application are described in sufficient detail below to enable those skilled in the art to implement the technical solutions of the present application. It should be understood that other embodiments may also be utilized, or structural, logical, or electrical changes may be made to the embodiments of the present application.

[0027] The term "personnel undergoing job changes" in this application refers to individuals whose job change records are present in the company's human resource management system. This includes former employees who have left their original employers, as well as current employees who are considering or have already completed job adjustments within their current employers. These individuals typically provide basic information, educational background, work experience, contract records, and reasons for job changes to the human resource management system (such as an operating system) during their employment.

[0028] The job recommendation process in this application only applies to employees whose job changes have been agreed to in the OS system through explicit agreement to the user agreement and privacy terms. With the employee's consent, the system can analyze, process, and match job information based on their historical data retained during their employment, thereby achieving accurate job recommendations. Since the data usage is based on the user's prior authorization and follows principles such as information anonymization, minimal use, and limited purpose, it does not constitute an infringement of employee privacy, nor does it involve the unauthorized acquisition or use of personal information.

[0029] Figure 1 This is a flowchart illustrating a method for recommending job positions to personnel undergoing job changes, according to one embodiment of this application. Figure 1 As shown, the method includes:

[0030] S110 will retrieve recruitment-related data of personnel whose positions have changed from the human resources system and perform structured processing to obtain employee profile feature data and job history data.

[0031] S120, the job resume data is converted into work experience text described in natural language;

[0032] S130, Generate an evaluation text reflecting career development based on job information and salary data in job resume data;

[0033] S140, input the job resume data into the trained skill prediction model, output skill keywords related to the job resume data, and then generate skill tag text;

[0034] S150, Generate the job application resume of the employee whose position has changed based on the employee profile feature data, work experience text, evaluation text and skill tag text;

[0035] S160, perform multi-dimensional feature vector matching between the job applicants' resumes and the job requirements to determine the positions to be recommended and recommend them to the job applicants.

[0036] In step S110, the structured processing includes: forming employee profile feature data from basic information, educational background, work experience, and other information related to job matching; and organizing job titles, service targets, functional categories, salary records, and reasons for leaving, arranged in chronological order, into job resume data. Through structured processing, redundancy and inconsistencies in the original data can be eliminated while retaining key information, providing a unified and standardized input data format for subsequent text generation and matching calculations, thereby providing high-quality input for large-scale models.

[0037] For example, recruitment-related data includes:

[0038] Basic Information: Name: Xiaoming; Gender: Male; Age: 30; Education: Master's Degree; Current Residence: No. xxxx, Zhangdong Road, Pudong New Area, Shanghai;

[0039] Education: Start Date: 2013-09 End Date: 2016-07 Degree: Master's Degree School: Shanghai University Major: Computer Networks;

[0040] Start Date: 2009-07 End Date: 2013-07 Education: Bachelor's Degree University: Shanghai University Major: Circuits and Systems;

[0041] Work experience:

[0042] Start Date: August 2022 End Date: December 2024 Company: C Industry: Internet Position: Senior Java Developer;

[0043] Start Date: July 2020 End Date: July 2022 Company: B Industry: Internet Position: Senior Java Developer

[0044] Start Date: September 2016 End Date: July 2020 Company: A Industry: Finance Position: Junior Development Engineer;

[0045] Current service information:

[0046] Contract start date: January 2025; Employer: D; Position: Java Supervisor;

[0047] Reason for leaving: Voluntary resignation.

[0048] In step S120, by mapping fields to semantic tags, information such as job title, target audience, and main responsibilities are combined according to common recruitment industry expressions, transforming previously scattered fields into fluent sentences that conform to recruitment reading habits. This process converts structured information into more easily understood and evaluated text descriptions, thereby improving the readability and persuasiveness of subsequent job matching.

[0049] In step S130, career development trends can be analyzed using job information and salary data, such as steady salary increases and promotion speed. These trends are then combined to generate evaluative text that reflects the patterns of career growth. Career development trends can reveal the career development status of individuals undergoing job changes over time, providing recruiters with a more intuitive reference for their abilities.

