An intelligent recruitment system and method based on LLM multi-terminal applications

Abstract

This invention relates to the field of enterprise recruitment technology, and discloses an intelligent recruitment system and method based on LLM multi-terminal applications. It includes an enterprise feature module to extract recruitment demand features and applicant supply features; a user screening module to screen candidate users; a virtual question-and-answer module to extract text data; a vector transformation module to identify soft skills features and generate intrinsic core vectors; and a recruitment recommendation module to calculate comprehensive scores and recommend candidates in order. This invention can perform a dual-integrated analysis of the fit between applicants and positions, considering both their hardware and software conditions. It not only overcomes the shortcomings of focusing solely on surface keyword matching while ignoring intrinsic soft skills, achieving a comprehensive quantitative evaluation of applicants, but also accurately uncovers the hidden potential value of applicants and positions, improving the value fit between personnel and positions and enhancing the reliability of intelligent recruitment results.

Description

Technical Field

[0001] This invention relates to the field of enterprise recruitment technology, and more specifically, to an intelligent recruitment system and method based on LLM multi-terminal applications. Background Technology

[0002] With the continuous development of artificial intelligence technology, the field of corporate human resources recruitment is undergoing a transformation from manual screening to intelligent recruitment. Traditional recruitment models consume a lot of time and energy to identify and screen the resumes of applicants, resulting in low recruitment efficiency. Therefore, intelligent recruitment systems based on large language models (LLM) are gradually becoming popular in the industry.

[0003] Reference patent application CN120069826A discloses an intelligent recruitment system and method based on LLM multi-terminal applications, including a multi-terminal interactive interface module, an information interaction and processing module, a large model analysis module, a database management module, and a result display and feedback module; the information interaction and processing module monitors various input information from multiple users in real time; the large model analysis module selects suitable LLM large models for recruitment scenarios during the model selection and adaptation process; the database management module regularly collects job information from various authoritative channels; and the result display and feedback module is responsible for receiving matching result data from the large model analysis module. Existing intelligent recruitment systems employ a keyword-matching-based approach, comparing and recruiting applicants based on structured information such as education, years of work experience, and skill tags. This approach focuses solely on the applicant's external hardware qualifications, neglecting the analysis and comparison of their internal software qualifications. This results in a limited and singular data dimension for applicant recruitment analysis. Furthermore, it lacks the ability to integrate and quantify the applicant's hardware and software qualifications, failing to provide a comprehensive, dual-dimensional assessment of the fit between the applicant and the position, thus reducing the reliability of the intelligent recruitment results.

[0004] In view of this, the present invention proposes an intelligent recruitment system and method based on LLM multi-terminal applications to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: an intelligent recruitment system based on LLM multi-terminal applications, applied to a recruitment platform, comprising: The enterprise characteristics module receives recruitment data from companies recruiting on the enterprise platform and extracts recruitment demand characteristics from the recruitment data. The user characteristics module receives job application data from users on the platform and extracts job supply characteristics from the application data. The user screening module calculates the job competency score of applicants based on the characteristics of recruitment needs and the characteristics of job supply, and selects candidate users from the applicants. The virtual Q&A module indexes virtual indicators based on job applicant supply characteristics, constructs a recruitment scenario that matches the virtual indicators, and extracts text data of candidate users in the recruitment scenario. The vector transformation module identifies the soft competency features of text data through the LLM model and transforms these soft competency features into vectors to generate intrinsic core vectors. The intrinsic core vectors include implicit competency vectors, growth potential vectors, and cultural value vectors. The recruitment recommendation module calculates the job value score through an intrinsic core vector, dynamically allocates value weight factors for the job value score and job competency score, calculates a comprehensive score, and recommends candidate users in order.

[0006] Furthermore, recruitment needs are characterized by identity characteristics, requirements characteristics, and compensation characteristics: The job requirements include age and gender; the job requirements include education level, major, years of work experience, and professional skills; and the compensation requirements include base salary and total salary.

[0007] Furthermore, the characteristics of job applicants include their identity, skills, and compensation. The information provided includes the applicant's age and gender; the information provided includes the applicant's education, major, years of service, and skills; and the information provided includes the applicant's base salary and total salary.

[0008] Furthermore, the selection method for candidate users is as follows: The applicant's age being within the recruitment age range and the applicant's gender being consistent with the recruitment gender are recorded as the first competency characteristic. The number of first competency characteristics is counted to obtain the first sub-competency value. The skills provided by applicants whose educational background is not lower than that of the job posting, whose major is within the scope of the professional direction, whose years of service are greater than or equal to the years of work experience, and whose skills are within the scope of professional skills are recorded as the second competency characteristics. The number of second competency characteristics is counted to obtain the second competency value. The compensation offered by applicants whose base salary and comprehensive salary fall within the range of base salary and comprehensive salary is recorded as the third competency characteristic. The number of third competency characteristics is then calculated to obtain the third sub-competency value. After assigning corresponding weight factors to the first, second, and third sub-competency values ​​respectively, the job competency score is calculated by weighted summation. When the job competency score is greater than or equal to the calibrated competency threshold, the applicant will be recorded as a candidate.