[0050] In step S140, the skill prediction model combines information such as job title and job description from the job resume data to infer the skills possessed by the employee in actual work, thereby supplementing skill information not explicitly listed in the resume and making the generated resume more complete and accurate. Furthermore, the skill tag text provides precise guidance for subsequent job recommendations.

[0051] Steps S120-S140 aim to generate high-quality resumes directly applicable to recruitment scenarios based on employee profile feature data and job experience data. Specifically, structured information from the data layer is semantically parsed and reorganized into natural language text with complete work experience, skill highlights, strengths descriptions, and scenario-based examples, thus forming a "recruitment-grade resume." This process utilizes the contextual understanding and language generation capabilities of the generative model, which not only preserves the authenticity and consistency of information but also extracts job matching points and professional advantages. This significantly improves the readability, persuasiveness, and job suitability of the generated resume, providing a reliable foundation for subsequent accurate job recommendations.

[0052] In step S150, the job application resume may include basic personal information, job experience description, career development evaluation and skill tags, which is similar in format to the resume provided by the job seeker and can perfectly match the current job recommendation system without any adjustments.

[0053] In step S160, job applicants' resumes are matched with job postings using multi-dimensional feature vector matching. Indicators such as skill similarity, job similarity, industry relevance, and geographic matching are combined to calculate a matching score between the job and the candidate. Based on this score, the recommended job postings are determined, and the recommendations are ultimately pushed to the applicants whose positions have changed. This approach allows for a comprehensive evaluation of the matching score across multiple dimensions, thereby improving the accuracy and effectiveness of job recommendations.

[0054] Through the above implementation methods, this approach can clean, structure, and semantically process multi-source data such as historical job positions and salary records of personnel undergoing job changes, thereby generating complete and naturally expressed job application resumes. These resumes can then be used for job recommendations, effectively utilizing a large amount of data on personnel undergoing job changes and providing high-quality candidate resources for recruitment. Furthermore, recommending suitable positions based on multi-dimensional matching calculations improves the accuracy of job recommendations for personnel undergoing job changes.

[0055] According to one embodiment of this application, structuring recruitment-related data includes:

[0056] Match the original job titles in the recruitment-related data with the standard job titles in the preset job dictionary, and output standardized job titles; and / or

[0057] Map the original industry description information in the recruitment-related data to industry categories, and output target industry tags that conform to the industry classification standards of the recruitment system; and / or

[0058] The rule engine is used to identify and delete sensitive information fields that are irrelevant to job recommendations.

[0059] Matching the original job titles in the recruitment-related data with standard job titles in a pre-defined job dictionary can resolve the inconsistency between the job titles recorded in the human resources system and the job dictionary in the recruitment system. For example, natural language processing and machine learning algorithms can be used to identify the core functions and hierarchical information in the job titles, such as standardizing the job title "Java Supervisor" to "Java Development Supervisor" on the recruitment end, to ensure the accuracy of subsequent job matching calculations.

[0060] The original industry description information in the recruitment-related data is categorized and mapped to target industry tags. For example, the industry of the employing unit D of the personnel whose positions have changed in the OS system is analyzed as "Internet / IT / Electronics / Telecommunications > Computer Software", etc., to achieve the unification and standardization of industry fields.

[0061] A rules engine is used to identify and delete sensitive information fields irrelevant to job recommendations to meet data privacy and compliance requirements. This sensitive information includes employee ID numbers, detailed residential information, and emergency contact information. Residential information is retained only at the city level and can be kept as information about the job seeker's preferred city, ensuring privacy without affecting the effectiveness of job recommendations.

[0062] Figure 2 This is a flowchart illustrating the method for generating work experience text in step S120. (Example) Figure 2 As shown, the method includes:

[0063] S121, the target fields in the job resume data are converted into semantic tags through field mapping, and the input template is generated by combining them in a preset order. The target fields include one or more of the following: job name, service object, and function.

[0064] The target field can include one or more of the following: job title, service recipient, and job function. For example, the field "Job Title: HR Outsourcing - Payroll Specialist" is mapped to the tag "Payroll Specialist," and the job function information is mapped to a pre-defined recruitment semantic tagging system. Subsequently, the semantic tags are combined in a pre-defined order to generate an input template, ensuring the structured and readable nature of the input information.

[0065] S122, The input template is used to generate a natural language description text that conforms to the recruitment context based on semantic association, and the description text highlights the job responsibilities and achievements;

[0066] In the above process, the correspondence between job responsibilities and results can be identified, and quantitative performance indicators can be embedded in the sentence structure, so that the generated description not only includes the scope of responsibilities, but also reflects performance. For example, "Calculate the salaries of 1000+ employees and handle social security and housing provident fund" can be transformed into the description "Responsible for the salary calculation, social security and housing provident fund payment and payroll report preparation of 1000+ employees, ensuring 100% accuracy of monthly salary payment, and optimizing the calculation process to improve efficiency by 20%".

[0067] S123, the descriptive text is optimized to obtain work experience text. The optimization includes one or more of the following: standardizing sentence structure and terminology to conform to common expressions used in the recruitment industry; retaining quantitative indicators to enhance information credibility; and enhancing the readability of grammatical structure. After the above processing, a work experience text that conforms to recruitment usage habits and highlights the value of the position can be obtained.

[0068] For example, the job resume data is as follows: Position: HR Outsourcing - Payroll Specialist; Client: An Internet Company; Responsibilities: Calculating salaries for 1000+ employees and handling social security and housing provident fund.

[0069] The aforementioned job resume data is input into a trained large model, which generates natural language descriptions that conform to the recruitment context based on semantic reasoning, and outputs them as work experience text.

[0070] The output work experience text is as follows: During my time as a payroll specialist at an internet company (HR outsourcing), I was responsible for the payroll calculation, social security and housing fund payment, and payroll report preparation for 1000+ employees, ensuring 100% accuracy in monthly payroll payments, and optimizing the calculation process to improve efficiency by 20%.

[0071] According to one embodiment of this application, a salary and job time series is generated based on job information and salary data in job resume data, and then an evaluation text reflecting career development is generated. The salary and job time series is a serialized data set that is arranged in chronological order and contains both job information and salary data, and is used to show the changing trends of salary and job.

[0072] Salary job time series are generated through the following steps:

[0073] 1. Obtain job information and salary data of employees who have changed positions during their employment, including but not limited to the job title, time of change, salary before and after the change, and year-end bonus data for each stage.

[0074] 2. The above job information and salary data are serialized according to time sequence to generate a salary-job time series. In the data structure of this salary-job time series, each time node is associated with a job name and a corresponding salary value, thus intuitively reflecting the correspondence between job adjustments and salary changes.

[0075] Among them, the salary and job time series not only retains the job and salary characteristics of employees who have changed jobs at different stages, but also forms a highly summarized and readable description of career development, providing data support and semantic information foundation for subsequent job matching and accurate recommendation.

[0076] By applying contextual reasoning to salary and job time series data through a large-scale model, self-evaluation and / or peer evaluation can be generated, amplifying employee strengths and value. See below for details. Figure 3 content:

[0077] Figure 3 This is a flowchart of a method for generating evaluation text based on job posting time series according to an embodiment of this application, as shown below. Figure 3 As shown, the method includes:

[0078] S310, extract salary nodes from the salary job time series and calculate salary change values, and compare the salary change values ​​with industry standard salaries to determine the salary increase;

[0079] The time-point information includes basic salary, year-end bonus, and other quantifiable compensation elements before and after the job change. For adjacent time points, the corresponding salary change value is calculated, and based on the industry salary database or preset industry salary statistics, the salary change value is compared with the industry standard salary level to determine the salary increase of the employee at each stage of the job change.

[0080] S320, Match the salary job time series with the career growth trajectory template to determine the job promotion speed;

[0081] The career development trajectory template includes various typical job promotion paths and their corresponding time cycles, such as the average promotion time from junior to intermediate and from intermediate to senior positions. Through matching analysis, the promotion speed of employees changing jobs can be determined and compared with the promotion cycle of similar positions in the industry to identify their career development pace and potential.

[0082] S330, Based on the salary increase and promotion speed, generate self-evaluation text and / or others' evaluation text reflecting the job stability and learning ability of employees whose positions have changed.