[0009] Furthermore, virtual indicators include identity indicators, ability indicators, and compensation indicators; Identity indicators include age and gender; ability indicators include education, major, years of service and skills; compensation indicators include base salary and total salary.

[0010] Furthermore, the method for constructing a recruitment scenario is as follows: When the first competency feature exists among the job applicant supply characteristics, the first competency feature is used as the first index to retrieve the identity indicator from the indicator database; When a second competency feature exists among the job application supply characteristics, the second competency feature is used as the second index to retrieve the competency indicators from the indicator database. When a third competency characteristic exists among the job applicant supply characteristics, the third competency characteristic is used as the third index to index the compensation indicator from the indicator database. Using the identity, capability, and compensation indicators of the candidate user index as the scenario boundaries, a virtual scenario with two question-and-answer ports is simulated, and the two question-and-answer ports are respectively denoted as the enterprise port and the user port; A bidirectional communication link is established between the enterprise port and the user port, and question-and-answer permissions and logic are configured on the communication link to enable the virtual scene to be transformed into a recruitment scene.

[0011] Furthermore, soft skills characteristics include skill breadth, skill depth, skill interaction value, communication skills, stress tolerance, learning potential, cultural fit, style matching, and ideological alignment. The method for generating the intrinsic core vector is as follows: The skill width, skill height, and skill interaction value are combined into a capability unit. The three vector elements that match the capability unit are retrieved from the vector lookup table. The three vector elements are normalized to generate width element value, height element value, and interaction element value. The width element value, height element value, and interaction element value are then concatenated and combined to form an invisible capability vector. Communication skills, stress resistance level, and learning potential are combined into potential units. Three vector elements that match the potential unit are retrieved from the vector lookup table. After normalizing the three vector elements, communication element values, stress resistance element values, and learning element values ​​are generated. The communication element values, stress resistance element values, and learning element values ​​are then concatenated and combined into a growth potential vector. Cultural fit, style matching, and concept alignment are combined into a value unit. Three vector elements that match the value unit are retrieved from the vector lookup table. After normalizing the three vector elements, cultural element values, style element values, and concept element values ​​are generated. The cultural element values, style element values, and concept element values ​​are then concatenated and combined to form a cultural value vector.

[0012] Furthermore, the calculation method for job value score is as follows: Retrieve the standard capability vector, standard potential vector, and standard value vector from the database; The ability value score is calculated by subtracting the width, height, and interaction values ​​of the stealth ability vector from the standard width, standard height, and standard interaction values ​​of the standard ability vector, summing the differences, and averaging them. The potential value score is calculated by subtracting the communication, resilience, and learning element values ​​of the growth potential vector from the standard communication, resilience, and learning values ​​of the standard potential vector, summing the differences, and averaging them. The cultural value vector is calculated by subtracting the cultural element value, style element value, and concept element value from the standard cultural value, standard style value, and standard concept value of the standard value vector, respectively. The differences are then added together and averaged to obtain the cultural value score. The job value score is calculated by adding the competency value score, potential value score, and cultural value score together.

[0013] Furthermore, the value weighting factors include hardware weighting factors and software weighting factors; The overall score is calculated as follows: Retrieve all job posting events within the historical time period, retrieve the first and second weight factors from the job posting events, and calculate the hardware weight factor and software weight factor by averaging all the first and second weight factors. The comprehensive score is calculated by weighting and summing the hardware weight factor, software weight factor, job value score, and job competency score.

[0014] A smart recruitment method based on LLM multi-terminal applications, applied to a recruitment platform, is implemented based on the aforementioned smart recruitment system based on LLM multi-terminal applications, including: S01: Receive recruitment data from companies recruiting on the enterprise platform and extract recruitment demand characteristics from the recruitment data; S02: Receive job application data from users on the platform and extract job supply characteristics from the application data; S03: Based on the characteristics of recruitment demand and job supply, calculate the job competency score of job applicants and select candidate users from the job applicants; S04: Index virtual indicators through job supply characteristics, construct recruitment scenarios that match the virtual indicators, and extract text data of candidate users in the recruitment scenarios; S05: Identify the soft capability features of text data through the LLM model, and transform the soft capability features into vectors to generate an intrinsic core vector; S06: Calculate the job value score through the intrinsic core vector, dynamically allocate the value weight factors of the job value score and job competency score, calculate the comprehensive score, and recommend candidate users in order.