[0083] After determining salary increases and promotion speed, these two indicators are combined to generate natural language text reflecting the job stability and learning ability of employees undergoing job changes. This natural language text can be output as a self-evaluation text and / or an evaluation text from others, depending on the usage scenario. The self-evaluation text can be written from a first-person perspective, highlighting the personal efforts, skill enhancements, and accumulated achievements of the employee undergoing job changes; the evaluation text from others, written from a third-person perspective, emphasizes external recognition and impressions of the employee's abilities and value, thus forming a multi-faceted presentation of their career development.

[0084] Compared to relying solely on resumes filled out by job seekers, this application generates additional evaluation texts based on job experience data and salary time-series analysis, allowing job recommenders to gain a more authentic and comprehensive understanding of the professional abilities and overall qualities of job seekers from a third-party perspective. This improves the accuracy of job matching and enhances the reference value for recruiters' decisions.

[0085] For example, the job information and salary data in the job resume are as follows: Xiaoming joined the company in 2020 with a two-year contract as a junior development engineer. He later signed a renewal contract, and his position was adjusted to senior development engineer with a monthly salary of 20,000 yuan. He then signed a revised contract with a monthly salary of 30,000 yuan. His annual bonus is approximately five months' basic salary.

[0086] Inputting the above information into the large model and utilizing contextual reasoning capabilities, the following evaluation text is output:

[0087] Self-Evaluation: Since joining the company in 2020, I have consistently maintained a proactive work attitude, continuously improving my professional skills and overall capabilities. From a junior development engineer to a senior development engineer, I have steadily progressed through in-depth technical expertise and project practice. During my time as a junior development engineer, I solidified my fundamental development skills and efficiently completed all assigned tasks, laying a strong foundation for future development. After being promoted to senior development engineer, I proactively undertook more complex technical research and development and project challenges, resolving numerous technical difficulties with my strong professional abilities and helping the team achieve key project goals. In my work, I always hold myself to high standards, emphasize teamwork, actively share technical experience, and promote overall team efficiency. The salary increase from 20,000 to 30,000 RMB per month, along with an annual bonus of approximately five months' basic salary, reflects the recognition of my work performance and abilities, and motivates me to continuously break through my limitations and create greater value for the company.

[0088] Evaluation from others: Since joining the company in 2020, Xiaoming's growth trajectory and work performance have been remarkable. His promotion from junior development engineer to senior development engineer demonstrates his rapid growth and deep accumulation of technical expertise. In his work, he has demonstrated strong problem-solving and project execution abilities, playing a key role in numerous important projects and providing strong support for the smooth progress of the team. The company's salary adjustment and year-end bonus fully reflect its high recognition of his work performance and professional capabilities. He actively collaborates within the team, is willing to share, and is an indispensable member of the team. His professionalism and work ability have been widely praised by his colleagues.

[0089] Figure 4 This is a flowchart illustrating a prediction model training method according to an embodiment of this application. Figure 4 As shown, the method includes:

[0090] S410, collect job descriptions and job skills information from a large number of resumes and job postings as pre-training data, and preprocess the pre-training data;

[0091] From a large amount of historical resume data and job posting information, job description fields and job skill fields related to skills are extracted to form a pre-training dataset. To ensure the effectiveness of subsequent training, preprocessing operations are performed on the pre-training data. The preprocessing includes: standardizing the field format; normalizing synonymous skill terms; and deleting invalid fields that are irrelevant to job skills to reduce training noise.

[0092] S420, The pre-training data is input into the model to be trained, and pre-training is performed through vectorization mapping until the preset training index is reached.

[0093] Pre-processed pre-training data is input into the model to be trained. Vectorization mapping technology is used to map the text information into a vector space, and parameter training is performed in this space. During the training process, the convergence of the model is determined by monitoring preset training metrics, which may include one or more of the following: TopN accuracy, correlation error threshold, and human reasonableness score.

[0094] S430: Collect task-specific data with job-skill correspondence labels, construct fine-tuning training data including input samples and output samples, and input the fine-tuning training data into the pre-trained model for fine-tuning until the preset training index is reached.