[0015] The technical advantages of this invention, a smart recruitment system and method based on LLM multi-terminal applications, are as follows: (1): This invention calculates job competency scores in the hardware dimension and job value scores in the software dimension, enabling a dual-integrated analysis of the fit between job applicants and positions based on their own hardware conditions and inherent software conditions. This dual-integrated and quantitative approach not only overcomes the shortcomings of traditional methods that only focus on surface keyword matching while ignoring inherent soft capabilities, but also achieves a comprehensive quantitative evaluation of job applicants. Furthermore, it can accurately uncover the hidden potential value of job applicants and positions, improve the value fit between personnel and positions, and enhance the reliability of intelligent recruitment results for enterprises.

[0016] (2): This invention quantifies the abstract implicit capability vector, growth potential vector and cultural value vector into calculable and comparable intrinsic core vectors, which can lay the foundation for accurate matching of applicants and positions, improve the accuracy of the analysis results of applicants and positions at the software condition level, and also reflect the potential long-term value of applicants in multiple ways, thereby ensuring that the recruited personnel can not only meet the short-term condition requirements of the position, but also play a potential economic role in the long-term planning of the position, thus improving the application effect of the intelligent recruitment system. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the architecture of an intelligent recruitment system based on LLM multi-terminal applications provided in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of a module of an intelligent recruitment system based on an LLM multi-terminal application provided in Embodiment 1 of the present invention; Figure 3 This is a flowchart illustrating an intelligent recruitment method based on an LLM multi-terminal application, as provided in Embodiment 2 of the present invention. Detailed Implementation

[0018] 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, and 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.

[0019] Example 1: Please refer to Figures 1-2As shown in the figure, the intelligent recruitment system based on LLM multi-terminal application described in this embodiment is applied to a recruitment platform and includes: The enterprise characteristics module receives recruitment data from recruiting companies within the enterprise platform, extracts the recruitment requirement characteristics of the positions from the recruitment data, and transmits it to the recruitment platform. Intelligent recruitment systems help recruiting companies publish job postings and accurately select suitable candidates from applicants based on the information collected from them. In this embodiment, the intelligent recruitment system consists of an enterprise platform, a user platform, and a recruitment platform. The enterprise platform is used by recruiting companies to collect recruitment information, the user platform is used by job seekers to collect job application information, and the recruitment platform provides an environment for intelligent analysis of recruitment and job application information to help recruiting companies conduct recruitment quickly and accurately.

[0020] It should be noted that the intelligent recruitment system has high device and system compatibility and is not limited to a single or fixed port and device. In this embodiment, the intelligent recruitment system can typically be applied to PC, APP, WeChat mini program, WeChat official account and other platforms.

[0021] Recruitment data consists of diverse data on job positions entered by recruiting companies on the company platform. This data can comprehensively represent the requirements and compensation of the positions. Due to the large amount of recruitment data and its relatively general nature, it is necessary to further analyze the recruitment data and extract recruitment demand characteristics that can concisely and accurately represent the requirements and compensation of the positions. Specifically, recruitment demand characteristics include identity characteristics, requirements characteristics, and compensation characteristics; identity characteristics refer to the job requirements for the applicant's identity, requirements characteristics refer to the job requirements for the applicant's abilities, and compensation characteristics refer to the job requirements for the applicant's salary.

[0022] Identity characteristics include age and gender; this can indicate the age and gender required for the position in the recruiting company.

[0023] Requirements include the candidate's educational background, major, years of work experience, and professional skills; these can describe the professional abilities required for the position by the recruiting company.

[0024] Compensation characteristics include base salary and total salary; this indicates the salary and benefits offered for this position by the recruiting company.

[0025] Once the recruitment needs characteristics are obtained, they can be transmitted to the recruitment platform for storage, thereby fulfilling the recruiting company's hard requirements for the position and ensuring that the recruitment data information on the company's platform can be effectively recorded and stored.

[0026] The user characteristics module receives job application data from users on the platform, extracts job application supply characteristics from the application data, and transmits it to the recruitment platform. Job application data consists of diverse data input by job applicants into the user platform, which can comprehensively represent the user's abilities and information from multiple perspectives. Due to the large amount of job application data and its relatively general nature, it is necessary to further analyze the job application data and extract job supply characteristics that can concisely and accurately represent the user's abilities and information. Specifically, the characteristics of job applicants include identity provision, skills provision, and compensation provision.

[0027] The identity information provided includes the applicant's age and gender; it can represent the identity information of the applicant.

[0028] Skills provision includes the applicant's educational background, major, years of experience, and skills; it is used to represent the applicant's ability to meet the job requirements.

[0029] The compensation package includes the applicant's base salary and comprehensive salary; it can express the applicant's salary and benefits requirements for the corresponding position.

[0030] Once the characteristics of job applicants are obtained, they can be transmitted to the recruitment platform for storage, thereby enabling job applicants to provide the hard requirements for the position and ensuring that the application data information on the user platform can be effectively recorded and stored.

[0031] The user screening module integrates and analyzes the characteristics of recruitment needs and the characteristics of job supply to calculate the job competency score of job applicants and select candidate users from the applicants. After obtaining the characteristics of recruitment demand and job supply, the recruitment platform needs to analyze and compare these characteristics to determine the degree of suitability between job applicants and the job in terms of hard requirements. The degree of suitability is then quantified to obtain a job competency score that can be expressed numerically.

[0032] The method for calculating job competency scores is as follows: An overlap analysis of identity characteristics and identity supply; The applicant's age being within the recruitment age range and the applicant's gender being consistent with the recruitment gender are recorded as the first competency characteristic. The number of first competency characteristics is counted to obtain the first sub-competency value. An overlap analysis was conducted on the required characteristics and skill supply. The skills provided by applicants whose educational background is not lower than that of the job posting, whose major is within the scope of the professional direction, whose years of service are greater than or equal to the years of work experience, and whose skills are within the scope of professional skills are recorded as the second competency characteristics. The number of second competency characteristics is counted to obtain the second competency value. An overlap analysis of compensation characteristics and compensation supply; The compensation offered by applicants whose base salary and comprehensive salary fall within the range of base salary and comprehensive salary is recorded as the third competency characteristic. The number of third competency characteristics is then calculated to obtain the third sub-competency value. After assigning corresponding weight factors to the first, second, and third sub-competency values ​​respectively, the job competency score is calculated by weighted summation. The formula for calculating the job competency score is as follows: ; In the formula, To score points for job competence, As the first piece's competence value, As the second piece's competence value, As the third child's competence value, , , These are the weighting factors for the first, second, and third sub-competency values, respectively. Greater than , Greater than , , , The sum is 1; for example, It is 0.5. It is 0.3. It is 0.2.

[0033] After obtaining the job competency scores of applicants, the suitability of the applicants for the job can be represented by the job competency scores. Then, candidates whose qualifications meet the job requirements can be selected from a large number of applicants. Specifically, the selection method for candidate users is as follows: The job competency scores of applicants are compared with the defined competency thresholds. When the job competency score is lower than the set competency threshold, it indicates that the applicant's hardware conditions are not well-suited to the job, and the applicant will not be recorded as a candidate. When the job competency score is greater than or equal to the calibrated competency threshold, it indicates that the applicant's hardware conditions are highly suitable for the job. In this case, the applicant is recorded as a candidate, and A candidates are obtained.

[0034] It should be noted that the calibrated competency threshold is the minimum value of the job competency score of candidate users who are suitable for the job in terms of hardware conditions, thereby ensuring that subsequent candidate users have a certain degree of suitability. In this embodiment, the calibrated competency threshold is obtained by collecting the minimum value of the job competency score of a large number of historically marked candidate users and then calculating the average value. The number of candidate users is not fixed and is a positive integer greater than or equal to 0.

[0035] The virtual Q&A module indexes virtual indicators based on job applicant supply characteristics, constructs a recruitment scenario that matches the virtual indicators, conducts virtual Q&A within the recruitment scenario, and extracts the text data of candidate users in the virtual Q&A. Since the number of identified candidate users may not be unique, and the candidate users may not be able to effectively meet the job requirements under the software conditions, it is necessary to further analyze the suitability of candidate users for the software conditions and the job.

[0036] Virtual metrics are recruitment assessment indicators that are indexed from the indicator database based on the job application supply characteristics of candidate users and are used to reflect the candidate users' own hardware conditions. The recruitment scenario is a virtual recruitment environment designed for recruitment companies and candidate users. Virtual indicators include identity indicators, ability indicators, and compensation indicators; among them, identity indicators, ability indicators, and compensation indicators are used to adapt to the first competency characteristic, the second competency characteristic, and the third competency characteristic, respectively, so as to meet the adaptation needs of different job supply characteristics.

[0037] In this embodiment, the identity indicators include age and gender, the ability indicators include education, major, years of service and skills, and the compensation indicators include base salary and total salary.

[0038] It should be noted that the number of virtual indicators is not fixed. It is determined by the specific type and number of the first, second, and third competency characteristics recorded in the job application supply characteristics of the candidate users. Therefore, the recruitment scenarios constructed based on the virtual indicators also correspond one-to-one with the candidate users to ensure that each candidate user can have a corresponding recruitment scenario.