[0095] To better align with the business needs of job skill prediction tasks, task-specific data with clear job-skill correspondence labels is collected. A fine-tuning training set with one-to-one correspondence between input and output samples is constructed. The fine-tuning data is then input into the pre-trained model for parameter fine-tuning to enhance the model's predictive ability in specific business scenarios. The same or similar indicators as those in the pre-training stage are used as the criteria for judging fine-tuning convergence.

[0096] S440 deploys the finely tuned model to the recommendation system and performs quantitative compression and / or periodic maintenance according to business needs.

[0097] The finely tuned model can be deployed to the recommendation system to predict skills based on job resume data. The model can also be quantized and compressed according to business needs to reduce the computational resource consumption in the deployment environment. At the same time, a regular maintenance mechanism can be set to periodically update or retrain the model parameters to ensure that the prediction results are consistent with the latest business data.

[0098] Figure 5 This is a flowchart illustrating a pre-training and fine-tuning method for a prediction model according to an embodiment of this application. Figure 5 As shown, the training process for the skill prediction model is as follows.

[0099] (1) Data Acquisition and Preparation

[0100] The training requires basic data from multiple data sources, including: a massive amount of resume information accumulated within the company and job postings published by clients.

[0101] These data contain fields such as job descriptions, job functions, and skill requirements, providing the model with rich semantic information. For example, a job description might state "responsible for microservice development and performance optimization," and the model can learn that "microservices" and "performance optimization" are skills related to this position.

[0102] Data acquisition and preparation provide the model with comprehensive and diverse corpora, ensuring that the model can learn the relationships between different jobs and skills, thereby improving prediction accuracy.

[0103] (2) Data cleaning and preprocessing

[0104] Process the raw data, including deleting records that are missing key fields (such as resumes that lack skill descriptions);

[0105] Standardize field naming (e.g., unify "Java Development" and "Java Engineer" into "Java Development Engineer");

[0106] Normalize synonyms (e.g., unify "JVM tuning" and "Java Virtual Machine optimization");

[0107] Remove fields that are irrelevant to skill prediction (such as personal interests and address information).

[0108] The preprocessed data is converted into numerical vectors through vectorization mapping (embedding) and stored in the pre-training dataset.

[0109] (3) Pre-training phase

[0110] In the pre-training phase, the training objectives and optimization strategies are first set, for example:

[0111] Training objective: To maximize the model's prediction accuracy in resume-skill matching;

[0112] Optimization strategy: Use the cross-entropy loss function to optimize the model parameters using gradient descent; set the learning rate to 0.001, the batch size to 128, and the number of training epochs to 10.

[0113] During training, performance metrics are used to determine whether the model's performance meets the requirements, for example:

[0114] A Top5 accuracy of ≥85% means that there is at least an 85% probability that the true skill of the sample will appear in the top five skills predicted by the model.

[0115] An average correlation error of ≤0.15 indicates a small correlation error between predicted skills and actual skills.

[0116] If the reasonableness score is ≥4 / 5, the prediction results will be evaluated by human resources experts to determine whether they are reasonable.

[0117] If the metrics do not meet the requirements, the model can be retrained by supplementing with high-quality sample data to improve its generalization ability on real business data.

[0118] (4) Fine-tuning stage

[0119] The fine-tuning phase uses task-specific data with a clear job-skill correspondence, for example:

[0120] Position: Java Development Engineer → Skills: Java, Microservices, MyBatis

[0121] The task data is constructed into input-output sample pairs and fed into the pre-trained model for targeted optimization training. The fine-tuning effect is also evaluated based on the aforementioned performance metrics.

[0122] By fine-tuning the model, we ensure that the predictions are more accurate for specific tasks or business scenarios.

[0123] (5) Model Deployment and Application

[0124] The finely tuned model can be deployed to the recruitment recommendation system. If necessary, quantization compression can be performed to reduce inference resource consumption (e.g., converting floating-point parameters to 8-bit integers). A regular maintenance strategy should also be implemented to ensure the model maintains stable performance even when data distribution changes.

[0125] For example, for employee Xiaoming, his past work experience is input into the model, and the model predicts that the top five key skills are: "Java", "Microservices", "MyBatis", "Performance Tuning", and "JVM".