[0039] Specifically, the method for constructing a recruitment scenario is as follows: When the first competency feature exists in the job application supply characteristics, the first competency feature is used as the first index, and the identity indicator corresponding to the first index is indexed from the indicator database; When a second competency feature exists in the job application supply characteristics, the second competency feature is used as the second index to index the corresponding ability indicators from the indicator database. When a third competency characteristic exists in the job application supply characteristics, the third competency characteristic is used as the third index, and the corresponding salary indicator is indexed from the indicator database. Using the identity, ability, and compensation indicators of the candidate user index as the scenario boundary, a virtual scenario with two question-and-answer ports is simulated, and the two question-and-answer ports are respectively denoted as the enterprise port and the user port; the scenario boundary refers to the question-and-answer background outline of the virtual recruitment process in the recruitment scenario, ensuring that the virtual questions and answers in the recruitment scenario can match the specific content of the identity, ability, and compensation indicators, thereby improving the rationality of the virtual questions and answers. Establish a bidirectional communication link between the enterprise port and the user port, and configure question-and-answer permissions and logic on the communication link to enable the virtual scene to be transformed into a recruitment scene.

[0040] It should be noted that question-and-answer permissions refer to the execution permissions of relevant data information when virtual question-and-answer occurs between the enterprise port and the user port in the communication link, ensuring that the data information during virtual question-and-answer does not exceed the scope of permissions. Question-and-answer logic refers to the logical rules for data interaction between the enterprise port and the user port in the communication link, ensuring that the data information during virtual question-and-answer is reasonable, smooth, and orderly.

[0041] Once the recruitment scenario is obtained, recruiting companies and potential candidates can conduct virtual recruitment Q&A sessions, enabling them to interact and get to know each other without meeting in person. This helps recruiting companies gain a deeper understanding of potential candidates. Text data refers to the aggregated data of text, voice, and images generated when recruiting companies and potential users engage in virtual Q&A in a recruitment scenario. In this embodiment, in order to improve the authenticity and reliability of text data, text data is recorded and extracted in text form. Therefore, it is necessary to convert the voice and image data in the virtual Q&A into text before recording and extracting them. Specifically, when extracting text data, the data format of all Q&A data in the recruitment scenario is queried one by one. Q&A data in audio format is recorded as voice data, and Q&A data in image format is recorded as image data. Voice data is converted into text data through speech recognition technology, and image data is converted into text data through computer vision technology.

[0042] The vector transformation module inputs text data into the LLM model, identifies the soft skills characteristics of candidate users through the LLM model, and transforms the soft skills characteristics into vectors to generate intrinsic core vectors. The intrinsic core vectors include implicit ability vectors, growth potential vectors, and cultural value vectors. The LLM model is a large language model based on deep learning technology that can identify and extract key true intentions from text data. In this embodiment, the input data of the LLM model is text data, and the output data of the LLM model is soft ability features, which can accurately consider the soft conditions corresponding to the candidate user's performance in the recruitment scenario.

[0043] Soft skills characteristics are used to assess the characteristics that candidate users possess in response to virtual questions and answers, demonstrating their inherent strengths and potential abilities for the position, and serve as a basis for judging whether candidate users can play a role and create value in the position. Specifically, soft skills characteristics include skill breadth, skill depth, skill interaction value, communication skills, stress tolerance, learning potential, cultural fit, style matching, and ideological alignment. In this embodiment, skill breadth refers to the scope of professional skills covered by the candidate user; skill depth refers to the extent to which the candidate user has mastered professional skills; skill interaction value refers to the candidate user's ability to integrate different skills; communication ability refers to the candidate user's ability to communicate within a team; stress resistance level refers to the candidate user's performance under work pressure; learning potential refers to the candidate user's learning ability; cultural fit refers to the degree to which the candidate user fits the corporate culture; style matching refers to the degree to which the candidate user matches the company's operating style; and philosophy alignment refers to the degree to which the candidate user aligns with the company's development philosophy.

[0044] In this embodiment, the LLM model is based on the deep learning model in the prior art. It is obtained by repeatedly training a large amount of text data and corresponding soft ability features, so that the LLM model can automatically and accurately extract the corresponding soft ability features from the input text data. During LLM model training, text data is used as input data and soft ability features are used as output data. The number of iterations and accuracy threshold of the LLM model are set, and the training stops when the LLM model reaches the training accuracy threshold. The training process of the LLM model in this embodiment is the prior art in this field and will not be described in detail here.

[0045] By inputting the text data of A candidate users one by one into the LLM model, and accurately identifying and extracting the true meaning in the text data, the corresponding soft ability features can be output.

[0046] The extracted soft skills features can be used to represent the soft conditions of candidate users from three aspects: implicit professional skills, job growth potential, and cultural and economic value. In order to better distinguish the meaning of the soft skills features, it is necessary to perform vector transformation on the soft skills features so that multiple different soft skills features can be converted into an internal core vector. Specifically, the intrinsic core vectors include implicit capability vectors, growth potential vectors, and cultural value vectors; that is, to spatially vectorize the candidate users' implicit professional capabilities, job growth potential, and cultural and economic value, respectively. Specifically, the method for generating the intrinsic core vector is as follows: Skill width, skill height, and skill interaction value are combined into a capability unit. Three vector elements matching the capability unit are retrieved from a vector lookup table. These three vector elements are normalized to generate width, height, and interaction element values. Finally, these values ​​are concatenated to form an implicit capability vector. The vector lookup table is a database table used to numerically represent the vector elements corresponding to soft capability features. It typically consists of two parts: the specific content of the soft capability feature and the numerical values ​​of the vector elements corresponding to that feature. Communication skills, stress resistance level, and learning potential are combined into potential units. Three vector elements that match the potential unit are retrieved from the vector lookup table. After normalizing the three vector elements, communication element values, stress resistance element values, and learning element values ​​are generated. The communication element values, stress resistance element values, and learning element values ​​are then concatenated and combined into a growth potential vector. Cultural fit, style matching, and concept alignment are combined into a value unit. Three vector elements that match the value unit are retrieved from the vector lookup table. After normalizing the three vector elements, cultural element values, style element values, and concept element values ​​are generated. The cultural element values, style element values, and concept element values ​​are then concatenated and combined to form a cultural value vector.

[0047] It should be noted that the element values ​​of the constructed implicit ability vector, growth potential vector, and cultural value vector are all between 0 and 1, which facilitates subsequent management operations and also ensures that the internal core vectors can maintain a unified calculation standard, thereby improving the accuracy of subsequent related numerical calculations.

[0048] The recruitment recommendation module calculates the job value score of candidate users through the intrinsic core vector, dynamically allocates the value weight factors of job value score and job competency score, calculates the comprehensive score of candidate users, and recommends candidate users in order based on the comprehensive score. The job value score is used to represent the degree of fit between the implicit ability vector, growth potential vector, and cultural value vector and the job at the software level, and serves as the fit numerical result of the candidate user in the three software condition dimensions. The method for calculating job value score is as follows: The standard ability vector, standard potential vector, and standard value vector are retrieved from the database. The standard ability vector, standard potential vector, and standard value vector refer to the minimum conditions for the implicit ability vector, growth potential vector, and cultural value vector to be suitable for the job, respectively. Therefore, the values ​​of each vector element in the standard ability vector, standard potential vector, and standard value vector are the minimum values ​​that meet the above conditions. The ability value score is calculated by subtracting the width, height, and interaction values ​​of the stealth ability vector from the standard width, standard height, and standard interaction values ​​of the standard ability vector, summing the differences, and averaging them. The formula for calculating the competence value score is: ; In the formula, Score the value of ability. The width of the element. This is the standard width value. The height element value, This is the standard height value. For interactive element values, Standard interaction value; The potential value score is calculated by subtracting the communication, resilience, and learning element values ​​of the growth potential vector from the standard communication, resilience, and learning values ​​of the standard potential vector, summing the differences, and averaging them. The formula for calculating the potential value score is: ; In the formula, Score for potential value. For communication element values, For standard communication values, The value of the compressive strength element. Standard compressive strength value, To learn element values, This is the standard learning value; The cultural value vector is calculated by subtracting the cultural element value, style element value, and concept element value from the standard cultural value, standard style value, and standard concept value of the standard value vector, respectively. The differences are then added together and averaged to obtain the cultural value score. The formula for calculating cultural value score is: ; In the formula, Score for cultural value. As for cultural elements, As the standard cultural value, For style element values, For standard style values, For concept element values, This is the standard concept value; The job value score is calculated by adding the competency value score, potential value score, and cultural value score together.

[0049] The job value score can only represent the suitability of candidate users with the job in multiple dimensions of software conditions. It cannot achieve a comprehensive evaluation of software and hardware conditions. Therefore, it is necessary to summarize and analyze the job value score and job competency score to obtain a comprehensive score that can analyze the suitability of hardware and software conditions from an overall perspective. When calculating the overall score, it is necessary to assign corresponding value weight factors to the job value score and the job competency score respectively, so as to reflect the importance of the job value score and the job competency score in the recruitment process. In this embodiment, the value weighting factors include hardware weighting factors and software weighting factors; the hardware weighting factors and software weighting factors are matched with job competency scores and job value scores, respectively.

[0050] Specifically, the overall score is calculated as follows: Retrieve all job posting events within the historical time period, retrieve the first and second weight factors from the job posting events, and calculate the hardware weight factor and software weight factor by averaging all the first and second weight factors. The comprehensive score is calculated by weighting and summing the hardware weighting factor, software weighting factor, job value score, and job competency score. The formula for calculating the overall score is: ; In the formula, For the overall score, To score the value of the job, Hardware weighting factor, It is a software weighting factor, and ,of The sum is 1, for example. It is 0.45. It is 0.55.