[0126] According to one embodiment of this application, performing multi-dimensional feature vector matching between the job resume and the job requirements includes:

[0127] The job applicants' resumes and job postings are each used to generate multi-dimensional feature vectors. A weighted similarity calculation is performed on these feature vectors to determine one or more job postings to be recommended. These recommended job postings are then recommended to the applicants who are changing jobs. The multi-dimensional feature vectors include one or more of the following: industry, position, desired work location, skills, salary requirements, age, education level, and years of work experience. Specifically:

[0128] First, analyze the existing job requirements of enterprises in their recruitment process, extract core requirement tags such as industry, position, expected work location, skills, salary requirements, age, education, and years of work experience, and transform them into multi-dimensional feature vectors.

[0129] Then, the job application resumes are analyzed, extracting relevant tags from profile feature data, evaluation text data, historical work experience, skill descriptions, expected salary, and job intentions, and generating multi-dimensional feature vectors consistent with job requirements. This design allows job seeker information to be compared with job requirements in the same feature space, which helps improve matching accuracy while taking into account the utilization of diverse information in resumes.

[0130] Next, the feature vectors of job requirements and job application resumes are matched. Weights are assigned to each feature dimension, and the degree of matching for each dimension is calculated using a pre-defined similarity function, such as skill similarity, salary requirement matching, and desired work location matching. Then, the matching results for each dimension are weighted and summed to obtain the final job matching score. The job matching score can be calculated using the following formula:

[0131]

[0132] Where TotalSimilarity is the job matching score, f i (a i ,b i Let w be the similarity of the i-th dimension feature. i Let a be the weight of the i-th dimension feature. i Let b be the i-th feature value in the multi-dimensional feature vector generated based on the job requirements. i Let be the i-th feature value in the multi-dimensional feature vector generated from the job applicant's resume.

[0133] Finally, based on the calculated job matching scores, jobs with scores above a preset threshold (e.g., 90 points) are selected from highest to lowest and recommended as candidate jobs to job seekers. This design allows the system to filter out jobs that highly match the job seeker's preferences, facilitating accurate recommendations and improving the efficiency and reliability of job matching.

[0134] According to one embodiment of this application, when the employee whose job is being changed is a departing employee, the departure time and a preset duration are compared, and based on the comparison results, urgent recruitment positions or long-term stable positions are recommended.

[0135] For example, for employees who have left their jobs within the past month, priority can be given to recommending urgently needed positions to meet their need for quick employment; for employees who have left their jobs more than six months ago, the preference is to recommend long-term stable positions to balance career stability and development plans. This design helps improve the timeliness of job matching while promoting the rationality of employees' career development.

[0136] Furthermore, personalized job recommendations can be implemented based on employees' current residence or desired work location. For example, positions matching the employee's current location or preferred region can be prioritized. This design makes job recommendations more geographically aligned with employees' actual circumstances, helping to improve the feasibility of job placement and employee satisfaction.

[0137] To further enhance the personalization of job recommendations and user experience, personalized recommendation messages can be generated for the job recommendation results, and follow-up can be conducted based on employee behavior.

[0138] Regarding recommendation generation, the system can automatically generate personalized recommendation reasons for each employee using a large model. For example, for employee Xiaoming, whose reason for leaving was voluntary resignation, the system can generate the following prompt:

[0139] "Dear job seekers, [Companies listed in the employee recommendation section] are actively recruiting! Multiple high-quality positions with generous salaries, comprehensive benefits, and fast promotion opportunities. Submit your resume or visit [website] for details, or call [number]."

[0140] We look forward to having you join us!

[0141] This design allows the recommended information to highlight the job's strengths, helping to attract employees' attention, increase their willingness to apply, and at the same time boost the exposure of the company's job postings and the conversion rate of applications.

[0142] In terms of follow-up, adjustments can be made dynamically based on employee behavior regarding recommended information to optimize recommendation frequency and channels. For example, if Xiaoming clicks on a recommended link and registers an account by entering his mobile phone number, but does not engage in any other activities within a certain period, updated job postings can be pushed to him again in a subsequent cycle. This ensures the continuity and effectiveness of recommended information while preventing frequent disruption to employees or being blocked by them, thus improving the rationality of the recommendation experience and customer acceptance.