[0051] After calculating the comprehensive score, the intelligent recruitment of candidate users A on the recruitment platform has completed the data analysis and calculation process. At this point, it is necessary to give corresponding conclusions on the intelligent recruitment operations of candidate users A based on the analysis and calculation results, so as to help recruiting companies quickly and accurately screen suitable candidates for positions. In this embodiment, when recommending candidate users to recruiting companies, the A candidate users are arranged into a recruitment queue in descending order of their comprehensive scores. Then, based on the number of positions available as set by the recruiting company, candidate users matching the number of positions are sequentially transmitted to the recruitment platform, starting from the head of the queue. This allows for the screening and recommendation of applicants who meet the job requirements in this intelligent recruitment process, thereby achieving intelligent recruitment screening for applicants and improving the convenience and accuracy of job recruitment for companies.

[0052] Example 2: Please refer to Figure 3 As shown, parts not described in detail in this embodiment are described in Embodiment 1. This embodiment provides an intelligent recruitment method based on an LLM multi-terminal application, applied to a recruitment platform, and implemented using an intelligent recruitment system based on an LLM multi-terminal application, including: S01: Receive recruitment data from companies recruiting on the enterprise platform and extract recruitment demand characteristics from the recruitment data; S02: Receive job application data from users on the platform and extract job supply characteristics from the application data; S03: Based on the characteristics of recruitment demand and job supply, calculate the job competency score of job applicants and select candidate users from the job applicants; S04: Index virtual indicators through job supply characteristics, construct recruitment scenarios that match the virtual indicators, and extract text data of candidate users in the recruitment scenarios; S05: Identify the soft capability features of text data through the LLM model, and transform the soft capability features into vectors to generate an intrinsic core vector; S06: Calculate the job value score through the intrinsic core vector, dynamically allocate the value weight factors of the job value score and job competency score, calculate the comprehensive score, and recommend candidate users in order.

[0053] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. An intelligent recruitment system based on LLM multi-terminal applications, applied to a recruitment platform, characterized in that, include: The enterprise characteristics module receives recruitment data from companies recruiting on the enterprise platform and extracts recruitment demand characteristics from the recruitment data. The user characteristics module receives job application data from users on the platform and extracts job supply characteristics from the application data. The user screening module calculates the job competency score of applicants based on the characteristics of recruitment needs and the characteristics of job supply, and selects candidate users from the applicants. The virtual Q&A module indexes virtual indicators based on job applicant supply characteristics, constructs a recruitment scenario that matches the virtual indicators, and extracts text data of candidate users in the recruitment scenario. The vector transformation module identifies the soft competency features of text data through the LLM model and transforms these soft competency features into vectors to generate intrinsic core vectors. The intrinsic core vectors include implicit competency vectors, growth potential vectors, and cultural value vectors. The recruitment recommendation module calculates the job value score through an intrinsic core vector, dynamically allocates value weight factors for the job value score and job competency score, calculates a comprehensive score, and recommends candidate users in order.

2. The intelligent recruitment system based on LLM multi-terminal application according to claim 1, characterized in that, Recruitment requirements include characteristics of identity, requirements, and compensation: Identity characteristics include age and gender when recruiting; Requirements include educational background, major, years of work experience, and professional skills. Compensation features include base salary and total salary.

3. The intelligent recruitment system based on LLM multi-terminal application according to claim 2, characterized in that, Job applicants' supply characteristics include their status, skills, and compensation. The information provided includes the applicant's age and gender; the information provided includes the applicant's education, major, years of service, and skills; and the information provided includes the applicant's base salary and total salary.

4. The intelligent recruitment system based on LLM multi-terminal application according to claim 3, characterized in that, The selection method for candidate users is as follows: The applicant's age being within the recruitment age range and the applicant's gender being consistent with the recruitment gender are recorded as the first competency characteristic. The number of first competency characteristics is counted to obtain the first sub-competency value. The skills provided by applicants whose educational background is not lower than that of the job posting, whose major is within the scope of the professional direction, whose years of service are greater than or equal to the years of work experience, and whose skills are within the scope of professional skills are recorded as the second competency characteristics. The number of second competency characteristics is counted to obtain the second competency value. The compensation offered by applicants whose base salary and comprehensive salary fall within the range of base salary and comprehensive salary is recorded as the third competency characteristic. The number of third competency characteristics is then calculated to obtain the third sub-competency value. After assigning corresponding weight factors to the first, second, and third sub-competency values ​​respectively, the job competency score is calculated by weighted summation. When the job competency score is greater than or equal to the calibrated competency threshold, the applicant will be recorded as a candidate.