[0143] Corresponding to the method embodiments of this application, this application also provides a job recommendation device for personnel undergoing job changes, such as... Figure 6 As shown, the recommended device 100 includes:

[0144] The structured processing module 110 is used to obtain recruitment-related data of personnel with job changes from the database of the human resources system and perform structured processing to obtain employee profile feature data and job resume data.

[0145] The experience text generation module 120 is used to convert the job resume data into work experience text described in natural language.

[0146] The evaluation text generation module 130 is used to generate evaluation texts reflecting career development based on job information and salary data in job resume data.

[0147] The skill text generation module 140 is used to input the job resume data into a trained skill prediction model, output skill keywords related to the job resume data, and then generate skill tag text.

[0148] The resume generation module 150 is used to generate job application resumes for the personnel whose positions have changed, based on the employee profile feature data, work experience text, evaluation text, and skill tag text.

[0149] The job recommendation module 160 is used to perform multi-dimensional feature vector matching between the job applicants' resumes and the job requirements to determine the jobs to be recommended and recommend them to the job applicants.

[0150] Figure 7 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of this application. The electronic device can be implemented as a server or other various terminal devices, such as desktop personal computers, tablet computers, laptop computers, mobile phones, etc., including a processor 601 and a memory 602. The memory 602 stores a program instruction set. When the processor 601 executes the program instruction set in the memory 602, it implements any of the aforementioned methods for recommending positions to personnel undergoing job changes.

[0151] Specifically, the processor 601 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of the present invention.

[0152] Memory 602 may include mass storage for data or instructions. For example, and not limitingly, memory 602 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 602 may include removable or non-removable (or fixed) media. Where appropriate, memory 602 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 602 is non-volatile solid-state memory.

[0153] Memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the method of recommending job positions to personnel undergoing job changes provided by this invention.

[0154] In one example, the electronic device may also include a communication interface 603 and a bus 604. The processor 601, memory 602, and communication interface 603 are connected via bus 604 and communicate with each other. Communication interface 603 is primarily used to enable communication between modules, devices, units, and / or equipment in the embodiments of the present invention. Bus 604 includes hardware, software, or both, coupling components of the online data traffic billing device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 604 may include one or more buses. Although specific buses are described and illustrated in the embodiments of the present invention, the present invention is contemplated by any suitable bus or interconnect.

[0155] The present invention also provides a computer-readable storage medium having stored thereon computer program instructions, which, when executed by a processor, implement any of the methods for recommending job positions to personnel undergoing job changes as described in the foregoing embodiments. The computer-readable storage medium can be any medium that can tangibly contain or store computer-executable instructions for use by or in conjunction with an instruction execution system, apparatus, or device. The storage medium can be a transient computer-readable storage medium or a non-transitory computer-readable storage medium. Non-transitory computer-readable storage media may include, but are not limited to, magnetic storage devices, optical storage devices, and / or semiconductor storage devices. Examples of such storage devices include, for example, magnetic disks, optical discs based on CD, DVD, or Blu-ray technology, and persistent solid-state storage such as flash memory and solid-state drives.

[0156] This invention also provides a computer program product comprising a set of computer program instructions, which, when executed by a processor, implement any of the methods for recommending job positions to personnel undergoing job changes as described in the foregoing embodiments. The computer program product includes, but is not limited to, application installation packages published on websites and in app stores, application plugins, and mini-programs that can run within certain applications.

[0157] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0158] The above embodiments are for illustrative purposes only and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the scope of the invention. Therefore, all equivalent technical solutions should also fall within the scope of the invention.

Claims

1. A method for recommending new positions to employees undergoing job changes, characterized in that, include: The recruitment-related data of personnel whose positions have changed will be obtained from the human resources system and processed in a structured manner to obtain employee profile feature data and job resume data. The job resume data is converted into work experience text described in natural language. An evaluation text reflecting career development is generated based on job information and salary data from job resume data. The job resume data is input into the trained skill prediction model, which outputs skill keywords related to the job resume data, and then generates skill tag text. The job application resume of the employee who changed positions is generated based on the employee profile feature data, work experience text, evaluation text, and skill tag text. The job applicants' resumes are matched with the job requirements using multi-dimensional feature vector matching to determine the positions to be recommended and then recommended to the applicants whose positions have changed.