5. The intelligent recruitment system based on LLM multi-terminal application according to claim 4, characterized in that, Virtual metrics include identity metrics, ability metrics, and compensation metrics; Identity indicators include age and gender; ability indicators include education, major, years of service and skills; compensation indicators include base salary and total salary.

6. The intelligent recruitment system based on LLM multi-terminal application according to claim 5, characterized in that, The method for constructing a recruitment scenario is as follows: When the first competency feature exists among the job applicant supply characteristics, the first competency feature is used as the first index to retrieve the identity indicator from the indicator database; When a second competency feature exists among the job application supply characteristics, the second competency feature is used as the second index to retrieve the competency indicators from the indicator database. When a third competency characteristic exists among the job applicant supply characteristics, the third competency characteristic is used as the third index to index the compensation indicator from the indicator database. Using the identity, capability, and compensation indicators of the candidate user index as the scenario boundaries, a virtual scenario with two question-and-answer ports is simulated, and the two question-and-answer ports are respectively denoted as the enterprise port and the user port; A bidirectional communication link is established between the enterprise port and the user port, and question-and-answer permissions and logic are configured on the communication link to enable the virtual scene to be transformed into a recruitment scene.

7. The intelligent recruitment system based on LLM multi-terminal application according to claim 6, characterized in that, Soft skills characteristics include skill breadth, skill depth, skill interaction value, communication skills, stress tolerance, learning potential, cultural fit, style fit, and ideological alignment. The method for generating the intrinsic core vector is as follows: The skill width, skill height, and skill interaction value are combined into a capability unit. The three vector elements that match the capability unit are retrieved from the vector lookup table. The three vector elements are normalized to generate width element value, height element value, and interaction element value. The width element value, height element value, and interaction element value are then concatenated and combined to form an invisible capability vector. Communication skills, stress resistance level, and learning potential are combined into potential units. Three vector elements that match the potential unit are retrieved from the vector lookup table. After normalizing the three vector elements, communication element values, stress resistance element values, and learning element values ​​are generated. The communication element values, stress resistance element values, and learning element values ​​are then concatenated and combined into a growth potential vector. Cultural fit, style matching, and concept alignment are combined into a value unit. Three vector elements that match the value unit are retrieved from the vector lookup table. After normalizing the three vector elements, cultural element values, style element values, and concept element values ​​are generated. The cultural element values, style element values, and concept element values ​​are then concatenated and combined to form a cultural value vector.

8. The intelligent recruitment system based on LLM multi-terminal application according to claim 7, characterized in that, The method for calculating job value score is as follows: Retrieve the standard capability vector, standard potential vector, and standard value vector from the database; The ability value score is calculated by subtracting the width, height, and interaction values ​​of the stealth ability vector from the standard width, standard height, and standard interaction values ​​of the standard ability vector, summing the differences, and averaging them. The potential value score is calculated by subtracting the communication, resilience, and learning element values ​​of the growth potential vector from the standard communication, resilience, and learning values ​​of the standard potential vector, summing the differences, and averaging them. The cultural value vector is calculated by subtracting the cultural element value, style element value, and concept element value from the standard cultural value, standard style value, and standard concept value of the standard value vector, respectively. The differences are then added together and averaged to obtain the cultural value score. The job value score is calculated by adding the competency value score, potential value score, and cultural value score together.

9. The intelligent recruitment system based on LLM multi-terminal application according to claim 8, characterized in that, Value weighting factors include hardware weighting factors and software weighting factors; The overall score is calculated as follows: Retrieve all job posting events within the historical time period, retrieve the first and second weight factors from the job posting events, and calculate the hardware weight factor and software weight factor by averaging all the first and second weight factors. The comprehensive score is calculated by weighting and summing the hardware weight factor, software weight factor, job value score, and job competency score.

10. An intelligent recruitment method based on LLM multi-terminal applications, applied to a recruitment platform, implemented based on an intelligent recruitment system based on LLM multi-terminal applications as described in any one of claims 1-9, characterized in that... include: S01: Receive recruitment data from companies recruiting on the enterprise platform and extract recruitment demand characteristics from the recruitment data; S02: Receive job application data from users on the platform and extract job supply characteristics from the application data; S03: Based on the characteristics of recruitment demand and job supply, calculate the job competency score of job applicants and select candidate users from the job applicants; S04: Index virtual indicators through job supply characteristics, construct recruitment scenarios that match the virtual indicators, and extract text data of candidate users in the recruitment scenarios; S05: Identify the soft capability features of text data through the LLM model, and transform the soft capability features into vectors to generate an intrinsic core vector; S06: Calculate the job value score through the intrinsic core vector, dynamically allocate the value weight factors of the job value score and job competency score, calculate the comprehensive score, and recommend candidate users in order.

Images