2. The method according to claim 1, characterized in that, Based on job information and salary data in job resume data, a salary and job time series is generated, which in turn generates an evaluation text reflecting career development. The salary and job time series is a serialized data set that is arranged in chronological order and contains both job information and salary data, used to show the changing trends of salary and job.

3. The method according to claim 2, characterized in that, Further includes: Salary nodes are extracted from the time series of the salary positions and salary change values ​​are calculated. The salary change values ​​are then compared with industry standard salaries to determine the salary increase. The salary job time series is matched with the career growth trajectory template to determine the job promotion speed; Based on the salary increase and promotion speed, generate self-evaluation texts and / or others' evaluation texts reflecting the job stability and learning ability of employees who have changed positions.

4. The method according to claim 1, characterized in that, Converting the job resume data into work experience text described in natural language includes: The target fields in the job resume data are converted into semantic tags through field mapping, and then combined in a preset order to generate an input template. The target fields include one or more of the following: job name, service object, and function. The input template is used to generate natural language description text that conforms to the recruitment context based on semantic association, and the description text highlights the job responsibilities and achievements. The work experience text is obtained by optimizing the output of the descriptive text. The optimization process includes one or more of the following: standardizing sentence structure and terminology, preserving quantitative indicators, and enhancing the readability of grammatical structure.

5. The method according to claim 1, characterized in that, The training method for the skill prediction model includes: Job descriptions and job skills information are collected from a large number of resumes and job postings as pre-training data, and the pre-training data is preprocessed. The pre-training data is input into the model to be trained, and pre-training is performed through vectorization mapping until the preset training index is reached. Collect task-specific data with job-skill correspondence labels, construct fine-tuning training data including input samples and output samples, and input the fine-tuning training data into the pre-trained model for fine-tuning until the preset training index is reached. Deploy the finely tuned model to the recommendation system and perform quantitative compression and / or regular maintenance as needed.

6. The method according to claim 1, characterized in that, The process of matching the job resumes with the job requirements using multi-dimensional feature vectors includes: The job applicants' resumes and job requirements are each used to generate multi-dimensional feature vectors. Based on the weighted similarity calculation of the two multi-dimensional feature vectors, one or more job positions to be recommended are determined and recommended to the job applicants.

7. The method according to claim 1, characterized in that, Structured processing of recruitment-related data includes: Match the original job titles in the recruitment-related data with the standard job titles in the preset job dictionary, and output standardized job titles; and / or Map the original industry description information in the recruitment-related data to industry categories, and output target industry tags that conform to the industry classification standards of the recruitment system; and / or The rule engine is used to identify and delete sensitive information fields that are irrelevant to job recommendations.

8. A device for recommending new positions to employees undergoing job changes, characterized in that, include: The structured processing module is used to perform structured processing on recruitment-related data of personnel whose positions have changed from the human resources system, to obtain employee profile feature data and job resume data. The experience text generation module is used to convert the job resume data into work experience text described in natural language; The evaluation text generation module is used to generate evaluation texts reflecting career development based on job information and salary data in job resume data. The skill text generation module is used to input the job resume data into a trained skill prediction model, output skill keywords related to the job resume data, and then generate skill tag text. The resume generation module is used to generate job application resumes for the personnel whose positions have changed, based on the employee profile feature data, work experience text, evaluation text, and skill tag text. The job recommendation module is used to perform multi-dimensional feature vector matching between the job applicants' resumes and the job requirements to determine the jobs to be recommended and recommend them to the job applicants.

9. An electronic device, characterized in that, This includes a processor and a memory, the memory storing a set of computer program instructions, which, when executed by the processor, implement the method for recommending positions to personnel undergoing job changes as described in any one of claims 1-7.

10. A computer program product, characterized in that, It includes a set of computer program instructions that, when executed by a processor, implement the method for recommending positions to personnel undergoing job changes as described in any one of claims 1-7.

Images